Skip to main content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

How Well Have Rural and Small Metropolitan Labor Markets Absorbed Welfare Recipients?

Publication Date
Mar 31, 2001

Prepared for:
Assistant Secretary for Planning and Evaluation
U.S. Department of Health and Human Services

Prepared by:
The Lewin Group: Mary Farrell, Selen Opcin, and Michael Fishman
Consultant: David Stapleton

"

Acknowledgments

Work on this project was funded by the Assistant Secretary for Planning and Evaluation (ASPE), Department of Health and Human Services (HHS) under a contract to The Lewin Group. This report has benefited greatly from the oversight and input of Laura Chadwick, the ASPE Project Officer, and her predecessor, Davy Norris.

The cooperation of staff from the Department of Labor and state staff in the regions have been critical to the completion of this report. From the Department of Labor, Bureau of Labor Statistics, Douglas Himes provided data from the Occupational Employment Statistics survey for the MSAs, Bernard Bell provided data from the ES-202 system for the MSAs, and Alan Eck provided the education and training codes. From the state labor agencies, William Niblack in Missouri, Jonathan Cole in New York, and Dwayne Stevenson in Oregon provided data from the ES-202 system for the non-MSA regions. From the welfare state agencies, Deborah Wood in Alabama, Richard Koon in Missouri, Chris Christmas in Mississippi, Jine Gleason in New York, Sue Johns in Oregon, Susan Banks in South Carolina, Harriet Drewery in Tennessee, Debra Tighe in Vermont, and Janet VanVleck and Beth Dorschner in Wisconsin provided AFDC/TANF caseload information.

At The Lewin Group, Stephanie Laud and Eunice Lee assisted with the data collection effort, Brian Simonson assisted with the analysis, and Darla Webb assisted with the report's production.

Acronyms And Abbreviations

AFDC
Aid to Families with Dependent Children
ARF
Area Resource File
ASPE
Assistant Secretary for Planning and Evaluation
BLS
Bureau of Labor Statistics
CEA
Council of Economic Advisors
CPS
Current Population Survey
DHHS
U.S. Department of Health and Human Services
EITC
Earned Income Tax Credit
ES-202
Covered Employment and Wages
HHS
U.S. Department of Health and Human Services
JOBS
Job Opportunity and Basic Skills Training Program
LD
Labor Demand
LS
Labor Supply
MSA
Metropolitan Statistical Area
NA
Not Applicable
NISP
National Industry Staffing Patterns
NSAF
National Survey of America's Families
OBRA
Omnibus Budget Reconciliation Act of 1981
OES
Occupational Employment Statistics
PRWORA
Personal Responsibility and Work Opportunity Reconciliation Act of 1996
SESA
State Employment Security Agency
SIC
Standard Industrial Classification
TANF
Temporary Assistance to Needy Families
USDA
United States Department of Agriculture

Obtaining a Printed Copy

To obtain a printed copy of this report, send or fax the title and your name and address to:

Human Services Policy, Rm 404E
Assistant Secretary for Planning and Evaluation
U.S. Department of Health and Human Services
200 Independence Ave, SW
Washington, DC 20201

Fax: (202) 690-6562

Chapter 1: Introduction

Purpose of Study

Since 1993, welfare recipients have been leaving the welfare rolls for work in record numbers. From January 1993 to January 1998, welfare caseloads declined by 33 percent nationally, and several studies have estimated that over half of the adults who have left welfare have entered the labor market.(7) The inflow of welfare recipients into the labor market can be attributed to two basic factors: welfare reform and the strong economy. Welfare reform is widely perceived to have begun with the passage of the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) in mid-1996, which increased the number of welfare recipients who were required to seek work. But even prior to this legislation, many states were reshaping their Aid to Families with Dependent Children (AFDC) programs under waivers, which likely increased the number of welfare recipients entering the labor force in at least some of these states. In addition, the strong economy from 1993 to 1998 increased the availability of low-skill jobs and undoubtedly lured many welfare recipients into the low-skill labor market.

Several other factors may have contributed to changes in labor market participation of welfare recipients and are worth mentioning. First, the federal government expanded the Earned Income Tax Credit (EITC) for working low-income families in the early- and mid-1990s, which most likely encouraged some welfare recipients to enter the labor force. Second, the minimum wage increased in 1997, which could have offset downward wage pressure from the entry of welfare recipients into the labor force. Third, some regions of the country experienced significant changes in population, which reduced or increased the number of low-skill workers in these areas. Finally, the recession of the early 1990s created a pool of unemployed low-skill workers who were available to take new jobs when the economy began to recover.

Policy-makers have been concerned about whether enough jobs will be available to employ the additional welfare recipients entering the labor market as a result of welfare reform. If a surplus of jobs is not available in particular areas, welfare recipients entry into the labor force might reduce low-skill wages and displace some workers. Policy-makers are especially concerned about the impact of welfare reform on rural and small metropolitan labor markets, because these markets might be less able to absorb the inflow of welfare recipients than urban labor markets.

The Assistant Secretary for Planning and Evaluation (ASPE) in the Department of Health and Human Services (DHHS) contracted with The Lewin Group to examine how well rural and small metropolitan labor markets can absorb welfare recipients, and to the extent feasible, estimate the impact of welfare reform on rural and small-metropolitan regions since 1993. This study uses an economic model to estimate the impact of welfare reform and improvements in the economy on the low-skill labor market, where most welfare recipients seek work.

A major challenge facing researchers in this area is to distinguish between entry due to reforms (welfare push) and entry due to the strong economy (demand pull). This decomposition is necessary if we are to anticipate future conditions in the low-skill labor market, when the economy might not be so strong. We attempted such a decomposition in this report.

We selected 12 regions for this study, which include a mix of rural and small metropolitan areas (the latter are defined as having fewer than 200,000 residents). The regions vary widely in population characteristics and labor market conditions and implemented a wide range of welfare policies. The study regions are listed below (Chapter 3 provides detailed information on each region).

Decatur and Florence, Alabama Central Oregon

Rural Mississippi

Medford-Ashland, Oregon

Joplin, Missouri

Florence, South Carolina

Southeast Missouri

State of Vermont

Jamestown, New York

Eau Claire, Wisconsin

North Country, New York

Wausau, Wisconsin

Welfare Reform

States began implementing significant reforms to their welfare programs in the past 10 years, many of which were designed to move welfare recipients into employment more rapidly. This section describes the types of welfare waivers approved between 1993 and 1996 and the major provisions of PRWORA.

1993 to 1996: Welfare Waivers

Prior to the enactment of PRWORA in mid-1996, the Family Support Act of 1988 (FSA) required a certain portion of AFDC recipients to participate in the Job Opportunity and Basic Skills Training (JOBS) program. States could apply for waivers to the AFDC/JOBS program to test new strategies. Waivers were granted to more than 40 states, many of which were statewide reforms. The waivers can be categorized by whether they were restrictive (i.e., made receipt more difficult or penalized welfare recipients for noncompliance) or liberalizing changes (made receipt easier or enabled recipients to combine work and welfare):(8)

Restrictive Changes

  • Time limits on benefits (24 states)
  • Tightened work requirements (31 states)
  • Benefits linked to school attendance or performance (26 states)
  • Limited benefits for additional children (family cap) (14 states)
  • Reduced benefits based on relocation (2 states)
  • Fingerprinting as condition of eligibility (1 state)

Liberalizing Changes

  • Enhanced earnings disregards (30 states)
  • Expanded eligibility for Unemployed Parent (UP or two-parent) families (25 states)
  • Increased resource limits (28 states)
  • Increased vehicle asset limit (25 states)
  • Expanded transitional medical and child care (21 states)

The Council of Economic Advisors (CEA) estimated that about 12 to 15 percent of the decline in welfare caseloads during the 1993 to 1996 period was due to waivers, particularly those that authorized more stringent JOBS sanctions.(9) Presumably many of those who left the rolls entered the labor market. In addition to increasing the number of welfare recipients who left the rolls, waivers that enhanced the earnings disregards and expanded eligibility for UP families allowed more individuals to combine work and welfare.

1996 to 1998: PRWORA

PRWORA ended the entitlement to individual families and gave states the flexibility to design their own programs, although attached conditions to the TANF block grant. Some of the more important conditions include:

  • Time limits. States must impose a five-year time limit on TANF-funded benefits, although they can exempt up to 20 percent of their caseloads. States can establish shorter time limits, without being penalized, or use state funds to provide assistance to welfare recipients after five years.
  • Work requirements. States must achieve minimum work participation rates. In 1997, states were required to have 25 percent of all families meeting the work participation requirement, increasing by 5 percentage points each year until reaching 50 percent in 2002. States can lower the work participation requirements by reducing their TANF caseload; the required work participation rate is reduced by the number of percentage points by which the states caseload is lower than its fiscal year 1995 caseload.
  • Sanctions. States are also required to sanction welfare recipients who refuse to meet work participation requirements. It is up to the states to determine the penalty. About 25 percent of states apply full-family sanctions; the remainder sanction adults only or reduce overall grant levels.

Growth of Economy

While economic conditions vary by region, the U.S. economy has had continued growth since 1993. From January 1993 to January 1999, employment increased by 18 million jobs nationally, the unemployment rate declined from 7.3 percent to 4.3 percent, and average inflation adjusted earnings increased from $12.41 per hour to $13.11 per hour.(10)

The additional jobs at higher hourly wages enticed some welfare recipients into the labor market. The CEA study estimated that between 1993 and 1996, about 26 to 36 percent of the welfare caseload decline was due to the economic expansion, and between 1996 and 1998, about 8 to 10 percent of the decline was due to the improved economy.(11) Several other studies have focused on the earlier pre-1996 period and found that economic factors explain somewhere between 25 and 50 percent of the change in caseloads.(12)

Other Factors

Factors other than the economy and welfare reform undoubtedly changed labor supply and demand. The EITC and increases in the minimum wage may have increased the incentive for work among low-income individuals, although the full extent of the effect of these two factors is largely unknown. Increasing the minimum wage could also result in a loss of jobs if employers are unwilling to pay the higher wage. In addition, changes in the population resulted in shifts of the supply curve. Finally, the recession in the early 1990s affected unemployment due to the fact that wages were slow to respond to the decline in the demand for labor. Each factor is discussed in more detail below.

EITC

The federal EITC, which was created in 1975, reduces taxes and supplements wages for working low  income taxpayers. The Omnibus Budget Reconciliation Act (OBRA) of 1992 increased the EITC by over 50 percent, phased in over three years, effective in 1994.(13) Previously, the OBRA of 1990 expanded the EITC (beginning in 1991) and also excluded the EITC in determining eligibility or benefit amounts for most federal means-tested programs (AFDC, Medicaid, Supplemental Security Income, food stamps, and low-income housing). In 1993, about 15 million taxpayers claimed the federal tax credit; by 1998, just under 19 million did so.(14)

In addition, 14 states(15) and the District of Columbia offer state EITCs that supplement the federal credit. These states use the federal eligibility rules, and offer an additional state credit that is a percentage of the federal credit. New Jersey limits the credit to families with incomes below $20,000.(16)

We believe the expansion primarily affected labor markets in the 1993 to 1996 period. A recent study separated the effect of EITC expansions from the effect of welfare policies and local labor market conditions, and found that EITC expansions have a large positive effect on employment of adults from welfare families.(17)

Increase in Minimum Wage

The federal minimum wage was $4.25 an hour from 1991 to 1997. In 1997, it increased to $5.15, where it has remained. In addition, some states have imposed minimum wages that exceed the federal rate. Among our study states, these include Oregon ($6.50 per hour) and Vermont ($5.75 an hour).

Economists have long debated the effect of minimum wage laws on employment. Some argue that it results in job loss; others argue that the minimum wage was well below the equilibrium rate in 1997, and increasing it had no effect on labor demand.

Population Growth

Growth in population shifts the labor supply curve outward, while a decline in population shifts the labor supply curve inward. Nationally, the number of individuals age 16 and over has been increasing about 1 percent annually, from 198 million in 1993, to 204 million in 1996, and to 208 million in 1998.

Effect of Recession

In 1993, the economy had just started to recover from a recession. In the presence of downward wage rigidity, we have reason to believe that the economy was not at equilibrium. The recession created an excess supply of individuals who were willing to work at the prevailing wage, but who were unemployed because wages did not adjust downward. When the economy began to recover, employers began to demand more labor. The presence of unemployed workers in the low-skill labor market would have allowed an increase in employment without an increase in wages.

Past research

A number of studies attempted to measure the impact of welfare reform on labor markets. Most studies focused on employment effects of welfare reform. These studies projected the number of welfare recipients who would leave welfare to enter the labor force and the number of low-skill job openings that would be available to welfare recipients. If there were more low-skill job openings than welfare recipients entering the labor market, then the conclusion was that the labor market would be able to absorb welfare recipients.

A study by Lerman, Loprest, and Ratcliffe(18) compared the growth of low-skill employment with the number of welfare recipients who entered the labor force between 1997 and 1998 in 20 metropolitan areas across the country. They found that, due to the growth in low-skill employment, only four metropolitan areas (Baltimore, New York City, St. Louis, and the District of Columbia) may have experienced an increase in unemployment due to welfare reform. A similar study by Leete and Bania(19) for the Cleveland-Akron metropolitan area found that welfare recipients would have to claim 34 to 61 percent of all low-skill job openings to be fully employed in the initial year of impact, meaning there would be enough jobs for welfare recipients. Most studies concentrated on large metropolitan areas, primarily because of data availability. Few studies to date have examined the impact of welfare reform on labor markets in rural or small metropolitan areas.

Past studies noted that low-skill wages might have to fall to employ all welfare recipients entering the labor market and that the fall in wages might cause other low-skill workers to lose their jobs (be displaced). Bartik(20) studied the wage and displacement effects of welfare reform using a model of labor demand and labor supply and elasticity assumptions. Based on policy simulations, this study predicted that wages would fall for workers with characteristics similar to welfare recipients. Despite theoretical discussions, we know of no study that has measured the actual impact of welfare reform on wages.

Study Overview

Basic Approach

Our approach began by comparing estimates of the number of welfare recipients entering the labor market to changes in the low-skill labor force in each of the regions. Were there enough jobs to absorb the increase of welfare recipients who left for work or who combined welfare and work?

We then attempted to estimate the effect of the welfare push on employment and wages in each of the regions. As discussed above, distinguishing between the effects of welfare push and demand pull is difficult for several reasons. One is that, for the most part, the economies of our 12 study regions still had not recovered fully from the 1991 recession (e.g., they had historically high unemployment rates in 1993). Hence, they could be reasonably characterized as having an excess supply of labor at going wage rates, which makes application of the demand/supply analysis problematic. In addition, factors other than welfare reform and economic expansion had impacts on some of these areas low-skill labor markets during this period. As noted above, three such factors are the expansion of the EITC, population growth, and increases in the minimum wage.

Contents of the Report

The report is organized as follows:

  • Chapter 2 discusses the unique nature of small and nonmetropolitan labor markets and some of the barriers to employment that exist in these areas.
  • Chapter 3 provides a description of the 12 regions selected for this study, including information on the characteristics of the population, the labor market conditions, the welfare policies, and the declines in the welfare caseloads.
  • Chapter 4 provides a detailed description of the methodology used for estimating low-skill employment and the impact of welfare reform and the economy on employment and wages.
  • Chapter 5 presents the findings from the analysis, projects the effect of a recession with and without welfare reform, and concludes with the implications for future research.

 

Endnotes

(7) From P. Loprest and S. Brauner (1999). Where Are They Now? What States Studies of People Who Left Welfare Tell Us. The Urban Institute. Washington, DC: A survey of state leaver studies found that employment rates ranged from 55 to 75 percent for continuous leavers who were surveyed at a point in time after leaving.

(8) 1998 Green Book. Committee on Ways and Means of the U.S. House of Representatives. Washington, DC.

(9) Council of Economic Advisors (1999). The Effects of Welfare Policy and the Economic Expansion on Welfare Caseloads: An Update. Washington, DC. This revises estimates from a 1997 CEA report that attributed about 30 percent of the caseload decline to welfare waivers.

(10)U.S. Department of Labor, Bureau of Labor Statistics. Employment is total non-farm payroll employment (seasonally adjusted); unemployment is the adjusted unemployment rate for the civilian labor force; and hourly rates are average hourly earnings of production workers (not seasonally adjusted).

(11) Council of Economic Advisors (1999).

(12) Blank, R (2000). Declining Caseloads/Increased Work: What Can We Conclude About the Effects of Welfare Reform? Paper prepared for the conference Welfare Reform Four Years Later: Progress and Prospects at the Federal Reserve Bank of New York.

(13) Meyer, B. and D. Rosenbaum (2000). Making Single Mothers Work: Recent Tax and Welfare Policy and Its Effects. National Bureau of Economic Research Working Paper 7491. Cambridge, MA.

(14) 1998 Green Book.

(15) Colorado, Illinois, Iowa, Kansas, Maine, Maryland, Massachusetts, Minnesota, New Jersey, New York, Oregon, Rhode Island, Vermont, and Wisconsin.

(16) Johnson, N. (2000). A Hand Up. How State Earned Income Tax Credits Help Working Families Escape Poverty in 2000: An Overview. Center on Budget and Policy Priorities. Washington, DC.

(17) Hotz, J. H., Mullin, C. H., and Scholz, J. K. (2000). The Earned Income Tax Credit and Labor Market Participation of Families on Welfare. Paper for the Joint Center for Poverty Research Conference on Means-Tested Transfers, December 7-8, 2000.

(18) Lerman, R. I., P. Loprest, and C. Ratcliffe (1999). How Well Can Urban Labor Markets Absorb Welfare Recipients? The Urban Institute New Federalism Number A-33. Washington DC.

(19) Leete, L. and N. Bania (1999). The Impact of Welfare Reform on Local Labor Markets. Journal of Policy Analysis and Management (18) 1.

(20) Bartik, T. (1999). Will Welfare Reform Cause Displacement? W.E. Upjohn Institute for Employment Research. Kalamazoo, MI.

Chapter 2: Rural And Small Metropolitan Labor Markets

As discussed in Chapter 1, most studies that have measured the impact of welfare reform on labor markets have focused on urban areas, and little is known about the effect on rural areas. This is because data are more readily available for large urban areas and a larger share of welfare recipients live in metropolitan areas (81 percent in 1997).(21) However, substantial differences exist between urban and rural labor markets in terms of the economic opportunities, the characteristics of the workforce, and the barriers to employment, suggesting that findings from urban studies might not readily apply to rural areas.

States emphasis on moving welfare recipients into jobs quickly might prove more difficult in rural settings for several reasons. First, there is some evidence that rural areas offer fewer economic opportunities. Second, individuals living in rural areas have lower education levels, on average, increasing the challenges for finding employment. Finally, job search, as well as educational, childcare, and transportation services, may be less available in rural areas to help welfare recipients find employment, obtain skills required for employment, and accept employment. This chapter discusses these obstacles in more detail.

Characteristics of Jobs

PRWORA stressed employment for all job-ready welfare recipients, which presumes that economic opportunities are available for those who seek them. Economic opportunities can mean the availability of jobs. Also important to consider are the types and quality (e.g., wages) of employment opportunities. If there are fewer opportunities available to rural welfare recipients, then PRWORA might have more negative consequences in these areas.

Lower Wages Across Industries

There is substantial evidence that metropolitan areas offer higher wages and salaries than nonmetropolitan areas.(22) The U.S. Department of Labor, Bureau of Labor Statistics (BLS) National Compensation Survey, collected wage and salary information from establishments with 50 or more workers in goods-producing and service industries, and state and local governments.(23) Analysis from this survey found that in 1997:

  • Metropolitan area workers overall earned an average of $15.73 per hour compared to $11.84 earned by workers in nonmetropolitan areas.
  • In private goods-producing industries, metropolitan area workers averaged $16.40 an hour, compared to $12.06 for similar workers in nonmetropolitan areas.
  • In the service industries, metropolitan workers averaged $14.44 an hour, compared to an hourly rate of $9.77 for workers in nonmetropolitan areas.
  • In each of the nine census regions, metropolitan areas paid higher wages.

In addition, the gap between metropolitan and nonmetropolitan wages may be growing. Between 1998 and 1999, real median income for households residing in metropolitan areas increased by 2.1 percent (for households in central cities, income rose by 5.0 percent). However, the median income of households outside metropolitan areas remained statistically unchanged.(24)

Industry and Occupation Mix

In rural areas, manufacturing jobs have traditionally paid higher salaries than other industries. While manufacturing jobs are more prevalent in rural areas, jobs in manufacturing have been declining over time. One study reported that between 1969 and 1992, rural manufacturing employment fell from 20 percent to 17 percent of total employment.(25) This has been accompanied by an increase in employment in the service industry, which traditionally pays lower wages, especially in nonmetropolitan areas (as discussed above).

One study examined the occupations in metropolitan and nonmetropolitan areas and found that about 54 percent of the nonmetropolitan labor force was employed in: service; agricultural, precision production, craft, and repair; or operators, fabricator, and laborer occupations. These are occupations where employees are likely to be paid hourly wages rather than salaries, and employees typically incur reductions in hours or layoffs when demand is slack. Consequently, nonmetropolitan labor forces tend to respond more quickly to business cycle movements than the metropolitan ones.(26)

Characteristics of workforce

On average, adult residents in rural areas have lower levels of educational attainment than more urban residents. According to the Census Bureaus Current Population Survey (CPS), 20 percent of rural residents 18 years and older did not have a high school diploma compared with 17 percent in metropolitan counties (see Exhibit 2.1). A significantly smaller percentage of rural residents had higher levels of education. Interestingly, this was not true when we limited the sample to welfare recipients in nonmetropolitan and metropolitan areas. Welfare recipients living in rural areas were slightly more likely to have obtained a high school diploma than welfare recipients living in large metropolitan areas. Thus, there is a smaller gap in education levels between welfare recipients and non-welfare recipients in rural areas than in urban areas, implying there might be more competition for low-skill jobs.

Exhibit 2.1
Education Level of Adults and Adult Welfare Recipients by Size of Geographic Area
Education Level Nonmetropolitan/ Not Identified (%) MSA: 100,000  250,000 (%) MSA: 250,000+ (%)
Total Population
No high school degree diploma 20.4 16.7 16.6
High school degree 38.9 32.7 31.3

Above high school

40.6 50.6 52.2

Total

100.0 100.0 100.0
Welfare Recipients

No high school degree diploma

33.9 32.6 40.0

High school degree

41.9 44.3 38.0

Above high school

24.3 23.1 22.0

Total

100.0 100.0 100.0

Source: Lewin calculations using the 1999 CPS March Supplement.

In addition, a research synthesis on the rural welfare population found the following:(27)

  • Rural welfare recipients were slightly more likely to be married and more likely to work than central city counterparts.
  • Spells on public assistance were shorter in rural areas than in urban areas.
  • In rural areas, the dollar amount of grants received by welfare families was significantly less than the dollar amount received by urban families. States with large rural populations tended to offer lower benefit grants.

Barriers to employment

Rural workers may face greater barriers to employment than urban workers. These include the following:

  • Child Care. Rural areas have fewer child care slots than urban areas and workers must travel greater distances from home to obtain child care. As a result, rural families depend more on child care by relatives and friends.(28)
  • Transportation. Public transportation is less available in rural areas and distances traveled between home and work are also longer in rural areas. Nationally, 40 percent of all rural residents live in areas with no form of public transportation.(29) Thus, access to personal transportation is critical.
  • Education and Training Services. To achieve self-sufficiency, welfare recipients may need access to additional education and training services that are more readily available in urban areas. In addition, they may not have the same level of access to job training centers.

 

Endnotes

(21)Current Population Survey 1998, US Census Bureau. These estimates are calculated from the number of respondents age 15 to 65 who indicated they received welfare in 1997.

(22)Unless otherwise noted, we used the term rural to mean nonmetropolitan, for simplicity, although these are not synonymous. The Office of Management and Budget defines nonmetropolitan counties as being outside the boundaries of metro areas and having no cities with as many as 50,000 residents. Metropolitan areas contain (1) core counties with one or more central cities of at least 50,000 residents or with a Census Bureau-defined urbanized area (and a total metro area population of 100,000 or more), and (2) fringe counties that are economically tied to the core counties. According to official federal definitions, rural areas comprise places (incorporated or unincorporated) with fewer than 2,500 residents and open territory. Urban areas comprise larger places and densely settled areas around them.

(23) This survey integrated the Occupational Compensation Survey Program (OCSP) with the Employment Cost Index and the Employee Benefit Survey. Results are reported in: U.S. Department of Labor (1999a). When it Comes to Pay, Does Location Matter?, Compensation and Working Conditions Online. Summer 2000, Vol 5, No. 2.

(24) U.S. Department of Commerce (2000). Money Income in the United States: 1999. Bureau of the Census, Current Population Reports, September.

(25) Parker, Timothy (1995). Understanding Rural America. Agriculture Information BulletinNo. 710. Washington, DC: USDA.

(26) Hamrick, Karen S. (1997). Rural Labor Markets Often Lead Urban Markets in Recessions and Expansions. Rural Development Perspectives, vol. 12, no. 3. Washington, DC: USDA.

(27) Marks, E., S. Dewees, T. Ouellette, R. Koralek (1999). Rural Welfare to Work Strategies: Research Synthesis. Macro International, Inc. Washington, DC.

(28) Rural Policy Research Institute (1999). Rural America and Welfare Reform: An Overview Assessment. University of Missouri, Columbia, MO.

(29) Dewees, S. (1998). The Drive to Work: Transportation Issues and Welfare Reform in Rural Areas. Southern Rural Development Center. Mississippi State, MS.

Chapter 3: Study Regions

Characteristics of population

Exhibit 3.3 summarizes the demographic characteristics of the population in each of the study regions, as well as the U.S. population for comparison purposes. As Exhibit 3.3 shows, the majority of residents in all sites are white, although the southern regions have a higher share of blacks. The southern states also have a relatively high share of adults without a high school degree (from 16 to 20 percent), although Southeast Missouri has the highest percent of high school dropouts (25 percent).

Exhibit 3.3
Characteristics of Population, Various Years
Region Population (in thousands) (1996) Rural Population (%) (1990) White (%) (1996) Black (%) (1996) No High School Degree (%) (1990) College Degree (%) (1990)

Decatur and Florence, Alabama

276 48.3 85.7 12.5 15.5 12.4

Rural Mississippi

1,952 62.0 61.0 37.7 19.6 11.1

Joplin, Missouri

146 40.8 95.5 1.1 10.6 12.8

Southeast Missouri

524 47.6 93.4 5.4 24.5 7.8

Jamestown, New York

141 47.6 93.5 2.2 8.8 14.2

North Country, New York

429 65.8 91.5 4.5 9.9 14.0

Medford-Ashland, Oregon

169 34.8 91.8 0.3 6.2 17.6

Central Oregon

131 64.5 91.8 0.2 7.5 13.7

Florence, South Carolina

123 47.7 59.5 39.6 15.7 14.8

Vermont

545 67.9 97.3 0.6 9.8 21.9

Eau Claire, Wisconsin

143 39.3 96.5 0.2 11.0 15.9

Wausau, Wisconsin

122 43.7 96.2 0.1 14.1 13.5

United States

265,284 24.8 83.5 8.9 14.5 13.4

Source: Lewin tabulations using the 2000 Area Resource File.

Labor Market Conditions

As Exhibit 3.4 shows, in three study regions  Rural Mississippi, Southeast Missouri, and Florence, South Carolina  over 20 percent of the residents were living in poverty in 1995. Correspondingly, these three regions have relatively low median household incomes.

Unemployment rates began increasing in most regions sometime after 1989 or 1990 and peaked by 1993 or 1994. In 1993, the unemployment rate reached 10 percent in Southeast Missouri, followed by North Country, New York (9.2 percent), and Central Oregon and Florence, South Carolina (both 8.7 percent). By 1996, unemployment rates had dropped in all regions except Rural Mississippi and the Oregon regions, which did not see a drop in unemployment until 1997. By 1998, seven regions had unemployment rates that were lower than the national average (5.3 percent) and two had rates near the national average (5.5 and 5.7 percent).

Exhibit 3.4 Economic Characteristics
Region Living in Poverty (%) (1995) Median household income ($) (1995) Unemployment Rate (%)
1989 1993 1996 1998

Decatur and Florence, Alabama

13.6 31,584 8.3 7.8 5.7 5.5

Rural Mississippi

22.9 22,861 8.3 6.9 6.9 6.2

Joplin, Missouri

15.0 28,705 5.5 6.2 4.3 4.0

Southeast Missouri

20.5 23,414 7.5 9.9 7.1 5.7

Jamestown, New York

15.8 29,568 6.1 6.8 5.2 5.2

North Country, New York

14.6 29,714 8.8 9.2 8.0 7.9

Medford-Ashland, Oregon

14.7 31,537 6.8 8.4 8.3 6.8

Central Oregon

12.0 33,373 6.6 8.7 8.8 7.1

Florence, South Carolina

20.5 29,541 4.7 8.7 8.4 4.5

Vermont

11.0 33,365 3.5 5.2 4.5 3.4

Eau Claire, Wisconsin (MSA)

10.4 33,538 4.3 5.6 3.7 3.3

Wausau, Wisconsin

7.5 40,078 4.4 5.1 3.9 3.5

United States

13.9 30,272 6.3 6.9 6.0 5.3

Source: Lewin tabulations using the 2000 Area Resource File.

Exhibit 3.5 presents the share of all jobs by major industry for the study regions and the U.S. As this exhibit shows, overall, the regions had a higher share of manufacturing jobs and a lower share of service jobs than the U.S. As discussed in Chapter 2, rural areas had a higher share of manufacturing employment than urban areas, although jobs in this industry continued to decline.

As Exhibit 3.5 also shows, there was some variation across the study regions:

  • Manufacturing jobs were well-represented in Wausau, Wisconsin; Decatur and Florence, Alabama; Jamestown, New York; and Joplin, Missouri (about 25 percent of jobs or higher).
  • Eau Claire had a relatively high share of jobs in the retail trade industry  Eau Claire Countys largest employer was a building materials dealer, which provided over 1,000 jobs.
  • Jobs in the transportation industry were prevalent in Joplin, Missouri.
  • Interestingly, while Vermont was one of the most rural regions, the share of jobs in each industry closely mirrored the U.S.

Finally, it is important to note that we selected regions where the predominant industry was not agriculture because of data limitations; specifically, the Occupational Employment Statistics (OES) survey, which is discussed in Chapter 4 and is used in our analysis, surveys only non-farm establishments.

Exhibit 3.5
Share of Employment in Major Industry By Region, 1998
  Decatur and Florence, AL (%) Rural MS (%) Joplin, MO (%) Southeast MO (%) Jamestown, NY (%) North Country, NY (%) Medford-Ashland, OR (%)

Agriculture, Forestry, and Fishing

0.8 2.0 0.8 1.6 1.0 1.0 3.2

Mining

1.1 0.7 0.3 0.9 0.3 0.4 0.1

Construction

6.5 4.8 3.5 4.5 2.7 3.7 5.0

Manufacturing

26.1 26.2 24.5 22.2 24.6 13.2 13.2

Transportation, Communications, Electric, Gas, and Sanitary Services

3.3 4.3 12.8 5.7 4.7 4.6 5.0

Wholesale Trade

5.4 4.1 3.8 5.1 3.5 2.9 4.1

Retail Trade

19.6 17.7 19.7 19.1 18.5 20.4 25.8

Finance, Insurance, and Real Estate

3.8 3.0 2.7 3.2 2.4 2.9 3.7

Services

26.8 31.6 28.3 32.2 35.1 37.1 34.4

Public Administration

6.8 5.7 3.5 5.6 7.1 13.7 5.4

Nonclassifiable

0.0 0.0 0.0 0.0 0.1 0.1 0.1

Total Employment

109,875 1,117,066 74,921 186,400 56,290 437,426 69,382

  Central Oregon (%) Florence, SC (%) Vermont (%) Eau Claire, WI (%) Wausau, WI (%) U.S. (%)

Agriculture, Forestry, and Fishing

3.1 0.6 1.2 0.5 1.2 1.5

Mining

0.1 0.1 0.2 0.1 0.2 0.5

Construction

7.3 5.1 5.4 4.5 4.4 4.9

Manufacturing

15.4 19.1 16.9 18.2 29.7 15.2

Transportation, Communications, Electric, Gas, and Sanitary Services

3.8 3.9 4.5 4.9 4.7 5.5

Wholesale Trade

4.7 4.5 4.4 3.8 7.5 5.5

Retail Trade

21.9 18.8 18.9 27.1 17.6 18.0

Finance, Insurance, and Real Estate

5.5 7.8 4.3 3.4 7.7 5.9

Services

31.6 33.1 38.4 31.8 22.4 36.7

Public Administration

5.1 6.8 5.9 5.8 4.6 6.1

Nonclassifiable

1.6 0.1 0.0 0.0 0.0 0.1

Total Employment

56,840 62,363 280,288 73,217 63,725 124,183,551

Source: Lewin tabulations using ES-202 data.

Welfare Policies

Waivers Operating in Study Regions

As mentioned in Chapter 1, prior to the passage of PRWORA, many states were operating their AFDC programs under waivers granted to them by DHHS to test innovative approaches. Waivers may have reduced the size of caseloads and increased the number of former and current welfare recipients entering the labor force between 1993 and 1996.

Waivers were granted to all states in our study except Alabama, although in New York and South Carolina, the waivers affected counties outside of the study regions. Missouri, Oregon, and Wisconsin approved statewide waivers in 1995 or 1996.(31) Mississippi and Vermont implemented waivers earlier, and as a result, waivers might have had a greater impact on declines in welfare caseloads in these states. Exhibit 3.6 describes the waivers and lists the dates of approval.

Exhibit 3.6 Waivers Covering the Study Regions
State Date Approved Description of Waiver
Mississippi

12/94

WorkFirst provided subsidized, private-sector employment for job-ready participants. It also provided for an Individual Development Account for family savings, to which employers contributed one dollar per hour of work. WorkFirst was implemented in six counties (four of which are included in the Rural Mississippi study region).
Missouri

4/95

The Mutual Responsibility Plan required AFDC recipients to sign and fulfill a self-sufficiency agreement that established a plan for work and placed a two-year time limit on benefits (with a two-year extension possible). At the end of the time limit, clients were required to participate in job search or work experience.
Oregon

3/96

Oregon Option limited AFDC to 24 months of benefits in any 7-year period, and required nearly all recipients to participate. Eligible participants were provided with subsidized public or private employment for up to nine months at minimum wage or better. Oregon also extended child care eligibility an additional 12 months for recipients who got jobs.
Vermont

4/93

The Family Independence Project enabled AFDC recipients to retain more income and accumulate more assets than is normally allowed. It also required AFDC recipients to work in a wage-paying job or participate in a community or public service job after they had received AFDC for 30 months (15 months for unemployed-parent families).
Wisconsin

6/96

Pay for Performance required AFDC applicants to meet with a financial planning resource specialist to explore alternatives to welfare. Those who still planned to apply were required to complete 60 hours of JOBS activities prior to approval. After approval, recipients were required to participate in JOBS activities for up to 40 hours per week.
Source: DHHS, ACF (1996). State Welfare Demonstrations. Washington, DC. Available on-line at: www.hhs.gov/news/press/1996pres/961007b.html.

After PRWORA

While variation in state AFDC programs existed prior to 1996, PRWORA gave states even more flexibility in designing their programs. Policies that could affect the number of welfare recipients moving to enter the labor market include the grant level, earnings disregards, time limits, work requirements, exemptions to the work requirements, and sanctions.

We selected regions that offered a range of these policies (see Exhibit 3.6 for a summary of the policies in the study regions). While we were not able to individually assess the impact of each policy on labor markets, we believe that variation in the caseload declines and the number of welfare recipients who were combining work and welfare partly reflected the broad range of state policies.

States establish the monthly grant, which depends on family size, family income, earnings disregards, and other factors. The size of the grant may affect an individuals willingness to forgo welfare benefits for work. As Exhibit 3.7 shows, the maximum grant levels for a family of three in 1998 varied dramatically from a low of $120 in Mississippi to a high of $673 in Wisconsin. While Mississippi had a very low grant level, it had one of the more generous earnings disregards policies. For the first six months, 100 percent of earnings could be disregarded, although individuals must obtain full-time work within 30 days of receiving welfare or participating in a work activity, such as job search. Conversely, Wisconsin had a high grant level, but did not disregard any earnings.

PRWORA established a 60-month federal time limit. Some states have imposed shorter time limits, while other states have agreed to provide assistance to families after they have reached the federal time limit. Among the study states, all except Oregon, New York, South Carolina, and Vermont have implemented the 60-month time limit. Oregon limited welfare benefits to 24 of the first 84 months; New York continued to provide assistance to families, partly in cash and partly in non-cash benefits after 60 months; South Carolina limited cash assistance to 24 of 120 months, imposing a lifetime limit of 60 months; and Vermont adopted a work trigger time limit, meaning that after 30 months (for single-parent households), families were required to work but could continue to receive assistance. We did not expect these time limits to greatly affect the caseloads as of 1998 because few recipients would have reached even their 24-month time limit.

Tough sanctions may have increased the number of welfare recipients who left welfare. Mississippi, South Carolina, and Tennessee imposed a full-family sanction, meaning the grant was terminated if the individual did not comply with program work requirements. Wisconsin reduced benefits for every hour of participation missed. Other states reduced the grant level by a portion until the individuals complied.

  Exhibit 3.7 Welfare Policies in Study States
  Maximum Grant for Family of Three (12/98) Earnings Disregards (as of 12/98) Time Limit Weekly Work Requirement for Single-Parent Families (as of 10/99) Exemptions to Work Requirement (as of 10/99) Sanction for Work Noncompliance (as of 4/00)

Alabama

$164 100% for 3 months; 20% in subsequent months 60 months (lifetime) 32-35 hours No exemption criteria First sanction: 25% until compliance; maximum sanction: termination for 6 months

Mississippi

$120 100% for 6 months if families obtain full-time work within 30 days of initial TANF receipt or participating in work activities; $90 in other months. 60 months (lifetime) 30 hours Caring for young child up to 1 year; disabled/temporary illness; caring for disabled household member; 60+years old; domestic violence victim; child care unavailable up to 6 years; pregnant (in third trimester); transportation unavailable. First sanction: termination for at least 2 months until compliance; maximum sanction: termination (permanent)

Missouri

$292 $120 and 1/3 of the remainder for 4 months; $120 for the next 8 months; $90 in subsequent months. 60 months (lifetime) 30 hours Caring for young child (up to 1 year); disabled/temporary illness; caring for disabled household member; 60+ years old; domestic violence victim; child care unavailable up to 6 years; pregnant (in third trimester); living in remote area First sanction: 25% until compliance; maximum sanction: 25% until compliance for at least 3 months

New York

$577 $90 and 45% of the remainder 60 months (safety net assistance is paid to family after time limit is reached) At county discretion Caring for young child up to 3 mths-1 year (at county's discretion); disabled/temporary illness; caring for disabled household member; 60+ years old; pregnant (in ninth month) First sanction: pro-rata reduction until compliance; maximum sanction: pro-rate reduction until compliance for at least 6 months

Oregon

$503 50% 24 in 84 months No Caring for young child up to 3 mths; 60+ years old; pregnant (in ninth month) First sanction: $50; maximum sanction: termination until compliance

South Carolina

$201 50% for 4 months; $100 in subsequent months 24 in 120 months; 60 months (lifetime) 30 hours Caring for young child (up to 1 year); disabled/temporary illness; caring for disabled household member; child care unavailable up to 12 years; transportation unavailable All sanctions: termination until compliance for 30 days

Vermont

$617 $150 and 25% of the remainder Not applicable a/ After reaching time limit only: 20 hours if youngest child is under age 13; 40 hours if 13 or older Caring for young child up to 3 years; disabled/temporary illness; caring for disabled household member; 60+years old; pregnant (at least in fourth month) First sanction: adult portion until compliance; maximum sanction: adult portion until compliance for at least 6 months

Wisconsin

$673 None 60 months (lifetime) At local discretion Caring for young child up to 12 weeks First sanction: pay-for-performance (per hour reduction); maximum sanction: termination (permanent)

Source: State Policy Documentation Project (www.spdp.org/tanf.htm)
a/ Vermont imposed a work-trigger time limit; after 30 months of assistance, most TANF recipients must enroll in a work activity, but their benefits are not reduced or canceled. Vermont was operating under a statewide waiver.

Welfare Caseloads

Exhibit 3.8 presents the average monthly caseload in each region at three points in time: 1993, 1996, and 1998. It also depicts the caseload as a percent of the civilian labor force, age 16 and over. As might be expected, while all regions experienced a reduction in caseloads between the three points in time, there was substantial variation in the size of the reduction across the regions (see Exhibit 3.9). In particular, the Wisconsin regions saw the most dramatic declines, while Vermont experienced relatively small reductions.

Exhibit 3.8 AFDC/TANF Caseloads in Study Regions
Region 1993 1996 1998
Monthly Caseload As Percent of Adult Civilian Labor Force Monthly Caseload As Percent of Adult Civilian Labor Force Monthly Caseload As Percent of Adult Civilian Labor Force

Decatur and Florence, Alabama

1,577 1.2 1,167 0.8 645 0.4

Rural Mississippi

45,384 5.3 36,565 4.2 19,096 2.2

Joplin, Missouri

2,081 2.9 1,906 2.4 1,271 1.6

Southeast Missouri

12,674 5.6 10,817 4.4 7,972 3.3

Jamestown, New York

3,154 4.6 2,516 3.7 1,975 2.9

North Country, New York

6,656 3.5 5,749 3.1 4,145 2.2

Medford-Ashland, Oregon

2,540 3.2 1,820 2.1 896 1.0

Central Oregon

1,342 2.1 1,026 1.5 635 0.9

Florence, South Carolina

2,619 4.3 2,469 4.1 1,665 2.6

Vermont

10,081 3.4 9,210 3.1 7,591 2.5

Eau Claire, Wisconsin

2,037 2.8 1,116 1.4 302 0.4

Wausau, Wisconsin

1,162 1.7 785 1.1 234 0.3

United States

4,963,000 3.8 4,628,000 3.4 3,305,000 2.4

Source: Lewin tabulations using the 2000 Area Resource File and information provided by state welfare agencies.

As might be expected, due to the implementation of PRWORA, caseload declines between 1996 and 1998 exceeded the declines between 1993 and 1996. Why the dramatic decline in welfare caseloads in Wisconsin? Partly, as we will see in Chapter 5, the strong labor market conditions contributed to the decline. Also, Wisconsin adopted one of the most stringent work requirements, requiring all who were job ready to either find unsubsidized work, take a trial job (with subsidies provided to the employer), or take a community service job. Vermont, on the other hand, provided relatively high grants, had a relatively generous earnings disregard, meaning individuals could combine work and welfare, and imposed no time limit on benefits (but does require individuals to work after two years).

Exhibit 3.9 Percent Reductions in AFDC/TANF Caseloads

Percent Reductions in AFDC/TANF Caseloads

Endnotes

(30) Jackson, Tennessee was an initial study region, but was excluded due to data limitations. This is discussed in Appendix F. [Back To Text]

(31) Oregon and Wisconsin had earlier waivers in place in selected counties outside the study regions. [Back To Text]

Chapter 4: Methodology

This chapter first describes the economic model used to measure the possible effect of welfare reform and the improved economy on low-skill employment and wages. Then it describes the data used in the analysis, assumptions about the elasticity of labor demand and supply, and estimates of welfare recipients in the labor force.

Economic Model

The economic model presented here provides a framework for studying the entry of large numbers of welfare recipients into the low-skill labor market as a result of (1) the pull of the economy, and (2) the push of welfare reform.

The pull of the economy refers to an increase in the labor demanded by firms at any given wage level, and the entry of welfare recipients into the low-skill labor market in response to better job opportunities. These welfare recipients would have entered the labor force even in the absence of welfare reform. The push of welfare reform refers to an increase in the number of persons willing to work at any given wage level as a direct result of welfare reform, and the consequent entry of welfare recipients into the labor market. These welfare recipients would not have entered the labor force in the absence of welfare reform.

The discussion of the model in the next section is divided into four subsections. Subsection A discusses the economic model. Subsection B stresses the importance of elasticities on the effect of welfare reform. Subsection C discusses unemployment. Subsection D discusses the effect of downward wage rigidities on unemployment.

Assessment of the Impact of Welfare Reform and Economic Expansion

The pull of the economy is modeled as a demand shift  an increase in the demand for low-skill labor at any wage level (i.e., an outward shift in the demand curve). The push of welfare reform is modeled as a supply shift  an increase in the supply of low-skill labor at any wage level (i.e., an outward shift in the supply curve). Both shifts are positive, i.e., they involve an increase in demand and an increase in supply at every wage.

The shifts can be illustrated on a standard labor demand/labor supply diagram (Exhibit 4.1). LS0 is the labor supply curve before welfare reform, and LS1 is the labor supply curve after welfare reform. LD0 is the labor demand curve before economic expansion, and LD1 is the labor demand curve after economic expansion.

The demand shift from LD0 to LD1 and the movement along LS0 from Point A to Point B represent the pull of the economy. It is important to note that the pull of the economy does not shift the supply curve. Employment increases from E0 to EB. The number of welfare recipients pulled into employment is less than this amount, because some who are pulled into employment by the expansion are not welfare recipients.

Exhibit 4.1
Demand and Supply for Low-Skill Labor

Demand and Supply for Low-Skill Labor

Symmetrically, the supply curve shift from LS0 to LS1 and the movement along LD1 from Point B to Point C represents the push of welfare reform. As a result, employment increases from EB to E1. As drawn, the new equilibrium is at a lower wage than the initial equilibrium  the upward pressure of economic expansion on wages is more than offset by the downward pressure of welfare reform. Under such a scenario, some individuals who are working in the initial equilibrium will not be willing to work in the new equilibrium, because of the lower wage. These workers are displaced by welfare reform. In the diagram, they are represented by the horizontal distance from EV to E0. The number of new workers is represented by the distance from EV to E1, and is exactly equal to the size of the shift in the supply curve.

It is important to keep in mind that the supply curve represents the supply of workers from two populations  adults who are in the group targeted by welfare programs and all other low-skill, working-age adults. Thus, the increased employment due to the pull of the economy that is represented in Exhibit 4.1 exceeds the number of welfare recipients who are drawn into employment by the economic expansion. Similarly, the workers displaced by welfare reform in Exhibit 4.1 might include some who are in the target group for welfare programs, and some of these might even enter welfare as a result.(32)

Points A and C represent initial and final equilibrium outcomes, before and after welfare reform (supply shift) and economic expansion (demand shift). The percentage changes in employment and wages between these two points can be expressed formally by the following equations, which show the effects of the demand and supply shifts on wages and employment.(33)

(i) % D employment = ( es * % D demand + ed * % D supply) / ( ed + es)

(ii) % D wage = (% D demand - % D supply) / ( ed + es)

where:

  • ed is the absolute value of the elasticity of demand  the absolute value of the percentage change in employment associated with a one percent increase in the wage rate along the demand curve (the value is positive);
  • es is the elasticity of supply  the percentage change in employment associated with a one percent increase in the wage rate along the supply curve (the value is positive);
  • % D wage is the percentage change in the equilibrium wage;
  • % D employment is the percentage change in equilibrium employment;
  • % D demand is the percentage change in labor demanded at a given wage level (i.e., the size of the horizontal shift of the demand curve, expressed in percent); and
  • % D supply is the percentage change in labor supplied at a given wage level (i.e., the size of the horizontal shift of the supply curve, expressed in percent).

We observe the initial (pre-reform) equilibrium point (Point A) and final (post-reform) equilibrium point (Point C) in the data collected (discussed in Section II.A below). These data give us the following information.

  • Employment at Point A;
  • Wages at Point A;
  • Employment at Point C; and
  • Wages at Point C

This information can be used along with information about the shapes of the supply and demand curves to obtain Point B in Exhibit 4.1. Point B is the wage and employment combination that would have been attained as a result of the economic expansion in the absence of welfare reform. The wage and employment information described above defines Points A and C (equilibrium outcomes before and after welfare reform).

Using the elasticity assumptions (that is, the value of the percentage change in employment associated with the percentage change in wages), we can draw labor demand and labor supply curves that pass through Points A and C. The intersection of the demand curve passing through Point C (LD1) and the supply curve passing through Point A (LS0) is Point B. Points A and C, along with the elasticity assumptions, are sufficient to produce the entire figure. We can also use this information to calculate other information of interest (e.g., the size of the shifts in the demand and supply curves, and the number of displaced workers).

The only remaining unknowns in the equations (i) and (ii) are the magnitudes of the demand and supply shifts (% D demand and % D supply). Because we have two equations and two unknowns, we can solve these two equations for % D demand and % D supply (see below). We can then use the equations to estimate the effects of the demand and supply shifts independently; that is, we can produce counterfactuals for the impact of economic expansion on employment and wages in the low-skill market, as well as counterfactuals for the impact of welfare reform on the same outcomes.

Solving equations (i) and (ii) for % D demand and % D supply, we get the magnitudes of the demand and supply shifts (equations (iii) and (iv)).

(iii) % D demand = % D employment + ed * % D wage

(iv) % D supply = % D employment - es * % D wage

Equations (i) through (iv) apply to small shifts in the supply and demand curves, but provide reasonable approximations for larger shifts if the elasticities of supply and demand are reasonably constant; they are exact if elasticities are constant. Constant elasticity functions are often used to represent supply and demand curves in the applied economics literature. We make use of elasticity estimates from the literature and the above relationships in our analysis.

The wage and employment equations can be used to analyze the impact of the supply shift alone (let % D demand be zero), or the demand shift alone (let % D supply be zero); that is, given the shift to the demand curve or the supply curve, and given the elasticities, we can use the equations to predict the impact on employment and wages. The percentage of workers displaced by an increase in supply can be derived from (iv).

(v) % displacement = % D supply - % D employment = - es * % D wage

Elasticity of Labor Demand and Labor Supply

The magnitude of the effect of welfare reform on wages and employment is highly dependent on the elasticity of labor demand and labor supply. The effect of the elasticity of labor demand and labor supply on the percentage change in wages and employment can be obtained by differentiating equations (i) and (ii) with respect to the elasticities. Results are summarized in Exhibit 4.2.

Exhibit 4.2 Effects of Elasticities on the Wage and Employment Equations
Demand/Supply Elasticity % D Wage (Absolute Value) % D Employment
% D demand > % D supply (wage increases; employment increases)

Elasticity of Demand

More elastic Lower Lower
Less elastic Higher Higher

Elasticity of Supply

More elastic Lower Higher
Less elastic Higher Lower
% demand < % D supply (wage falls; employment increases)

Elasticity of Demand

More elastic Lower Higher
Less elastic Higher Lower

Elasticity of Supply

More elastic Lower Lower
Less elastic Higher Higher

Note: These results assume that the elasticities of demand and supply are greater than zero.

The direction of the percentage change in the wage is determined by the relative magnitudes of the supply and demand shifts; elasticities only affect the magnitude of the percentage change in the wage. If the labor demand and supply curves are more responsive to the wage (i.e., more elastic), then only a small change in the wage will be needed to restore the labor market to equilibrium. For example, consider the response of the economy if there is a positive demand shift. More workers will be demanded. Wages will rise. If the labor supply curve is very elastic, just a small increase in the wage will be sufficient to draw enough new workers into the labor market to satisfy the increased demand. Alternatively, consider the response of the economy if there is a positive supply shift. Wages will fall. If the labor demand curve is elastic, a small decrease in the wage will be sufficient to increase the quantity of labor demanded to absorb the increased supply. In either case, the equilibrium will be restored with a small change in the wage. Regardless of the relative magnitude of the demand and supply shifts, a higher elasticity of demand and/or supply will result in a smaller absolute change in the wage.

In equation (i) (the % D employment equation), if the demand shift is greater than the supply shift (% D demand > % D supply), then a higher elasticity of demand will result in a lower percentage change in employment, while a higher elasticity of supply will have the opposite effect. If the supply shift is greater than the demand shift (% D supply > % D demand), the results are reversed: the higher the elasticity of demand, the higher the percentage change in employment, and the higher the elasticity of supply, then the lower the percentage change in employment.

Unemployment Rate

One purpose of this project is to assess the impact of welfare reform on the unemployment rate. The unemployment rate, as defined by the Bureau of Labor Statistics (BLS), is an estimate of the percentage of persons who want to work who are not employed. Hence, it is intended to capture involuntary unemployment and does not include those who do not want to work (i.e., who are voluntarily unemployed).

In the conceptual model we have presented thus far, there is no involuntary unemployment, by definition. When shifts to the demand or supply curves occur, wages adjust so that everyone who is willing to work at the prevailing wage is employed. While there are some who might be willing to work only at a higher wage, they are voluntarily unemployed.

As discussed above, the influx of workers into the labor market as a result of welfare reform will lead to the displacement of existing workers unless the expansion of demand is sufficient to completely offset the negative effect of the supply shift on wages. Displacement occurs because workers who were employed before welfare reform lose their jobs to welfare recipients who are willing to work for a lower wage. From a theoretical standpoint, displaced workers might be voluntarily unemployed, because they are not willing to work at the lower wage that now prevails in the labor market. However, from a practical standpoint, BLS defines displaced workers as unemployed as long as they are looking for work. If displaced workers eventually decide that they cannot get a job at a wage they would be willing to accept, they might drop out of the labor force and no longer be counted as involuntarily unemployed by BLS.

We assume that job displacement does increase the unemployment rate in the short-run, as displaced workers are not perfectly aware that the prevailing wage in the labor market has fallen. Displaced workers are replaced by welfare recipients who are willing to work for a lower wage. We assume displacement is the primary means through which welfare reform leads to unemployment in the low-skill labor market.

An inability of wages to adjust to new conditions in the long run would, however, result in long-term unemployment. One reason why this might occur is downward wage rigidity, an issue that we turn to next.

Downward Wage Rigidity

The above discussion assumes that the responsiveness of wage adjustments to shifts in either curve are symmetric: in percentage terms, the effects of a positive shift in a curve on wages and employment are equal, but opposite in sign, to the effects of a negative shift of the same size. Economists have long argued, however, that institutional factors, such as the minimum wage and union contracts, result in asymmetric responses to demand shifts. More specifically, a downward shift in demand, induced by a recession or some other factor, might have very little negative impact on wages, but result in large reductions in employment, while an outward shift in demand of equivalent size, starting from the same point, results in higher wages and small increases in employment.

This phenomenon is very relevant for our analysis because many low-skill workers receive wages that are close to the minimum wage and because the starting point of our analysis, 1993, is near the beginning of a recovery from a major recession. If unemployment in an area in 1993 is still high relative to historical values, then the number of low-skill workers who are willing to work at current wage rates might be substantially higher than the number who are employed, and outward demand shifts are likely to increase employment, with little effect on wages, while outward supply shifts might simply result in greater unemployment, with little effect on employment or wages.

This is illustrated in Exhibit 4.3, in a stylized fashion. The initial supply curve is kinked; it is horizontal at wage level W0, to the left of Point B, then follows the upward sloping LS0. W0 represents a peak wage from the most recent expansionary period. The demand curve that generated that wage rate (i.e., pre-recession peak demand) is represented by LD-1, and peak employment is E-1. The initial demand curve LD0 passes through the horizontal portion of the initial supply curve at E0 (Point A).

Exhibit 4.3
Demand and Supply for Low-Skill Labor in the Presence of Downward Wage Rigidity

Demand and Supply for Low-Skill Labor in the Presence of Downward Wage Rigidity

Consider first an outward shift in demand (LD1), holding supply constant. The wage remains constant until the demand curve shifts past the previous peak demand curve at Point B, then rises along LS0 to Point C, where LD1 intersects LS0. This scenario can be used to illustrate what happens if we ignore the kinked supply curve in the analysis. Suppose we, mistakenly, assume that the initial supply curve is LS*  which also passes through LD0 at Point A, but is not kinked. The employment and wage change generated by the fixed kinked supply curve and the shift in demand is identical to what would be observed if LS* was the initial supply curve and supply shifted from LS* to LS0 as demand shifted from LD0 to LD1 (Point C). Thus, if we ignored the kinked nature of the hypothetical true supply curve, we could easily mistake an increase in employment and wages that is due solely to a demand shift as the result of both a shift in demand and a smaller shift in supply. The size of the false supply shift is the horizontal distance from Point A to Point B. In this stylized model, this is identical to the drop in employment from peak pre-recession employment, E-1, to E0. While the increase in employment from E0 to E1 is entirely due to the demand shift, the analysis would attribute part of the change to the false supply shift  represented by the distance between E1 and E*.

Next, consider an outward shift in supply, to LS1, holding demand constant, again starting at Point A. As the wage rate is already at the pre-expansion peak, it cannot decline further. There is no increase in employment, because demand does not shift. The number of persons seeking jobs at the existing wage rate is E3 (Point D), and E3 E1 workers are unemployed. Hence, under this stylized scenario, the push of welfare reform would just increase unemployment without increasing employment or depressing wages. This scenario further illustrates what happens if the analysis were to ignore the kinked supply curve. It would appear that the shift in supply due to welfare push was the horizontal distance from LS* to LS1, greater than the horizontal distance which represents the actual shift, from LS0 to LS1, again by amount E-1 E0.

The analysis becomes more complex if there is both a demand and a supply shift, because the order in which they shift matters. If supply shifts first, to LS1, then demand shifts to LD1, the wage rate stays constant and employment expands to E2; E3 E2 workers are unemployed. If, instead, demand shift firsts, then supply, wages increase to W1, employment increases to E1, and the shift in supply has no impact on wages or employment, but results in unemployment of E4 E1.

In a less stylized version of the above example, the supply curve would be more elastic at wage rates below previous peaks, and less elastic at higher wage rates. Such a supply curve would yield findings that are qualitatively similar to the stylized version presented above, with outward shifts in supply generating only modest wage reductions and outward shifts in demand generating only modest increases up to the point where demand passes the previous peak, and more pronounced increases thereafter.

In the decomposition of employment and wage changes into changes due to welfare push and changes due to the economy, we do not attempt to model kinks in the supply curve like those that appear in this stylized example, because we do not have a sound empirical basis for differentiating between the supply elasticity for low-skill workers during the early recovery from a recession and the elasticity as economic expansion passes the earlier peak. Instead, we use a constant elasticity model with an elasticity based on findings in the literature. The analysis of the kinked supply curve, above, is very important, however, in helping us interpret the findings.

Summary

The economic model presented here provides a framework for analyzing the entry of large numbers of welfare recipients into the low-skill labor market. Better job opportunities and welfare reform are the two main reasons for this phenomenon.

The economic model is an equilibrium model of low-skill labor demand and labor supply. In the economic model, a shift in labor demand represents the increase in job opportunities and a shift in labor supply represents the increase in labor market entry of welfare recipients due to work requirements and time limits. Both of these changes increase employment, but have opposite impacts on prevailing wages.

If the equilibrium wage falls, some former workers will be displaced; i.e., their employers will replace them with welfare recipients who have entered the labor market and are willing to work for lower wages. This is likely to result in short-term unemployment, as those displaced continue to seek work at a higher wage. In the long-term some of these individuals are likely to become discouraged and leave the labor force, mitigating the impact on the unemployment rate.

Calculating Low-Skill Employment and Wages

To calculate low-skill employment and wages, we calculated employment and wages by occupation. We first classified occupations as low-skill, medium-skill, and high-skill, and then aggregated employment and wages across all occupations that were classified as low-skill.

Calculating Employment and Wages by Occupation

To estimate employment and wages by occupation, we obtained employment and wage data by industry from the BLS Covered Employment and Wages (ES-202) system. We distributed employment and wages in each industry across occupations using the National Industry Staffing Patterns (NISP).

It would have been preferable to use employment and wage data by occupation instead of by industry, obviating the need for an industry to occupation conversion. The Occupational Employment Statistics (OES) survey, on which the NISP is based, tabulates employment and wage data by occupation. However, wage data are only available for 1998. In addition, the OES did not survey all agricultural, forestry, fishing, and private household industries. Therefore, we used the ES-202 for employment and wages.

1.Covered Employment and Wages (ES-202)

The ES-202 program provides comprehensive employment and wage information by industry for workers covered by state Unemployment Insurance laws and federal workers covered by the Unemployment Compensation for Federal Employees program. Employment is reported monthly for covered workers who were working or who received pay during the pay period including the 12th of the month. Wages, from payroll records, are reported quarterly for all covered workers who received pay during the quarter.

The ES-202 data are coded according to the Standard Industrial Classification (SIC) system, which classifies establishments by industry activity. While a four-digit SIC level is assigned to all reporting establishments, employment and wages are often aggregated to the two-digit or one-digit level. Examples of industries in these categories include the following:

Four-Digit Two-Digit One-Digit

Wheat Grains

Agricultural Production Crops Agriculture, Forestry, Fishing

Creamery Butter

Food and Kindred Products Manufacturing

Womens Clothing

Stores Apparel & Accessory Stores Retail Trade

Junior Colleges

Educational Services Services

Because of confidentiality concerns, ES-202 data are not disclosed for any level in which the universe has three or fewer employers or is dominated by a single employer that represents more than 80 percent of employment. We obtained ES-202 data at the two-digit SIC level rather than four-digit SIC to reduce the number of undisclosed employment records.(34) When we encountered undisclosed cells at the two-digit SIC level, we allocated the remaining employment and payroll at the one-digit SIC level (total minus all disclosed employment within an industry level) equally across all undisclosed cells.

2.National Industry Staffing Patterns (NISP)

For this study, we use the NISPs occupational estimates by industry, which are produced from the national OES, to convert employment and payroll (total wages) by industry for each region to employment and payroll by occupation. Prior to 1996, the OES surveyed only one-third of the industries in the sample each year, taking three years to fully complete the survey. This was done to reduce respondent burden. In addition, it only collected employment data. In 1996, OES began to survey all industries in each year, but recommended using three years of data to reduce sampling error. As discussed above, it began collecting wage information in 1996.

We obtained the NISP at the two-digit level from BLS for 1990 through 1998. We combined 1990 to 1992, 1993 to 1995, and 1996 to 1998 NISPs to obtain the industry occupation matrix for 1993, 1996, and 1998. Because NISP is only produced at the national level, we made the implicit assumption that there were no significant regional differences in the occupational distribution for each industry. We tested this assumption using the Current Population Survey (CPS), and found a consistent pattern of occupational employment in each industry across four regions (Northeast, Midwest, South, West) and in urban versus rural areas. These results are presented in Appendix A.

The 1998 NISP was used to distribute payroll for 1993, 1996, and 1998. We assumed there were no significant differences in the occupational distribution within an industry over time. We tested this assumption using the CPS. The CPS shows a declining percentage of payroll in low-skill occupations over time, while we assumed no decline in the proportion of total payroll allocated to low-skill occupations (see Appendix B). We are not too concerned about the latter assumption because there is no decline in the share of employment in low-skill occupations from the CPS, while there is an increase over time from the NISP (i.e., we found employment in low-skill occupations grew at a faster rate than employment in high-skill occupations). Thus, in both cases, wages per low-skill worker grew at a slower rate than wages per high-skill worker.

Classifying Occupations by Education and Training Requirements

We classified occupations into three skill categories (low, medium, and high) using the BLS education and training requirements for occupation groups. The BLS requirements are straightforward and consistent with the CPS and OES. BLS classifies occupations in 11 groups according to their education and training requirements. We grouped occupations that require short-term (less than one month) on-the-job training as low-skill. This is the occupation category that requires the least level of education and training in the BLS classification. Examples of occupations in the low-skill category are retail salespersons, office clerks, cashiers, truck drivers, personal care and home health aides. We grouped occupations that require more than short-term on-the-job training, but less than a bachelors degree, as medium-skill. These occupations may require more on-the-job training, more work experience, vocational training, or an associates degree. We grouped occupations that require a bachelors degree or higher as high-skill.

One disadvantage of the BLS classification is that it does not indicate whether occupations require a high school degree. This should not be a problem, however, as CPS tabulations show that there is a mix of high school graduates and high school drop-outs in the population who received income from TANF in 1998 (see Exhibit 2.1).

Other researchers (Leete and Bania(35); Lerman, Loprest, and Ratcliffe(36)) have defined low-skill jobs to include both short-term and moderate-term training occupations. Leete and Bania also included long-term training occupations in their definition, but limited the percentage of welfare recipients who were eligible for these jobs. We chose to include only short-term training occupations in our definition because we believed these occupations would be affected most by welfare reform.

As discussed above, we assumed there were no regional differences between the occupational distribution of employment within an industry (i.e., differences in skill level resulted from differences in the industry mix of employment across regions). Appendix C presents a comparison of the percentage of 1998 employment by skill level from our estimates (ES-202 merged with the NISP) and the OES, which we were able to obtain for selected regions only. The percentage of employment that was low-skill was slightly lower from the ES-202/NISP than from OES.

Assumptions of Elasticity of Labor Demand and Labor Supply

Our assumption is that welfare reform will have the largest effect on the low-skill labor market. Hence, the relevant elasticities of labor demand and labor supply are those for low-skill labor. Bartik presents a summary of elasticity estimates used in studies that examine the effect of welfare reform on wages and displacement.(37) The three studies cited in the exhibit (Mishel and Schmitt,(38) Holzer,(39) and Bernstein(40))used a labor demand elasticity of -0.3. Holzer and Bernstein used a labor supply elasticity of 0.4; Mishel and Schmitt used a labor supply elasticity of zero. All three studies repeated their calculations for alternative labor demand and labor supply elasticity assumptions.

Because elasticities for different types of labor can vary, it is necessary to use an elasticity estimate for workers who are similar in characteristics to welfare recipients. The labor demand estimates used in the studies cited above are taken from the minimum wage literature. The labor supply estimates used in these studies are taken from the literature on the decline in employment among low-skill adult males in the 1980s.

Based on these studies, we used a labor demand elasticity of -0.3 and a labor supply elasticity of 0.4. We were not able to find any studies that compared the differences in the elasticities across rural and urban areas. We were also not able to find any studies that compared the differences in the elasticity of labor supply before and after welfare reform.

Assumptions about the elasticity of labor supply and labor demand are critical to our analysis. As was discussed in Section 4.A and will be discussed in more detail in Chapter 5, the elasticity assumptions are instrumental in determining the size of the demand and supply shifts from the employment and wage data that we collected. Therefore, we used alternative labor demand and labor supply elasticities to test the sensitivity of our results to the elasticity assumptions. These results are presented in Appendix D. We found that our basic findings are not affected much by reasonable changes in the elasticities as a result of the small size of the increase in employment due to welfare reform relative to the low-skill labor market.

Estimating Welfare Recipients in Labor Force

To estimate the number of welfare recipients in the labor force, we collected caseload information from all regions and estimated labor force participation from a combination of caseload employment reports, state estimates produced by DHHS, and a study of TANF leavers from the National Survey of Americas Families (NSAF).

The change in welfare caseloads over time is due to two effects: the change in the number of welfare recipients who enter the welfare rolls (inflow) and the change in the number of welfare recipients who leave the rolls (outflow).

Change in Inflow

Changes in inflow can be caused by the economy and by welfare reform. For example, potential welfare recipients who might have entered the rolls under previous economic conditions (e.g., the 1991 recession) may be less likely to apply if they can find jobs easily in a strong economy. Potential welfare recipients might also be diverted from welfare due to welfare reforms, such as time limits, stringent work requirements, and state welfare diversion programs that either require an applicant to look for work before being approved for benefits or offer a one-time lump sum payment to help potential clients to avoid welfare altogether. These potential recipients might instead rely on family support for income or enter the labor market.

Change in Outflow

Changes in outflow also are due to the economy and welfare reform. For example, in a strong economy, more recipients might leave welfare due to better job opportunities in the labor market. Welfare reform policies also play a role. More recipients might leave welfare due to time limits and stringent work requirements. Recipients who leave may rely on family support for income or might enter the labor market. Some might have been working off the books while on welfare and could continue to rely on their shadow labor market activities for income after they leave the rolls.

For this analysis, we are measuring the change in stock between two points in time and not focusing on changes in flows. Therefore, our analysis assumes there is no net effect on the labor market when a person leaves welfare, but is replaced by another person who enters welfare.

Our analysis uses the number of welfare recipients who are newly employed. That is, our estimates include those who were not in the labor force initially (e.g., in a given month in 1993), but who entered at a later point in time (e.g., in a given month in 1996). The number of welfare recipients entering the labor market is estimated using the following equation:

(vi) (C0 C1) *(L1 W0) + C1 * (W1 W0)

where:

C0 = Caseload at time(0)
C1 = Caseload at time(1)
W0 = Percent of caseload in labor force at time(0)
W1 = Percent of caseload in labor force at time(1)
L1 = Percent of leavers in labor force at time(1)

Note that C0 C1 represents the change as a result of welfare recipients leaving (net of those arriving) and others diverted from entering the rolls. While the formula appears to assume that welfare leavers and welfare stayers  those who continued to be on the rolls at time(1)  had equal labor force participation rates at time(0), this is not a necessary condition. We can assume that welfare leavers had a higher rate of labor force participation in the initial period than did stayers and still get the same results.(41)

Exhibit 4.4 presents each regions monthly caseload, for each relevant year, along with estimates of the percentage of welfare recipients and welfare leavers who were in the labor force. These estimates came from the following sources:

  • The state welfare agencies supplied the average monthly caseload estimates for the specific study regions.
  • A few of the state welfare agencies (Alabama, Missouri, New York, and Wisconsin) were able to estimate the percent of the caseload employed in the study regions. For Mississippi, Oregon, South Carolina, Tennessee, and Vermont regions, we used the state average, reported by DHHS.
  • The Urban Institute analyzed the NSAF, which surveyed families between February and November 1997. Approximately 61 percent of the families that had been on welfare at some point since 1995 and had left and remained off welfare at the survey date were employed.(42) Similar rates were found in the studies of welfare leavers funded by ASPE.
  • We estimated that a smaller percent of leavers (50 percent) were employed between 1993 and 1996. We know of no study that has estimated this percentage for this pre-PRWORA time period.

As this exhibit shows, New York, Vermont, and Wisconsin welfare recipients were more likely to be in the labor force while on welfare. These states offer relatively high cash grants that enable individuals with earnings to remain eligible for welfare. The Southern states, on the other hand, offer lower grants and have a lower share of the welfare population employed.

Exhibit 4.4
Monthly Caseloads and Participation in Labor Force
1993 1996 1998
  Monthly Caseload Caseload in Labor Force (%) Monthly Caseload Caseload in Labor Force (%) Leavers in Labor Force (%) Monthly Caseload Caseload in Labor Force (%) Leavers in Labor Force (%)

Decatur and Florence, Alabama

1,577 1.0 1,167 1.1 50.0 645 8.7 61.0

Rural Mississippi

45,384 9.1 36,565 8.1 50.0 19,096 7.6 61.0

Joplin, Missouri

2,081 4.5 1,906 4.6 50.0 1,271 10.1 61.0

Southeast Missouri

12,674 4.6 10,817 8.6 50.0 7,972 14.1 61.0

Jamestown, New York

3,154 17.0 2,516 24.0 50.0 1,975 27.0 61.0

North Country, New York

6,656 9.0 5,749 15.8 50.0 4,145 20.9 61.0

Medford-Ashland, Oregon

2,540 12.7 1,820 11.8 50.0 896 3.8 61.0

Central Oregon

1,342 12.7 1,026 11.8 50.0 635 3.8 61.0

Florence, South Carolina

2,619 6.8 2,469 9.8 50.0 1,665 16.8 61.0

Vermont

10,081 12.0 9,210 23.1 50.0 7,591 22.7 61.0

Eau Claire, Wisconsin

2,037 28.2 1,116 27.9 50.0 302 13.6 61.0

Wausau, Wisconsin

1,162 22.5 785 23.1 50.0 234 15.8 61.0

United States

4,963,000 7.8 4,628,000 10.3 50.0 3,305,000 15.6 61.0
Source: Lewin calculations using data provided by state welfare agencies and DHHS.

Endnotes

(32) Technically, there are two supply curves behind the supply curve drawn, one for each of the population groups. The sum of labor supplied at a given wage from the two groups corresponds to total labor supplied at that wage on the supply curve shown. The shift in the labor supply curve for the welfare target group corresponds to the shift in the total labor supply curve. [Back To Text]

(33) The derivation of these equations can be found in Freeman (1977). They hold only approximately, except for infinitesimally small shifts. A simple way to derive them is to begin with the assumption that the demand and supply curves are linear in natural logarithms (i.e., assume that the wage and employment axes in Exhibit 4.1 are natural log scales). For small changes, changes in logs are equivalent to percentage changes in levels. The slope of the supply curve on the log scales is the inverse of the supply elasticity and the slope of the demand curve is the negative of the inverse of the demand elasticity. Given these slopes, Equations (i) and (ii) can be derived via the use of geometry. If percentage changes in the equation are replaced by changes in logarithms, and if the demand and supply curves are linear in the logarithms, the equations apply exactly. [Back To Text]

(34) We obtained data for all of our metropolitan regions from the BLS and for nonmetropolitan regions from State Employment Security Agencies (SESA) for 1993, 1996, and 1998. [Back To Text]

(35) Leete, L. & N. Bania (1999) [Back To Text]

(36) Lerman, R., P. Loprest, & C. Ratcliffe (1999). [Back To Text]

(37) Bartik, T. J. (1999). Displacement and Wage Effects of Welfare Reform. W.E. Upjohn Institute for Employment Research. Kalamazoo, MI. [Back To Text]

(38) Mishel, L. & J. Schmitt (1995). Cutting Wages By Cutting Welfare. Economic Policy Institute. Washington, DC. [Back To Text]

(39) Holzer, H. J. (1996). Employer Demand, AFDC Recipients, and Labor Market Policy. Institute for Research on Poverty Discussion Paper No. 1115-96. Michigan State University. East Lansing, MI. [Back To Text]

(40) Bernstein, J. (1997). Welfare Reform and the Low-Wage Labor Market: Employment, Wages, and Wage Policies. Economic Policy Institute Technical Paper #226. Washington, DC. [Back To Text]

(41) If we assume there are two groups  leavers (C0  C1) and stayers (C1)  and each group has an average employment rate at time(0) of WL0 and WS0, respectively, then the following equation measures the increase in total employed from time(0) to time(1):

(vii) (C0  C1) * (L1  WL0) + C1 * (W1  WS0).

The following identity also holds:

(viii) ( C0  C1) * WL0 + C1 * WS0 = W0 * C0 {total employed at time(0)}

Substituting the right-hand side of Eq. (viii) in Eq. (vii), yields the following:

(C0  C1) * L1 + C1 * W1  W0 * C0, which is equal to Eq. (vi)  [Back To Text]

(42) Loprest, P. (1999). Families Who Left Welfare: Who Are They and How Are They Doing? Washington, DC: The Urban Institute. [Back To Text]

Chapter 5: Findings

In this chapter, we first provide a descriptive account of changes in low-skill employment, low-skill wages, and welfare caseloads between 1993 and 1998 in our 12 regions. We then decompose the changes into supply-induced or demand-induced changes generated by the economic model. We further decompose the supply-induced changes into several factors including welfare reform, reduction of excess supply of labor, and population growth. In addition, we provide an upper bound estimate on the effect welfare reform could have had on employment, wages, and displacement in the low-skill labor market. Finally, we summarize the findings and discuss the implications for further research.

Change in Employment, Wages, and Welfare Caseloads between 1993 and 1998

In this section, we provide a descriptive account of changes in employment and wages in low-skill, medium-skill, and high-skill labor markets in our 12 regions. The descriptive analysis gives an overview of how employment and wages were changing in the selected regions over the time period being studied. We also provide a descriptive account of changes in the number of welfare recipients in the labor force. In addition, we compare the increase in the number of welfare recipients in the labor force to the increase in low-skill employment. This comparison is the first step in determining whether the low-skill labor market in each region was able to absorb the increase in the number of welfare recipients in the labor force.

Employment and Wages

1. Low-Skill Employment and Wages: 1993 to 1996

Exhibit 5.1 presents the percentage change in employment and wages by skill level between 1993 and 1996. As this exhibit shows, there was a sizeable increase in low-skill employment in all regions except the New York regions between 1993 and 1996, which experienced a decline in population between these two years. Five of the 12 regions experienced a low-skill employment increase of more than 10 percent, and 10 of the 12 regions experienced a low-skill employment increase of more than 5 percent. The largest employment increase was in Eau Claire, Wisconsin at 15.5 percent; low-skill employment declined by 0.3 percent in North County, New York. The average increase in low-skill employment for the 12 regions was 9.2 percent, compared to a national average of 8.7 percent.

However, low-skill wages (in 1998 dollars) declined in 8 of the 12 regions between 1993 and 1996. The biggest decline was in Central Oregon at 3.8 percent. The average decrease in low-skill wages for the 12 regions was 0.4 percent, compared to a national average of 0.1 percent. Changes in employment were not closely related to changes in wages. For example, low-skill employment increased by roughly the same amount in Alabama and Southeast Missouri (9.2 and 9.3 percent, respectively), but low-skill wages declined by 0.8 percent in Alabama and increased by 0.4 percent in Southeast Missouri. We think this was due to the variation in the supply and demand forces that were responsible for the changes across regions.

Exhibit 5.1
Percent Change in Employment and Wages by Skill Level, 1993-1996
Region Employment Wages
Low Skill (%) Medium Skill (%) High Skill (%) Low Skill (%) Medium Skill (%) High Skill (%)

Decatur and Florence, Alabama

9.2 4.1 6.5 -0.8 2.9 -0.8

Rural Mississippi

13.5 5.9 9.5 1.2 5.3 1.7

Joplin, Missouri

10.8 6.5 11.0 6.6 5.3 1.9

Southeast Missouri

9.3 4.9 11.1 0.4 5.6 1.4

Jamestown, New York

1.9 -1.2 1.1 -2.7 0.3 -2.2

North Country, New York

-0.3 -0.2 4.1 -0.4 2.2 -2.7

Medford-Ashland, Oregon

12.8 10.0 13.7 -1.8 1.0 -0.8

Central Oregon

14.0 12.1 15.4 -3.8 -1.0 -0.8

Florence, South Carolina

8.1 0.5 12.6 -0.5 2.5 -0.5

Vermont

7.4 5.7 7.6 -2.0 1.4 -1.7

Eau Claire, Wisconsin

15.5 13.0 14.2 -1.5 0.8 -2.0

Wausau, Wisconsin

8.6 9.9 11.0 1.0 2.1 1.8

Average

9.2 5.9 9.8 -0.4 2.4 -0.4

United States

8.7 5.7 8.3 -0.1 2.8 0.7

Source: Lewin calculations using ES-202, NISP, and BLS education and training requirements data.
Note: Percentage change calculated as a difference of the logs.

2.Low-Skill Employment and Wages: 1996 to 1998

Exhibit 5.2 presents the percentage change in employment and wages by skill level between 1996 and 1998. While low-skill employment increased in all regions between 1996 and 1998, the increases were generally less pronounced than between 1993 and 1996 in most regions. None of the 12 regions experienced a low-skill employment increase of more than 10 percent. Seven of the 12 regions experienced low-skill employment increases of more than 5 percent. The largest employment increase was in Medford-Ashland, Oregon (8.1 percent) and the smallest increases were in Jamestown, New York and Decatur and Florence, Alabama (2.5 percent each). The average increase in low-skill employment for the 12 regions was 5.7 percent, compared to a national average of 7.1 percent.

Low-skill wages increased in all 12 regions between 1996 and 1998. The largest increase was in Eau Claire, Wisconsin (6.2 percent). The average increase in wages was 2.1 percent, compared to a national average of 3.8 percent. Again, changes in wages and employment were not closely related. For example, Decatur and Florence, Alabama and Jamestown, New York had the same low-skill employment increases (2.5 percent), but very different low-skill wage increases (1.1 and 3.0 percent, respectively). It appears that the supply and demand forces behind the employment increases varied across the areas.

Exhibit 5.2
Percent Change in Employment and Wages by Skill Level, 1996-1998
  Employment Wages
Region Low Skill (%) Medium Skill (%) High Skill (%) Low Skill (%) Medium Skill (%) High Skill (%)

Decatur and Florence, Alabama

2.5 -1.9 2.3 1.1 2.6 1.4

Rural Mississippi

6.8 1.6 0.2 1.9 6.6 5.7

Joplin, Missouri

7.9 5.3 6.0 1.4 3.7 3.9

Southeast Missouri

4.8 2.1 4.0 0.8 3.4 1.9

Jamestown, New York

2.5 -1.0 1.7 3.0 5.2 3.7

North Country, New York

4.5 0.7 3.4 2.3 6.8 3.4

Medford-Ashland, Oregon

8.1 4.9 6.1 2.1 4.0 3.8

Central Oregon

7.5 4.7 8.5 1.7 5.4 4.4

Florence, South Carolina

6.4 3.3 8.0 0.6 5.3 3.3

Vermont

5.0 2.3 3.2 2.5 7.1 3.9

Eau Claire, Wisconsin

4.4 6.9 5.3 6.2 6.2 7.1

Wausau, Wisconsin

7.4 3.0 4.1 1.9 4.7 5.2

Average

5.7 2.7 4.4 2.1 5.1 4.0

United States

7.1 2.9 5.3 3.8 7.5 6.8

Source: Lewin calculations using ES-202, NISP, and BLS education and training requirements data.
Note: Percentage change calculated as a difference of the logs.

3.Changes by Other Skill Levels

Employment and wages in medium-skill and high-skill occupations followed the same trends as employment and wages in low-skill occupations. Over time, the increase in employment was higher in the 1993 to 1996 period than in the 1996 to 1998 period at all skill levels while the increase in wages was higher in the 1996 to 1998 period than in the 1993 to 1996 period at all skill levels.

Employment in low-skill occupations increased more than employment in medium-skill occupations in both the 1993 to 1996 and 1996 to 1998 periods and more than employment in high-skill occupations in the 1996 to 1998 period. Wages in low-skill occupations increased less than wages in medium-skill occupations in both the 1993 to 1996 and 1996 to 1998 periods, and less than wages in high-skill occupations while in the 1996 to 1998 period.

Welfare Recipients' Participation in the Labor Force

We first estimated the number of current and former welfare recipients who entered the labor force between 1993 and 1996, and between 1996 and 1998, and then compared these estimates to the increase in low-skill employment presented in Section I.A above. A region where the number of welfare recipients entering the labor force is significantly smaller than the increase in low-skill employment implies that the growth in jobs could accommodate the inflow of welfare recipients. Note that welfare recipients may have entered the labor force because of welfare reform or the improved economy. Conversely, a large number of recipients entering the labor market relative to the increase in low-skill jobs would lead us to believe that unemployment would increase and/or wages would decline.

Using the estimates presented in Exhibit 5.3 and the methodology outlined in Chapter 4, we estimated the increase of welfare recipients in the labor force and compared them to the increase in low-skill employment. The increase of welfare recipients included former recipients and current recipients and netted out the number who were in the labor force in the earlier year. As Exhibit 5.3 shows, the increase in low-skill employment exceeded the increase in welfare recipients entering the labor force for all regions, except North Country in the 1993 to 1996 period. North Country experienced a reduction in low-skill employment in the 1993 to 1996 period, which coincided with a decline in the total population. Employed welfare recipients as a percent of new low-skill employment ranged from 2.5 percent in Joplin, Missouri to 89 percent in Jamestown, New York between 1993 and 1996 and from 8 percent in Central Oregon to 52 percent in Southeast Missouri between 1996 and 1998.

Exhibit 5.3
Comparing Welfare Recipient Entrants and Increases in Low-Skill Employment
Region 1993-1996 1996-1998
Increase of Welfare Recipients in Labor Force (a) Increase in Low-Skill Employment (b) Recipients Employed/ Low-Skill Employment (%) (a/b) Increase of Welfare Recipients in Labor Force (c) Increase in Low-Skill Employment (d) Recipients Employed/ Low-Skill Employment (%) (c/d)

Decatur and Florence, Alabama

203 3,877 5.2 361 1,118 32.3

Rural Mississippi

3,235 38,716 8.4 9,129 21,729 42.0

Joplin, Missouri

82 3,271 2.5 428 2,644 16.2

Southeast Missouri

1,275 6,766 18.8 1,934 3,737 51.8

Jamestown, New York

387 436 88.8 259 590 44.0

North Country, New York

762 -155 n/a 936 2,732 34.3

Medford-Ashland, Oregon

253 3,633 7.0 383 2,544 15.1

Central Oregon

109 3,152 3.5 142 1,874 7.6

Florence, South Carolina

139 1,865 7.5 528 1,573 33.6

Vermont

1,355 8,213 16.5 586 5,836 10.0

Eau Claire, Wisconsin

197 4,360 4.5 227 1,378 16.4

Wausau, Wisconsin

109 2,046 5.3 192 1,895 10.1

Average

    15.3     26.1

United States

255,443 4,090,284 6.2 846,346 3,585,571 23.6

Source: Lewin calculations using ES-202, NISP, BLS education and training requirements data, and data provided by state welfare agencies.

During the 1993 and 1996 period, the increase in low-skill employment dwarfed the increase of welfare recipients in the labor market in all but the New York regions. Thus, it appears that the low-skill labor market could absorb the inflow of welfare recipients during this period without a serious effect on employment or wages.

For most regions, the share of recipients as a percent of low-skill employment also increased in the 1996 and 1998 period. This was due, in part, to the fact that caseloads declined significantly during this period and a higher percent of welfare recipients were combining work and welfare than from 1993 to 1996. However, even between 1996 and 1998, welfare recipients accounted for only about one quarter of new low-skill employment. Coupled with the fact that unemployment declined in each of these regions between 1996 and 1998, it appears that the regions could accommodate the inflow of welfare recipients.

The average share of welfare recipients as a percent of low-skill employment in the 12 regions is slightly higher than the national average. This indicates that rural areas had a higher percentage of welfare recipients in the low-skill labor market than urban areas. However, it is not clear whether wages declined. Nor is it clear how the dual effects  welfare reform and the improved economy  impact the employment and wage results. These are discussed below.

Decomposing the Effect of Welfare Reform and Economic Expansion

In this section, we describe our estimates of the effect of welfare push on employment and wages in each of the regions. As discussed earlier, distinguishing between the effects of welfare push and demand pull is difficult, for several reasons. One is that, for the most part, these regions still had historically high unemployment rates in 1993, following the 1991 recession. Hence, their economies could be reasonably characterized as having an excess supply of labor at going wage rates, which makes application of the demand/supply analysis problematic. In addition, factors other than welfare reform and economic expansion had effects on some of these regions low-skill labor markets during this period. The EITC, population growth, and increases in the minimum wage were three such factors.

We also present the results of the demand/supply decomposition described in Chapter 4. Interpretation of the findings from this analysis was problematic for the reasons described above, especially during the 1993 to 1996 period. More specifically, we found supply shifts that were much larger than could be credibly attributed to welfare reform. Hence, in the subsequent section we analyzed the possible explanations of the estimated supply shifts, and used our estimates of increases in employment for the welfare population to limit the contribution of welfare reform to supply shifts and, more importantly, employment and wage growth.

Demand and Supply Shifts

Following the methodology presented in Section 4.1, we calculated the magnitude of the demand shift (% D demand) and the supply shift (% D supply) in each regions low-skill labor market over the 1993 to 1996 and 1996 to 1998 periods. The estimated demand shift presumably represents the impact of economic expansion on the labor market  increased demand for goods and services produced by industries employing low-skill labor increases the demand for low-skill labor at any wage rate. The supply shift presumably reflects any factor that shifts the supply curve, including welfare reform. Population growth, the EITC and possibly other factors could affect supply as well. This simple analysis ignores the possibility that, because of the earlier recession, there was an excess supply of labor in the regions at the beginning of the period; i.e., substantially more people wanted to have jobs at prevailing wages than were actually employed.

It is important to keep in mind that the demand and supply shifts represent shifts in the demand and supply curves, respectively; i.e., they indicate the change in demand or supply holding wages constant. While of some interest in themselves, they are intermediate results that are needed to decompose changes in employment and wages into changes due to each of the shifts. Employment increases with both a positive demand shift and a positive supply shift. Wages increase with a positive demand shift and decrease with a positive supply shift; if the demand shift is greater than the supply shift, then wages rise, and vice versa.

Exhibit 5.4 presents the magnitudes of the demand and supply shifts in the 1993 to 1996 and 1996 to 1998 time periods, relative to employment levels.(43)

Exhibit 5.4
Magnitudes of Demand and Supply Shifts
  1993-1996 1996-1998
Region Demand Shift Supply Shift Demand Shift Supply Shift

Decatur and Florence, Alabama

9.0 9.6 2.8 2.1

Rural Mississippi

13.8 13.0 7.4 6.1

Joplin, Missouri

12.8 8.1 8.4 7.4

Southeast Missouri

9.4 9.1 5.0 4.5

Jamestown, New York

1.0 3.0 3.4 1.3

North Country, New York

-0.4 -0.1 5.2 3.6

Medford-Ashland, Oregon

12.2 13.5 8.7 7.2

Central Oregon

12.9 15.5 8.0 6.8

Florence, South Carolina

8.0 8.3 6.6 6.1

Vermont

6.8 8.3 5.7 4.0

Eau Claire, Wisconsin

15.0 16.0 6.3 1.9

Wausau, Wisconsin

8.9 8.2 7.9 6.6

Average

9.1 9.4 6.3 4.8

United States

8.7 8.8 8.2 5.6

Source: Lewin calculations using ES-202, NISP, and BLS education and training requirements data.

In both periods, the estimated supply and demand shifts were large and comparable in magnitude. In the earlier period, the demand shifts were generally smaller than the supply shifts, which explains the wage declines observed in some areas. The opposite was true in the later period for all areas, as needed to explain wage increases.

The decomposition of employment changes due to supply and demand shifts appears in Exhibit 5.5. The economic expansion, which increased the demand for labor, played a much larger role in increasing low-skill employment than welfare reform or other supply factors. While the contribution of the demand shift to employment growth was larger than the contribution of the supply shift in both periods, the demand shift was more pronounced in the 1996 to 1998 period (64 percent, on average) than in the 1993 to 1996 period (57 percent).

Exhibit 5.5
Decomposition of Percent Change in Employment in Low-Skill Labor Markets
  1993-1996 1996-1998
Region Total Shift in Demand Shift in Supply Total Shift in Demand Shift in Supply

Decatur and Florence, Alabama

9.2 5.3 4.0 2.5 1.6 0.9

Rural Mississippi

13.5 7.7 5.8 6.8 4.2 2.6

Joplin, Missouri

10.8 6.2 4.6 7.9 4.8 3.2

Southeast Missouri

9.3 5.3 4.0 4.8 2.9 1.9

Jamestown, New York

1.9 1.1 0.8 2.5 1.9 0.6

North Country, New York

-0.3 -0.1 -0.1 4.5 3.0 1.6

Medford-Ashland, Oregon

12.8 7.3 5.5 8.1 5.0 3.1

Central Oregon

14.0 8.0 6.0 7.5 4.6 2.9

Florence, South Carolina

8.1 4.6 3.5 6.4 3.8 2.6

Vermont

7.4 4.3 3.2 5.0 3.3 1.7

Eau Claire, Wisconsin

15.5 8.8 6.6 4.4 3.6 0.8

Wausau, Wisconsin

8.6 4.9 3.7 7.4 4.5 2.8

Average

9.2 5.2 4.0 5.6 3.6 2.1

United States

8.7 5.0 3.7 7.1 4.7 2.4

Source: Lewin calculations using ES-202, NISP, and BLS education and training requirements data.

The decomposition of wage changes due to supply and demand shifts appears in Exhibit5.6. As this exhibit shows, the contribution of the supply shift to wage growth was almost as large or larger than the contribution of the demand shift in the earlier period, while the demand shift was substantially larger than the supply shift in the later period. This is a direct consequence of the essentially stagnant wage growth in the first period and the substantial positive wage growth in the second period. Further, in almost all regions, the magnitude of the supply effect was larger in the first period than in the second, while the reverse was true for the magnitude of the demand effect. This is contrary to what we expected for the impacts of both welfare reform and economic expansion. That is, we expected that the impact of welfare reform would be greater in the later period, after PRWORA was enacted, while the effect of economic expansion would be greater in the earlier period, on the heels of the recession. It seems likely, therefore, that the estimated supply shift in the earlier period generally overstated the real shift in supply because of excess labor supply at the beginning of the period. For the same reason, the estimated demand shift likely understated the effect of economic expansion. We consider the interpretation of the estimated supply shift further in the next section.

Exhibit 5.6
Decomposition of Percent Change in Wages in Low-Skill Labor Markets
  1993-1996 1996-1998
Region Total Shift in Demand Shift in Supply Total Shift in Demand Shift in Supply

Decatur and Florence, Alabama

-0.8 12.8 -13.7 1.1 4.1 -3.0

Rural Mississippi

1.2 19.7 -18.6 1.9 10.6 -8.7

Joplin, Missouri

6.6 18.2 -11.6 1.4 11.9 -10.5

Southeast Missouri

0.4 13.4 -13.0 0.8 7.2 -6.4

Jamestown, New York

-2.7 1.5 -4.2 3.0 4.8 -1.8

North Country, New York

-0.4 -0.6 0.1 2.3 7.5 -5.2

Medford-Ashland, Oregon

-1.8 17.5 -19.3 2.1 12.4 -10.3

Central Oregon

-3.8 18.4 -22.2 1.7 11.4 -9.8

Florence, South Carolina

-0.5 11.4 -11.9 0.6 9.4 -8.7

Vermont

-2.0 9.8 -11.8 2.5 8.2 -5.7

Eau Claire, Wisconsin

-1.5 21.5 -22.9 6.2 9.0 -2.8

Wausau, Wisconsin

1.0 12.8 -11.8 1.9 11.3 -9.5

Average

-0.4 13.0 -13.4 2.1 9.0 -6.9

United States

-0.1 12.5 -12.5 3.8 11.8 -7.9

Source: Lewin calculations using ES-202, NISP, and BLS education and training requirements data.

Analysis of Supply Shifts and the Maximum Impact of Welfare Reform

In this section we analyze the estimated supply shifts further and also produce estimates of the maximum impact that welfare reform could have had on employment and wages during both periods.

We begin by assessing the plausible magnitudes of three factors that could account for the estimated supply shifts: welfare reform, reduction of excess supply of labor following the 1991 recession; and population growth.

1.Increase of Welfare Recipients in the Labor Force

The impact of welfare reform on the supply shift could be no larger than the change in employment in the welfare population. Both demand pull and welfare reform push contributed positively to this growth, so the impact of welfare reform could be no larger than the total change in employment. We calculated the increase of welfare recipients in the labor force based on caseload declines, estimates of the percentage of leavers in the labor force, and estimates of the percentage of welfare recipients combining work and welfare from caseload reports.(44)

2.Reduction of Excess Supply of Labor

Because wages tend to be rigid downward, there was an increase in unemployment at the prevailing wage between 1989 and 1993.(45) When economic recovery began in 1993, these previously unemployed workers began to find employment at the prevailing wage. There was an increase in low-skill employment without an increase in low-skill wages. Because we did not model downward wage rigidity, this observation was interpreted as a supply shift in the economic model during this time period. We predicted that the economy would have completely recovered from the recession by 1996 in most regions.

To assess the possible contribution of reductions in the excess supply of labor following the 1991 recession, we estimated the difference between employment in the base year for each of the two periods (i.e., 1993 or 1996) and what employment would have been had the employment rate (employment divided by the adult population) in that year been the same as in 1989. We interpreted the latter as an estimate of pre-recession peak employment, adjusted for population growth. If the difference was positive, we interpreted it as the excess labor supply remaining in the base year as the result of the recession. If the difference was negative, we assumed that no excess labor supply remained. We calculated the percentages by dividing the difference by base-year employment. Note that these percentages applied to all skill levels, combined. We did not have the information needed to compute estimates by skill level. We suspect that relative excess labor supply for low-skill workers would be greater than for all workers combined, so these estimates likely understated the contribution of excess labor supply to the estimated supply shifts.

3.Population Growth

The final factor we considered explicitly was population growth. The contribution of population growth to supply shifts was estimated as the percentage increase in the entire population over each period. We assumed that the percent increase in the low-skill labor force due to population growth would be similar.(46)

All regions, with the exception of the New York regions, experienced some population growth that contributed to the supply shift. In particular, the Oregon regions experienced significant growth during this period.

4.Other Factors

As indicated earlier, at least two other factors could have contributed to change in the supply curve over this period  the EITC and the increase in the minimum wage. The EITC would have shifted the supply curve out, and the increase in the minimum wage would have effectively made the supply curve more inelastic at wage rates near the minimum and generated excess supply.

Results of the analysis for 1993 to 1996 and 1996 to 1998 appear in the first four columns of Exhibits 5.7 and 5.8, respectively.

For the earlier period (Exhibit 5.7), the maximum that welfare reform could have contributed to the supply shift was well below the estimated supply shift in all but one region (North Country, New York). Population growth and excess supply helped explain the large estimated shifts in most areas, but if we used the values in the table to estimate the contributions of welfare reform, excess supply, and population growth to the estimated supply shift, there was a considerable residual in several areas. We do not have good explanations for all of these residuals. Those in Alabama and Mississippi could have reflected the fact that these areas economies had high unemployment rates throughout the 1980s, so our estimates might have substantially underestimated the size of excess supply in 1993.(47) The source of the exceptionally high residual for Eau Claire, Wisconsin is unknown. The EITC and minimum wage might explain some of the residual shift. Measurements errors and errors in the elasticity estimates could also have been contributing factors.

For the later period (Exhibit 5.8), the maximum that welfare reform could have contributed to the supply shift was below the estimated supply shift in all regions. There was excess labor supply from the recession in only 4 of 12 regions. The residuals were smaller than they were for the earlier period, but they were still not negligible.

It is apparent from this analysis that the estimated supply shift was a poor indicator of the impact of welfare reform on supply, with the possible exception of a few regions in the 1993 to 1996 period. The maximum estimate of the impact of welfare reform on the supply shift, which was based on analysis of caseload data, appears to be much more useful for our purposes. Based on this estimate, plus the estimated demand and supply elasticities, we also estimated the maximum impact of welfare reform on both employment and wages in each period. These estimates appear in columns seven of Exhibits 5.7 and 5.8. For comparison purposes, we also show the total change in employment and wages (columns six and nine), and the total that the supply and demand analysis attributed to supply shifts (columns eight and eleven).

Exhibit 5.7
Factors Contributing to Estimated Supply Shift, 1993-1996
  Factors Contributing to Estimated Supply Shift Employment Growth Wage Growth
Region Total Welfare Reform (max) Net Job Loss Pop Growth Residual Total Welfare Reform (max) Supply Shift Total Welfare Reform (max) Supply Shift
Decatur and Florence, Alabama 9.6 0.5 0.0 1.1 8.0 9.2 0.2 4.0 -0.8 -0.7 -13.7
Rural Mississippi 13.0 1.0 0.0 2.6 9.4 13.5 0.5 5.8 1.2 -1.7 -18.6

Joplin, Missouri

8.1 0.3 0.9 4.3 2.6 10.8 0.1 4.6 6.6 -0.4 -11.6

Southeast Missouri

9.1 1.8 2.7 2.3 2.3 9.3 0.8 4.0 0.4 -2.6 -13.0

Jamestown, New York

3.0 0.3 0.7 -1.1 3.1 1.9 0.7 0.8 -2.7 -2.4 -4.2

North Country, New York

-0.1 0.7 0.4 -1.4 0.2 -0.3 0.6 -0.1 -0.4 -1.8 0.1

Medford-Ashland, Oregon

13.5 1.0 1.8 6.6 4.1 12.8 0.4 5.5 -1.8 -1.4 -19.3

Central Oregon

15.5 0.6 2.3 11.3 1.3 14.0 0.2 6.0 -3.8 -0.7 -22.2

Florence, South Carolina

8.3 0.5 4.4 3.1 0.3 8.1 0.3 3.5 -0.5 -0.9 -11.9

Vermont

8.3 1.3 1.8 2.2 3.0 7.4 0.5 3.2 -2.0 -1.8 -11.8

Eau Claire, Wisconsin

16.0 0.8 1.5 1.5 12.4 15.5 0.3 6.6 -1.5 -1.1 -22.9

Wausau, Wisconsin

8.2 0.5 0.8 1.8 5.2 8.6 0.2 3.7 1.0 -0.7 -11.8

Average

9.4 0.8 1.4 2.9 4.3 9.2 0.4 4.0 -0.4 -1.4 -13.4

United States

8.8 0.6 1.7 2.9 3.6 8.7 0.2 3.7 -0.1 -0.8 -12.5

Source: Lewin calculations using ES-202, NISP, BLS education and training requirements data, and data provided by state welfare agencies.

Exhibit 5.8
Factors Contributing to Estimated Supply Shift, 1996-1998
Factors Contributing to Estimated Supply Shift Employment Growth Wage Growth
Region Total Welfare Reform (max) Net Job Loss Pop Growth Residual Total Welfare Reform (max) Supply Shift Total Welfare Reform (max) Supply Shift

Decatur and Florence, Alabama

2.1 0.8 -2.8 1.4 -0.1 2.5 0.4 0.9 1.1 -1.2 -3.0

Rural Mississippi

6.1 2.8 -1.5 1.1 2.2 6.8 1.3 2.6 1.9 -4.2 -8.7

Joplin, Missouri

7.4 1.2 -1.3 2.1 4.1 7.9 0.6 3.2 1.4 -1.9 -10.5

Southeast Missouri

4.5 2.4 -0.4 1.0 1.1 4.8 1.1 1.9 0.8 -3.6 -6.4

Jamestown, New York

1.3 1.1 -0.9 -1.8 2.0 2.5 0.5 0.6 3.0 -1.6 -1.8

North Country, New York

3.6 1.5 -0.9 -1.5 3.6 4.5 0.7 1.6 2.3 -2.3 -5.2

Medford-Ashland, Oregon

7.2 1.2 1.6 2.7 3.3 8.1 0.5 3.1 2.1 -1.8 -10.3

Central Oregon

6.8 0.5 2.4 6.3 0.0 7.5 0.3 2.9 1.7 -0.8 -9.8

Florence, South Carolina

6.1 2.1 4.0 1.1 2.9 6.4 0.9 2.6 0.6 -3.2 -8.7

Vermont

4.0 0.5 1.0 0.3 3.2 5.0 0.2 1.7 2.5 -0.7 -5.7

Eau Claire, Wisconsin

1.9 0.7 -0.6 0.4 0.8 4.4 0.3 0.8 6.2 -1.1 -2.8

Wausau, Wisconsin

6.6 0.8 -0.5 1.1 4.8 7.4 0.3 2.8 1.9 -1.1 -9.5

Average

4.8 1.3 0.0 1.2 2.3 5.7 0.6 2.1 2.1 -2.0 -6.9

United States

5.6 1.7 0.1 1.9 1.9 7.1 0.7 2.4 3.8 -2.5 -7.9

Source: Lewin calculations using ES-202, NISP, BLS education and training requirements data, and data provided by state welfare agencies

Employment and Wages

In the 1993 to 1996 period, we estimated that, at the maximum, welfare reform could have accounted for employment growth of 0.1 to 0.8 percent, with an average of 0.4 percent. Presumably the contribution was generally less than the small maximums reported. Welfare reform could have reduced wage growth by as much 2.6 percent in one region, but the average estimate was only 1.4 percent. Again, in all likelihood the actual reductions in wage growth due to welfare reform were smaller.

In the 1996 to 1998 period, we estimated that, at the maximum, welfare reform could have accounted for employment growth of 0.5 percent to 2.8 percent, with an average of 1.3 percent. Welfare reform could also have reduced wage growth by as much 0.7 to 4.1 percent; the average estimate was only 2 percent. The actual increases in employment and reductions in wage growth due to welfare reform were presumably smaller than these maximum estimates.

Displacement

Based on the estimate of the maximum impact of welfare reform on the supply shift, from the analysis of caseload data, we also estimated the maximum percentage of workers displaced by welfare reform in each period. We present two estimates for each region (see Exhibit 5.9). The gross displacement estimate is an estimate of what the maximum displacement due to welfare reform would have been in the absence of any other shift in the demand or supply curves. The net displacement estimate is the amount of displacement that occurred after the effects of other factors that shifted the demand and supply curves. A negative value for net displacement means that workers other than welfare recipients were induced to accept jobs by wage growth.

In the 1993 to 1996 period, we estimated that, in the absence of any other change, welfare reform would have displaced no more than 1.0 percent of workers in any one region, with an average maximum of 0.5 percent in the 12 regions. The average is just slightly higher than the U.S. maximum for this period of 0.3 percent. Net displacement was smaller than the maximum gross displacement in six of the regions, indicating that other factors generally offset any displacement due to welfare reform. In four of these regions, net displacement was negative, indicating that other low-skill workers were induced to enter employment. In five of the other six regions, displacement was essentially the same as the maximum due to welfare reform, or just slightly higher, after considering all supply and demand factors. In Central Oregon, displacement was much higher after considering all factors, reflecting the high rate of growth of the regions population.

In the 1996 to 1998 period, welfare reform by itself would have displaced no more than 1.7 percent of workers in any region, with an average of 0.8 percent  slightly below the national average of 1.0 percent. The potential for displacement due to welfare reform was larger in the 1996 to 1998 period, because the decrease in caseloads was larger. During this period, however, demand growth was so large in every region that net displacement was negative. That is, the estimates imply that no displacement due to welfare reform actually occurred.

Exhibit 5.9
Percent of Labor Displaced
1993 - 1996 1996 - 1998
Region Welfare Reform Gross (max)a/ Netb/ Welfare Reform Gross (max)a/ Netb/

Decatur and Florence, Alabama

0.3 0.3 0.5 -0.4

Rural Mississippi

0.7 -0.5 1.7 -0.8

Joplin, Missouri

0.2 -2.7 0.8 -0.6

Southeast Missouri

1.0 -0.2 1.4 -0.3

Jamestown, New York

1.0 1.1 0.6 -1.2

North Country, New York

0.7 0.2 0.9 -0.9

Medford-Ashland, Oregon

0.5 0.7 0.7 -0.8

Central Oregon

0.3 1.5 0.3 -0.7

Florence, South Carolina

0.4 0.2 1.3 -0.3

Vermont

0.7 0.8 0.3 -1.0

Eau Claire, Wisconsin

0.4 0.6 0.4 -2.5

Wausau, Wisconsin

0.3 -0.4 0.4 -0.7

Average

0.5 0.1 0.8 -0.9

United States

0.3 0.0 1.0 -1.5

Source: Lewin calculations using ES-202, NISP, BLS education and training requirements data, and data provided by state welfare agencies.
a\ Maximum displacement due to welfare reform if other factors did not affect supply and demand.
b\Actual displacement as result of all factors affecting supply and demand. Negative values imply that low-skill workers outside of the welfare population were induced to work by wage increases.

The maximum displacement due to welfare reform in each region is small, even in the absence of other changes, because the maximum number of welfare recipients entering employment due to welfare reform is a small share of the low-skill labor market in each region. In addition, labor supply is fairly inelastic, so that the number of existing workers who leave employment as wages fall is small.

These estimates essentially assume that low-skill workers are perfectly substitutable across all low-skill jobs in the region, and that low-skill workers themselves are indifferent to which low-skill job they have, holding wages and benefits constant. This might be true in general, but violated in specific instances. For instance, workers displaced by welfare reform in one area of a region might not be the same workers who accept job openings in another area. Similarly, retail trade workers displaced by welfare reform might not be the same workers who accept unskilled job openings in construction. Thus, displacement could occur at a level that cannot be observed in the data. Nonetheless, the fact that displacement at the level we are able to observe is small, given our maximum estimate of employment increases due to welfare reform and no other changes, indicates that the low-skill labor markets in these areas are able to absorb welfare recipients with little negative impact on existing workers.

Projecting the Effect of a Recession

Welfare reform happened at a propitious time. In 1996, the economy was in the midst of the longest peacetime expansion in American history, and in September 2000 the unemployment rate reached a 30-year low of 3.9 percent. Based on our analysis, we found that the strong economy was able to absorb the inflow of welfare recipients into the low-skill labor market with negligible adverse effects. In fact, welfare reform might have helped ease the shortage of low-skill workers in the labor force, and hence might have helped sustain a high rate of economic growth by easing the inflationary pressure on low-skill wages.

The next economic downturn will be a significant test of welfare reform. Will there be a higher increase in unemployment and a higher decrease in wages during a recession as a result of the increased number of welfare recipients in the labor force? Will rural and small metropolitan areas be more adversely affected by welfare reform when the economy slows than urban areas?

We projected the wage and employment effects of a recession both under the status quo and in the absence of welfare reform. In these projections, we assumed that a recession would cause the demand curve to shift back by the same percentage as it shifted out in the 1993 to 1996 period.(48) In 1993, the economy was beginning to recover from the 1991 recession, so we assumed that the shift in the demand curve in this period was representative of the amount the economy would contract in a recession. This estimate was likely an overstatement of the demand shift in a recession, because the 1993 to 1996 demand shift likely included both a recovery from the recession and growth in the economy.

To simulate the 1998 low-skill labor market conditions in the absence of welfare reform, we shifted the supply curve back by the same percentage as the maximum outward shift attributable to welfare reform in the 1996 to 1998 period. Because PRWORA passed in 1996, we assumed that the decline in welfare caseloads represented the maximum impact of welfare reform on the supply curve in this period. We took the equilibrium wage and employment resulting from this shift as the equilibrium if welfare reform had not happened.

If, as we have assumed throughout, the demand and supply curves are linear on the log scale, the effect of a demand shift (i.e., a fixed percent reduction in demand) on wages and employment is the same in percentage terms regardless of the starting equilibrium point. The percentage impacts on both wages and employment, under both the status quo and in the absence of welfare reform, are shown in Exhibit 5.10. These figures assume that wages adjust downward. The effect of a recession on the economy is substantial, with an average decrease of 5 percent in employment and 13 percent in wages across the 12 regions, slightly worse than for the U.S. as a whole.

Exhibit 5.10
Percentage Effect of a Recession on Employment and Wages
Region Shift In Demand (%) Change in Employment (%) Change in Wages (%)

Decatur and Florence, Alabama

-9.0 -5.1 -12.8

Rural Mississippi

-13.8 -7.9 -19.7

Joplin, Missouri

-12.8 -7.3 -18.2

Southeast Missouri

-9.4 -5.4 -13.4

Jamestown, New York

-1.0 -0.6 -1.5

North Country, New York

0.4 0.2 0.6

Medford-Ashland, Oregon

-12.2 -7.0 -17.5

Central Oregon

-12.9 -7.4 -18.4

Florence, South Carolina

-8.0 -4.6 -11.4

Vermont

-6.8 -3.9 -9.8

Eau Claire, Wisconsin

-15.0 -8.6 -21.5

Wausau, Wisconsin

-8.9 -5.1 -12.8

Average

-9.1 -5.2 -13.0

United States

-8.7 -5.0 -12.5

Source: Lewin calculations using ES-202, NISP, BLS education and training requirements data, and data provided by state welfare agencies.

While the model implies that the percent change in employment in wages is the same under the two scenarios, the change in the levels of employment and wages differ because equilibrium wage and employment levels under any demand scenario are affected by the presence or absence of welfare reform. In the absence of welfare reform, wages would be higher and employment would be lower. Hence, if the percentage shift in the demand curve associated with a recession was the same with or without welfare reform in place, as we assumed, the changes in the levels of employment and wages are different. The simulated recession reduces employment somewhat more under the status quo welfare system than in the absence of welfare reform because welfare reform resulted in higher employment in 1998. The level of employment in the recession, nonetheless, was higher under the status quo welfare system than under the no reform scenario. Conversely, the recession reduced wages somewhat less under the status quo welfare system than in the absence of welfare reform because welfare reform depressed wages somewhat in 1998. The level of wages in the recession, however, is somewhat lower under the status quo welfare system than under the no reform scenario. Exhibit 5.11 presents the employment effect of a recession under the two scenarios. Exhibit 5.12 presents the wage effect.

Exhibit 5.11
Absolute Effect of a Recession on Employment
Status Quo Without Welfare Reform
Region Equilibrium Employment Reduction in Employment Resulting Employment Equilibrium Employment Reduction in Employment Resulting Employment

Decatur and Florence, Alabama

45,057 2,316 42,742 44,903 2,308 42,595

Rural Mississippi

328,993 25,979 303,014 325,046 25,667 299,378

Joplin, Missouri

34,672 2,529 32,142 34,493 2,516 31,977

Southeast Missouri

80,174 4,305 75,869 79,349 4,260 75,089

Jamestown, New York

24,171 145 24,026 24,057 144 23,913

North Country, New York

61,568 -136 61,705 61172 -135 61,308

Medford-Ashland, Oregon

32,803 2,295 30,509 32,634 2,283 30,352

Central Oregon

25,935 1,914 24,021 25,879 1,910 23,969

Florence, South Carolina

25,461 1,162 24,299 25,232 1,152 24,080

Vermont

120,398 4,698 115,700 119,675 4,670 115,006

Eau Claire, Wisconsin

31,818 2,730 29,088 31,668 2,718 28,950

Wausau, Wisconsin

26,625 1,359 25,266 26,488 1,352 25,136

Average

69,806 4,108 65,698 69,216 4,070 65,146

United States

52,450,640 2,612,898 49,837,742 52,270,810 2,603,940 49,666,870

Source: Lewin calculations using ES-202, NISP, BLS education and training requirements data, and data provided by state welfare agencies.

Exhibit 5.12
Absolute Effect of a Recession on Wages
Status Quo Without Welfare Reform
Region Equilibrium Wages Reduction in Wages Resulting Wages Equilibrium Wages Reduction in Wages Resulting Wages

Decatur and Florence, Alabama

16,405 2,108 14,297 16,593 2,132 14,461

Rural Mississippi

14,910 2,943 11,966 15,506 3,061 12,445

Joplin, Missouri

16,743 3,054 13,690 17,030 3,106 13,924

Southeast Missouri

13,617 1,828 11,790 14,084 1,891 12,194

Jamestown, New York

15,676 235 15,441 15,922 238 15,684

North Country, New York

15,810 -87 15,897 16,148 -89 16,238

Medford-Ashland, Oregon

16,087 2,813 13,274 16,363 2,861 13,501

Central Oregon

16,603 3,063 13,540 16,722 3,085 13,637

Florence, South Carolina

15,373 1,754 13,618 15,834 1,807 14,027

Vermont

16,528 1,612 14,916 16,858 1,644 15,214

Eau Claire, Wisconsin

15,442 3,313 12,130 15,685 3,365 12,320

Wausau, Wisconsin

16,912 2,158 14,754 17,202 2,195 15,007

Average

15,842 2,066 13,776 16,162 2,108 14,054

United States

19,380 2,414 16,967 19,602 2,441 17,161

Source: Lewin calculations using ES-202, NISP, BLS education and training requirements data, and data provided by state welfare agencies.
Note: Estimated annual wages per person.

The absolute differences between the wage and employment levels in a recession and under the status quo and in the absence of welfare reform are very small. This finding mirrors the results from the 1996 to 1998 period. Welfare recipients were a small percentage of the low-skill labor force, so their presence in the labor market does not lead to a big increase in employment or a big decrease in wages.

The assumptions of the model might be wrong. One might argue, for instance, that the magnitude of the percentage shift in the demand curve is dependent on welfare reform. We think that arguments could be made in either direction, and are not aware of a persuasive argument that would dominate in one direction or the other.(49) Welfare reform contributed somewhat to the growth in the economy over the period under study, by increasing the supply of low-skill labor. Now that the economy has adapted to welfare reform, there appears to be no obvious reason to think that contractions will be larger, in percentage terms, than they have been in the past.

Welfare reform might also have changed the elasticity of supply, but it is also difficult to determine the direction of change, if any. While the labor supply of a given low-income worker with a family might be less elastic than before welfare reform, because welfare benefits are more difficult to obtain, a larger share of low-income workers might be from families that would qualify for welfare benefits in the event of job loss.

Summary and Implications for Further Research

Overall, we found that between 1993 and 1996, welfare reform had a minimal impact on job displacement and wage reduction. It could have more adversely affected labor markets after 1996, when PRWORA was enacted, but even during this period, the strong economy helped rural and small metropolitan labor markets absorb the inflow of welfare recipients. Even regions that experienced dramatic declines in caseloads, such as the Wisconsin regions, experienced no adverse effects. In a less robust economy, welfare reform likely would have depressed low-skill wage growth and displaced some low-skill workers who were not welfare recipients. The size of these effects, however, appears to be small relative to the effects of other factors whose fluctuations affect low-skill labor markets  such as population growth and the business cycle  because welfare recipients who enter the labor market are a fairly small share of the low-skill labor force in each area.

Several limitations of this study deserve mention:

  • The magnitude of the estimated effects of welfare reform and the economy are dependent on the estimates of the elasticity of labor demand and labor supply. Only national-level estimates are available, and even among these there is a significant range. Differences in elasticities across rural and urban areas may exist. We conducted a sensitivity analysis (incrementing and decrementing the elasticities by 0.1) and found that our basic findings are not affected much by reasonable changes in the elasticities as a result of the small size of the increase in employment due to welfare reform relative to the low-skill labor market.
  • The estimates of welfare recipients who entered the labor market relied on estimates obtained from a national leaver study. If differences in the percent of leavers who entered the labor market differ between rural and urban areas, then the maximum effect of welfare reform on employment and wages could differ from the ones presented.
  • We cannot account for the effect of the EITC in our estimates. We believe the expansion primarily affected the 1993 to 1996 period, and may have affected some of the unexplained shift in the supply curve during this period. A recent study separated the effect of EITC expansions from the effect of welfare policies and local labor market conditions, and found that EITC expansions have a large positive effect on employment of adults from welfare families.(50)
  • We allocated industry payroll to low-skill, medium-skill, and high-skill occupations in each of the three years based on the payroll distribution of occupations within industries in 1998; distributions for earlier years did not exist. The extent to which the percent distribution differed between 1993 and 1998 affects our low-skill wage estimates, but not the employment estimates.
  • We found that welfare reform did not significantly affect the market for low-skill labor because the market is large relative to the number of welfare recipients entering it. Because we defined the low-skill market narrowly  assuming that welfare recipients sought jobs that required little education or on-the-job training  we believe we might be over-estimating, and not under-estimating, the effect of welfare reform on labor markets.

Future studies could augment this study by collecting additional information and using other data sources not available to us at the time of this study. In addition, more will be learned about the effect of time limits on reducing caseloads and the effect of welfare reform during periods of slow economic growth. Future studies could address the following:

  • In the next few years, as more welfare recipients reach time limits and as the economy slows and, perhaps, enters a recession, will welfare recipients begin to displace low-skill workers and affect wage levels? Will they be more or less vulnerable to job loss than other low-skill workers? To conduct future analysis, researchers will be able to rely on the Occupational Employment Statistics survey, which began collecting wage information starting in 1996. This survey is also a better source for employment information by occupation. Finally, some of the confounding factors we encountered  such as the EITC and the effects of the earlier recession will not have as large an effect on the estimates going forward.
  • Are welfare leavers in rural areas more or less likely to enter the labor force than their counterparts in urban areas? To date, leaver studies have focused primarily on urban areas and offer little information on the labor force participation of welfare leavers living in more rural areas. An analysis of national data sets, such as the National Survey of Americas Families, focusing on welfare recipients living in rural areas could supplement this study. In addition, some leaver studies include an adequate sample size of individuals residing in rural areas; future research could analyze these subsamples in more detail.
  • Are welfare offices in rural areas able to offer the job search and job training assistance available in urban areas? What percent of the welfare population in rural areas are exempt from time limits and work participation requirements because of the lack of services? Field research or phone surveys targeting these study regions could provide useful information on the issues facing welfare offices in rural labor markets, and provide context for some of the findings from this study.
  • What types of jobs are welfare recipients taking? Did we narrow the field to the correct set of occupations? Are these jobs likely to evaporate during a downturn? Field research conducted in some of these study regions would supplement the analysis from this study.

Endnotes

(43) The estimates presented are changes in natural logarithms over the relevant period, which can be interpreted as approximate percentages. [Back To Text]

(44) Some individuals who would have become welfare recipients in the absence of welfare reform might have been diverted from entering welfare as a result of welfare reform. As discussed in Chapter 4, caseload declines capture the decrease in both the stock and the flow of welfare recipients. Therefore, welfare diversion is accounted for in our estimate of caseload declines, although it is not possible to separate out its effect. [Back To Text]

(45) The unemployment rate rose in all regions except Decatur and Florence, Alabama and rural Mississippi between 1989 and 1993. [Back To Text]

(46) We assume that the elasticity of supply used in the analysis does not capture any possible effects of wage changes on migration between states. [Back To Text]

(47) Decatur and Florence, Alabama had an unemployment rate of 11.7 percent in 1986 compared to 8.3 percent in 1989; and rural Mississippi had an unemployment rate of 12.9 percent in 1986 compared to 8.3 percent in 1989. [Back To Text]

(48) For simulation purposes, we used the change in the logarithm of employment, holding price constant, as the measure of a percentage shift. [Back To Text]

(49) One could argue that welfare reform had an impact on the automatic stabilizer feature of transfer programs. In general, these programs pump more government money into the hands of consumers when they lose earnings during recessions, dampening the reduction in demand. Because it might be harder for a given family to get assistance under TANF if a breadwinner loses a job than it was under AFDC, the automatic stabilizer feature of this transfer program could have been weakened. At the same time, however, under TANF a larger share of those who lose jobs because of a recession might be in TANFs target population and able to obtain benefits. Further, a larger share of those who are in TANFs target population might qualify for Unemployment Insurance benefits, another automatic stabilizer, than would have in the absence of welfare reform. [Back To Text]

(50) Hotz, J. H., Mullin, C. H., and Scholz, J. K. (2000). The Earned Income Tax Credit and Labor Market Participation of Families on Welfare. Paper for the Joint Center for Poverty Research Conference on Means-Tested Transfers, December 7-8, 2000. [Back To Text]

Appendices

Comparison of NISP Employment Across Regions Using the CPS

We used the National Industry Staffing Patterns (NISP) to convert employment by industry to employment by occupation in our study regions. By using a national distribution of occupations by industry for 12 different regions, we implicitly made the assumption that the distribution of employment by occupations in an industry did not change by region. We used the CPS Outgoing Rotation Groups for 1994, 1996, and 1998 to test the plausibility of this assumption.(1) We compared the distribution of occupations by skill level (low, medium, and high) by one-digit industry for urban (MSA) versus rural (non-MSA) areas and for four geographic regions (Northeast, Midwest, South, and West).

The CPS had a higher percentage of medium-skill workers and a lower percentage of low-skill workers compared to the NISP. This difference between the two data sources stems from the nature of the two surveys. The NISP (derived from the Occupational Employment Statistics (OES)) is an employer survey, in which employers answer questions about the occupations of their employees. The CPS is a household survey, in which individuals themselves answer questions about their occupation. We believe that there is occupation creep in the CPS, which means that individuals are likely to report being in higher-skill occupations than they actually are. For example, a food preparation worker (low-skill) might report himself as a restaurant cook (medium-skill). We do not think that this issue presents a problem for our analysis, because occupation-creep is likely to be consistent across regions.

We found that the distribution of skill levels across one-digit industries was consistent across four regions and urban versus rural areas for most industries. Agriculture and mining were the two industries that varied across rural versus urban areas and across regions. We concluded that we had a reasonable degree of confidence in the NISP, with the caveat that we might have introduced some measurement error in the agriculture and mining industries. It is important to note that we selected areas that did not have a large share of agricultural or mining employment.

Exhibit A.1
Percent Distribution of Employment, 1994
By Industry, Skill Category, and Region
Industry Skill Level U.S. MSA Non-MSA Northeast Midwest South West
Agriculture Low-Skill 41.6 52.1 32.1 42.3 27.6 45.5 54.4
Medium-Skill 49.2 35.8 61.4 45.8 64.3 45.3 35.9
High-Skill 9.2 12.2 6.5 11.9 8.0 9.2 9.7
Mining Low-Skill 11.1 9.2 13.3 18.0 15.1 9.5 11.6
Medium-Skill 60.1 49.1 73.2 61.7 66.9 57.6 64.4
High-Skill 28.8 41.7 13.5 20.4 18.1 32.9 24.0
Construction Low-Skill 15.3 15.0 16.2 14.7 14.1 16.4 15.0
Medium-Skill 69.3 68.3 72.7 71.6 72.1 68.5 66.2
High-Skill 15.4 16.7 11.1 13.7 13.8 15.1 18.8
Manufacturing Low-Skill 19.4 18.1 23.2 17.8 20.9 19.7 18.0
Medium-Skill 58.3 55.6 66.0 55.9 59.5 61.8 52.1
High-Skill 22.3 26.4 10.8 26.3 19.6 18.6 30.0
Transportation Low-Skill 42.6 41.2 49.1 41.3 43.8 43.3 41.2
Medium-Skill 39.3 39.5 38.5 40.5 38.3 38.7 40.2
High-Skill 18.1 19.4 12.5 18.2 17.9 18.0 18.6
Wholesale Trade Low-Skill 26.1 25.1 30.8 26.1 27.0 26.0 25.1
Medium-Skill 59.9 59.7 61.3 59.2 59.8 60.3 60.1
High-Skill 14.0 15.3 7.9 14.7 13.2 13.7 14.8
Retail Trade Low-Skill 47.2 47.5 45.9 48.2 47.5 47.4 45.6
Medium-Skill 47.6 46.9 50.2 47.0 46.8 47.7 48.9
High-Skill 5.2 5.6 3.9 4.8 5.7 4.8 5.6
Finance Low-Skill 21.3 20.8 25.1 19.6 21.9 22.2 21.1
Medium-Skill 31.2 30.8 34.2 29.3 31.7 32.1 31.4
High-Skill 47.5 48.5 40.7 51.2 46.5 45.7 47.5
Services Low-Skill 24.6 24.0 27.2 22.2 25.0 24.5 26.5
Medium-Skill 36.5 35.9 39.2 35.9 37.5 37.6 34.3
High-Skill 38.9 40.1 33.6 41.8 37.5 37.8 39.2
Public Low-Skill 17.0 16.8 18.0 14.7 16.8 17.3 18.5
Medium-Skill 58.9 58.2 61.8 63.2 59.2 58.0 56.8
High-Skill 24.1 25.0 20.3 22.1 24.1 24.7 24.6
Total Low-Skill 28.2 27.8 29.9 26.6 28.3 28.6 29
Medium-Skill 46.3 44.7 52.2 45 47.8 47.3 43.9
High-Skill 25.5 27.6 17.9 28.4 23.8 24.1 27.1
Source: Lewin calculations using the CPS Outgoing Rotation Groups.

Exhibit A.2
Percent Distribution of Employment, 1996
By Industry, Skill Category, and Region
Industry Skill Level U.S. MSA Non-MSA Northeast Midwest South West
Agriculture Low-Skill 46.3 54.7 36.6 51.8 34.7 45.6 58.7
Medium-Skill 43.6 33.6 55.1 32.8 56.6 43.5 32.5
High-Skill 10.1 11.7 8.3 15.5 8.7 10.9 8.8
Mining Low-Skill 11.7 13.1 10.1 19.1 14.7 9.2 14.5
Medium-Skill 64.4 52.0 78.0 61.2 70.5 63.2 64.7
High-Skill 24.0 34.9 11.9 19.7 14.8 27.6 20.8
Construction Low-Skill 15.6 15.3 16.9 16.2 15.0 15.9 15.3
Medium-Skill 67.5 66.7 70.5 66.8 68.8 68.3 65.4
High-Skill 16.9 18.0 12.7 17.0 16.2 15.8 19.4
Manufacturing Low-Skill 20.0 18.3 25.2 17.9 22.4 19.9 18.2
Medium-Skill 56.3 54.1 63.1 54.5 56.5 60.1 51.1
High-Skill 23.8 27.6 11.8 27.6 21.1 20.1 30.7
Transportation Low-Skill 42.2 41.0 48.1 41.3 45.6 41.8 39.9
Medium-Skill 38.1 38.2 37.5 40.5 36.2 37.6 39.0
High-Skill 19.7 20.8 14.4 18.2 18.3 20.6 21.1
Wholesale Trade Low-Skill 26.8 25.2 35.3 28.0 26.3 26.6 26.8
Medium-Skill 59.3 59.8 56.8 56.2 60.5 61.2 57.8
High-Skill 13.9 15.0 8.0 15.8 13.2 12.3 15.4
Retail Trade Low-Skill 47.6 47.4 48.3 47.3 48.8 46.8 47.6
Medium-Skill 47.0 46.7 48.4 47.4 45.5 48.0 47.0
High-Skill 5.4 5.9 3.4 5.3 5.7 5.2 5.4
Finance Low-Skill 21.1 20.5 25.6 19.4 21.9 22.1 20.5
Medium-Skill 29.2 28.8 32.4 27.4 27.5 30.6 30.9
High-Skill 49.7 50.7 42.0 53.3 50.6 47.3 48.5
Services Low-Skill 23.9 23.4 26.3 23.0 24.2 23.5 25.0
Medium-Skill 35.5 34.7 39.4 34.8 36.4 36.4 33.7
High-Skill 40.7 41.9 34.3 42.2 39.5 40.1 41.3
Public Low-Skill 16.1 15.3 19.7 14.2 16.9 16.4 16.3
Medium-Skill 59.2 59.2 59.3 61.8 60.0 57.6 59.4
High-Skill 24.7 25.5 21.0 24.1 23.1 26.0 24.3
Total Low-Skill 28.2 27.5 31.0 26.9 28.9 28 28.8
Medium-Skill 44.9 43.7 50.4 43.5 45.8 46.4 42.9
High-Skill 26.9 28.8 18.6 29.6 25.3 25.6 28.3
Source: Lewin calculations using the CPS Outgoing Rotation Groups.

Exhibit A.3
Percent Distribution of Employment, 1998
By Industry, Skill Category, and Region
Industry Skill Level U.S. MSA Non-MSA Northeast Midwest South West
Agriculture Low-Skill 47.3 54.0 38.5 49.6 34.9 44.8 60.6
Medium-Skill 41.1 32.8 52.0 39.1 55.0 41.6 28.4
High-Skill 11.6 13.3 9.5 11.3 10.1 13.5 11.0
Mining Low-Skill 12.6 11.6 13.8 7.2 16.7 11.9 13.6
Medium-Skill 59.7 48.6 73.3 69.4 62.4 57.3 62.7
High-Skill 27.7 39.9 13.0 23.4 20.9 30.8 23.8
Construction Low-Skill 14.1 13.7 15.8 14.4 13.5 13.7 15.2
Medium-Skill 68.6 67.8 71.5 68.1 71.1 68.9 65.9
High-Skill 17.3 18.5 12.8 17.5 15.4 17.5 18.9
Manufacturing Low-Skill 19.4 18.0 23.8 17.5 21.8 19.5 17.1
Medium-Skill 56.2 53.9 63.7 54.3 55.8 60.1 52.0
High-Skill 24.4 28.1 12.5 28.2 22.4 20.4 30.9
Transportation Low-Skill 43.1 41.9 48.8 41.8 45.3 43.4 41.3
Medium-Skill 36.5 36.4 37.2 39.0 36.0 34.9 37.5
High-Skill 20.5 21.7 14.0 19.2 18.7 21.7 21.2
Wholesale Trade Low-Skill 26.2 26.0 27.7 25.3 26.5 25.5 27.7
Medium-Skill 58.2 57.4 63.1 57.8 59.8 59.6 55.2
High-Skill 15.5 16.7 9.2 16.9 13.8 15.0 17.1
Retail Trade Low-Skill 47.5 47.8 46.4 47.9 47.5 47.3 47.5
Medium-Skill 46.9 46.4 49.3 46.7 46.8 47.1 47.0
High-Skill 5.6 5.9 4.3 5.4 5.7 5.7 5.5
Finance Low-Skill 20.9 20.2 26.9 19.2 22.0 21.6 20.4
Medium-Skill 27.6 27.6 27.9 25.9 26.3 29.4 27.9
High-Skill 51.5 52.3 45.2 54.9 51.7 49.0 51.6
Services Low-Skill 23.6 23.0 26.4 22.5 23.6 22.7 25.6
Medium-Skill 34.9 34.1 39.3 35.7 35.2 35.9 32.6
High-Skill 41.5 42.9 34.2 41.7 41.2 41.3 41.8
Public Low-Skill 15.6 15.5 15.9 16.4 14.6 15.0 16.7
Medium-Skill 59.2 57.7 65.5 63.4 60.9 57.5 57.0
High-Skill 25.3 26.8 18.6 20.2 24.5 27.5 26.3
Total Low-Skill 27.9 27.4 30.1 26.9 28.1 27.5 29.1
Medium-Skill 44.4 42.9 50.7 43.6 45.3 45.6 41.9
High-Skill 27.7 29.7 19.1 29.5 26.6 26.8 29.0
Source: Lewin calculations using the CPS Outgoing Rotation Groups.

Comparison of NISP Payroll Across Regions and Time using the CPS

We used the NISP to convert payroll by industry to payroll by occupation in our study regions. This conversion methodology is similar to the one we employed to convert employment by industry to employment by occupation, with one exception: we only had the 1998 NISP for payroll, whereas we had the NISP for employment for all years. Therefore, by using the 1998 NISP to convert payroll by industry to payroll by occupation for 1993, 1996, and 1998, we were implicitly making the assumption that the distribution of payroll by occupation in an industry did not change by region or by year. We used the CPS March Supplement for 1994, 1996, and 1998 to test the plausibility of this assumption. We compared the distribution of payroll by occupation by skill level in an industry for urban (MSA) versus rural (non-MSA) areas and for four geographic regions (Northeast, Midwest, South, and West) across the three years. We found that the distribution of payroll by skill level in one-digit industries was consistent across four regions and urban versus rural areas for most industries.

Examining the CPS employment and payroll shares over time, we found that the share of low-skill employment was relatively constant over time (about 28 percent of employment was low-skill), while the percent of payroll declined over time (the share of low-skill payroll declined from 18 to 16 percent). This implied that the gap between low-skill and high-skill wages increased over time. Examining the NISP, we found a slight increase in low-skill employment over time (from 41 to 42 percent). Therefore, by applying the 1998 NISP wage distribution to the earlier years, and not allocating a smaller share of payroll in the earlier years (because the share of low-skill employment was smaller), we were implicitly assuming that the gap between low-skill and high-skill wages increased over time, similar to the CPS analysis.

Exhibit B.1
Percent Distribution of Payroll, 1994
By Industry, Skill Category, and Region
Industry Skill Level U.S. MSA Non-MSA Northeast Midwest South West
Agriculture Low-Skill 50.9 49.8 52.5 39.2 43.9 49.5 62.0
Medium-Skill 25.2 23.5 28.2 35.1 24.7 25.6 21.6
High-Skill 23.9 26.7 19.3 25.7 31.5 24.9 16.4
Mining Low-Skill 9.1 8.3 10.2 25.5 8.3 7.7 8.8
Medium-Skill 50.0 36.6 71.1 46.7 50.0 45.5 64.0
High-Skill 40.9 55.0 18.7 27.8 41.8 46.8 27.2
Construction Low-Skill 14.5 13.3 19.8 14.2 13.6 15.2 14.5
Medium-Skill 61.1 60.1 64.9 62.5 65.2 60.1 57.3
High-Skill 24.4 26.6 15.3 23.3 21.2 24.7 28.2
Manufacturing Low-Skill 14.0 12.8 18.7 12.3 16.9 13.6 11.7
Medium-Skill 49.9 47.0 61.9 46.4 52.3 53.0 45.2
High-Skill 36.1 40.3 19.5 41.3 30.9 33.3 43.2
Transportation Low-Skill 33.1 31.9 39.3 32.7 36.1 31.9 32.0
Medium-Skill 39.8 38.9 44.9 36.8 38.5 42.1 40.6
High-Skill 27.1 29.2 15.8 30.5 25.3 26.0 27.4
Wholesale Trade Low-Skill 17.4 16.4 23.4 18.3 19.9 15.3 16.8
Medium-Skill 64.0 63.9 64.0 60.4 63.2 67.2 62.8
High-Skill 18.7 19.7 12.7 21.3 16.9 17.5 20.4
Retail Trade Low-Skill 34.5 34.4 34.7 34.3 35.5 33.6 34.7
Medium-Skill 54.0 53.5 56.6 54.9 51.2 55.4 54.1
High-Skill 11.6 12.1 8.7 10.8 13.3 11.0 11.1
Finance Low-Skill 12.3 12.0 15.5 11.4 13.4 12.7 11.5
Medium-Skill 27.5 27.1 30.9 24.3 29.6 26.8 30.4
High-Skill 60.3 60.9 53.6 64.3 56.9 60.5 58.1
Services Low-Skill 12.7 12.5 13.8 12.0 12.2 12.2 14.4
Medium-Skill 31.9 31.7 33.2 30.9 32.4 34.3 29.3
High-Skill 55.4 55.9 53.0 57.1 55.4 53.5 56.3
Public Low-Skill 10.8 10.6 11.8 10.3 11.6 11.5 9.6
Medium-Skill 60.3 59.5 64.7 62.7 66.0 56.0 60.7
High-Skill 28.8 29.9 23.5 27.0 22.4 32.5 29.8
Total Low-Skill 17.6 17.0 21.0 16.4 18.8 17.4 17.9
Medium-Skill 43.3 42.0 50.1 40.8 44.8 45.1 41.4
High-Skill 39.1 41.0 28.8 42.8 36.3 37.5 40.7
Source: Lewin calculations using the CPS March Supplement.

Exhibit B.2
Percent Distribution of Payroll, 1996
By Industry, Skill Category, and Region
Industry Skill Level U.S. MSA Non-MSA Northeast Midwest South West
Agriculture Low-Skill 42.7 47.2 36.7 37.4 45.0 38.0 47.6
Medium-Skill 34.1 25.4 45.6 22.3 34.5 42.2 28.9
High-Skill 23.2 27.4 17.7 40.2 20.5 19.8 23.5
Mining Low-Skill 10.1 8.1 12.4 1.9 12.3 12.4 6.1
Medium-Skill 55.8 44.7 68.7 67.6 60.4 57.3 47.5
High-Skill 34.1 47.2 18.9 30.5 27.2 30.3 46.5
Construction Low-Skill 11.7 11.3 13.9 13.3 11.0 12.6 9.7
Medium-Skill 59.5 57.7 69.5 60.7 63.8 58.0 56.6
High-Skill 28.8 31.0 16.6 26.0 25.1 29.4 33.7
Manufacturing Low-Skill 13.3 12.2 18.6 10.6 16.1 12.6 12.7
Medium-Skill 47.5 44.1 63.5 42.6 51.1 51.8 40.8
High-Skill 39.2 43.8 17.9 46.8 32.8 35.6 46.6
Transportation Low-Skill 33.5 32.3 40.7 30.8 40.2 31.6 32.2
Medium-Skill 37.3 36.8 40.5 38.9 34.3 38.0 37.7
High-Skill 29.2 30.9 18.9 30.3 25.5 30.4 30.1
Wholesale Trade Low-Skill 16.8 15.4 27.4 17.3 14.6 17.1 18.5
Medium-Skill 63.9 64.2 61.9 63.0 70.8 61.7 59.9
High-Skill 19.2 20.4 10.7 19.7 14.5 21.2 21.6
Retail Trade Low-Skill 33.1 32.6 36.5 33.2 32.0 33.8 33.3
Medium-Skill 55.8 56.0 54.4 56.1 56.2 54.1 57.6
High-Skill 11.1 11.4 9.1 10.7 11.8 12.1 9.0
Finance Low-Skill 10.6 10.2 14.4 8.4 11.2 12.6 10.3
Medium-Skill 24.8 24.3 30.0 19.7 22.7 25.8 33.8
High-Skill 64.6 65.4 55.6 72.0 66.1 61.7 55.9
Services Low-Skill 11.1 11.0 12.1 10.9 10.5 10.3 13.0
Medium-Skill 28.5 28.0 32.0 27.4 28.8 29.2 28.1
High-Skill 60.5 61.1 55.9 61.6 60.7 60.5 58.9
Public Low-Skill 10.3 9.6 14.0 9.0 11.7 9.7 11.0
Medium-Skill 59.0 59.2 57.7 63.8 59.1 58.1 56.7
High-Skill 30.7 31.1 28.2 27.2 29.2 32.1 32.3
Total Low-Skill 16.4 15.6 20.7 14.7 17.3 16.2 17.1
Medium-Skill 41.2 39.8 49.4 38.3 42.9 42.4 40.1
High-Skill 42.5 44.6 29.9 47.0 39.9 41.3 42.8
Source: Lewin calculations using the CPS March Supplement.

Exhibit B.3
Percent Distribution of Payroll, 1998
By Industry, Skill Category, and Region
Industry Skill Level U.S. MSA Non-MSA Northeast Midwest South West
Agriculture Low-Skill 56.5 61.6 47.1 58.1 37.1 57.6 64.3
Medium-Skill 19.5 20.8 17.2 16.6 24.3 21.2 16.7
High-Skill 24.0 17.7 35.7 25.3 38.7 21.2 19.0
Mining Low-Skill 6.0 6.2 5.8 34.0 4.1 4.5 7.1
Medium-Skill 56.8 41.5 78.4 64.0 63.7 51.4 68.5
High-Skill 37.2 52.3 15.8 2.0 32.2 44.1 24.4
Construction Low-Skill 11.5 10.5 16.6 9.1 13.4 10.9 12.4
Medium-Skill 62.5 62.1 64.5 65.1 66.3 59.1 62.0
High-Skill 26.0 27.4 18.8 25.8 20.4 29.9 25.6
Manufacturing Low-Skill 12.7 11.7 16.6 10.7 14.8 13.3 10.3
Medium-Skill 48.7 45.9 60.6 44.8 50.1 53.3 43.5
High-Skill 38.6 42.3 22.8 44.5 35.2 33.4 46.2
Transportation Low-Skill 34.0 33.2 39.7 35.1 35.1 33.6 32.6
Medium-Skill 34.3 33.6 38.9 36.3 36.0 33.6 31.9
High-Skill 31.7 33.2 21.4 28.7 28.9 32.8 35.5
Wholesale Trade Low-Skill 15.8 15.4 19.2 14.2 17.3 13.8 18.3
Medium-Skill 62.0 60.7 73.1 60.4 65.9 63.3 57.5
High-Skill 22.2 23.9 7.8 25.4 16.7 22.8 24.2
Retail Trade Low-Skill 33.6 33.0 37.1 31.6 32.4 34.1 35.6
Medium-Skill 54.8 54.6 55.6 56.2 54.8 55.5 52.5
High-Skill 11.6 12.4 7.3 12.2 12.8 10.4 12.0
Finance Low-Skill 10.6 10.0 18.9 8.3 12.3 10.7 12.2
Medium-Skill 22.9 23.0 21.3 20.4 25.2 24.7 21.5
High-Skill 66.4 66.9 59.8 71.3 62.5 64.6 66.3
Services Low-Skill 11.2 11.0 12.9 10.9 10.0 10.9 13.0
Medium-Skill 27.9 27.3 32.5 29.7 26.6 27.5 28.1
High-Skill 60.9 61.7 54.5 59.4 63.3 61.6 59.0
Public Low-Skill 11.7 11.9 10.2 10.8 13.8 11.2 11.6
Medium-Skill 58.4 56.7 68.2 63.5 60.3 54.3 59.2
High-Skill 29.9 31.3 21.7 25.7 25.9 34.5 29.2
Total Low-Skill 16.4 15.9 20.1 14.9 16.5 16.6 17.6
Medium-Skill 40.3 38.9 48.9 39.1 41.7 41.0 38.7
High-Skill 43.3 45.2 31.0 46.0 41.8 42.3 43.7
Source: Lewin calculations using the CPS March Supplement.

Comparison of Low-Skill Employment Estimates Calculated using ES-202/NISP and OES

As discussed in Chapter 4, to estimate the low-skill employment for each region and year we converted industry employment to occupational employment by using the National Industry Staffing Patterns (NISP), a data file that includes occupational employment within each industry. This file is based on a national sample only. We took the following steps:

  1. Calculated the percent of employment within each of the 801 occupations for each two-digit industry from the NISP. For example, we estimated the percent of all workers in the Legal Services industry who are file clerks (2 percent in 1998).
  2. Applied the percent distribution calculated in (1) to the industry employment for each region (from ES-202) to estimate number employed by occupation.
  3. Classified employment by skill level (low-skill, medium-skill, and high-skill) depending on the education and training requirements of the occupation.

By applying this national percent distribution to our regional industry employment data, we were assuming that there were no regional differences between the occupational distribution for each industry. To test the validity of this approach, we compared the occupation estimates generated from the ES-202 and NISP with estimates from the Occupational Employment Survey (OES). The OES data are only available for 1998, which precluded us from using the data for the analysis. Also, the OES excludes agricultural, fishing and forestry and private household industries from the survey.

Exhibit C.1 shows a comparison of the percentage of 1998 employment by skill level from the ES-202 merged with the NISP and the OES for selected regions (MSAs only). As this exhibit shows, the percentage of employment that was low-skill was slightly lower from ES-202/NISP than from OES for all regions except Vermont.

Exhibit C.1
Percent of Employment By Skill Level
Region Skill Level ES-202/NISP (%) OES (%)
Decatur and Florence, Alabama Low-Skill 41 47
Medium-Skill 41 37
High-Skill 18 16
Joplin, Missouri Low-Skill 46 50
Medium-Skill 38 37
High-Skill 16 13
Jamestown, New York Low-Skill 43 50
Medium-Skill 38 32
High-Skill 19 18
Medford-Ashland, Oregon Low-Skill 47 49
Medium-Skill 35 34
High-Skill 18 17
Florence, South Carolina Low-Skill 41 50
Medium-Skill 39 35
High-Skill 20 16
Vermont Low-Skill 43 42
Medium-Skill 36 38
High-Skill 21 20
Eau Claire, Wisconsin Low-Skill 43 50
Medium-Skill 38 31
High-Skill 19 19
Wassau, Wisconsin Low-Skill 42 46
Medium-Skill 40 35
High-Skill 18 19
Source: Lewin calculations using ES-202, NISP, OES, and BLS education and training requirements data.

Sensitivity Of Changes In Wages ANd Employment To Alternative Elasticity Assumptions

The elasticity assumptions were instrumental in determining the size of the demand and supply shifts from the employment and wage data that we collected. Therefore, we used alternative labor demand and labor supply elasticities to test the sensitivity of our results to the elasticity assumptions. In the report, we used a supply elasticity of 0.3 and a demand elasticity of 0.4. We used three sets of alternative elasticity assumptions to conduct the sensitivity analysis. We incremented the assumed elasticities by 0.1, we decremented the assumed elasticities by 0.1, and we used a supply elasticity of zero.

Exhibits D.1 and D.2 present the decomposition of the change in employment into the change due to the demand shift and the change due to the supply shift under the three different elasticity assumptions. Change in employment due to the shift in demand and shift in supply was not sensitive to small changes in the elasticity assumptions where elasticities were incremented or decremented by 0.1, but was sensitive to large changes in the elasticity assumptions. With a supply elasticity of zero, all the change in employment was due to the shift in supply; the shift in demand had no effect on employment. The average percent change in employment attributable to supply increased from 2.1 percent to 5.6 percent in the 1996 to 1998 period. However, we did not believe that a supply elasticity of zero was plausible. Therefore, we felt confident that the employment findings were robust to the elasticity assumptions.

Exhibits D.3 and D.4 present the decomposition of the change in wages into the change due to the demand shift and the change due to the supply shift under the three different elasticity assumptions. We found that the decomposition of the change in wages was more sensitive to the elasticity assumptions than the decomposition of the change in employment. The decomposition depended on the sum of the demand and supply elasticities. A higher sum decreased the percentage change in wages attributable to the shift in demand (or supply). The change in wages became smaller, because the demand and supply curves were more elastic; i.e., they were more responsive to changes in wages. A smaller change in wages was needed to bring about a change in employment. Hence, when we incremented the elasticity assumptions by 0.1, the percentage change in wages attributable to demand decreased from 9 to 7 percent. The reverse was true for a lower sum of the demand and supply elasticities. When we decremented the elasticity assumptions by 0.1, the percentage change in wages attributable to demand increased from 9 to 12 percent. A supply elasticity of zero increased the percentage change in wages attributable to demand to 16 percent; however, as discussed above, we do not believe this elasticity assumption was plausible.

Based on the sensitivity analysis, we concluded that our employment findings were robust to the alternative elasticity assumptions, but our wage findings were not. However, our basic findings are not affected much by reasonable changes in the elasticities as a result of the small size of the increase in employment due to welfare reform relative to the low-skill labor market.

Exhibit D.1
Percent Change in Employment, 1993-1996
Region Overall Demand=0.3
Supply=0.4
Demand=0.4
Supply=0.5
Demand=0.2
Supply=0.3
Demand=0.3
Supply=0
Demand Supply Demand Supply Demand Supply Demand Supply
Decatur and Florence, Alabama 9.2 5.1 4.1 5.0 4.3 5.4 3.8 0.0 9.2
Rural Mississippi 13.5 7.9 5.6 7.7 5.7 8.2 5.2 0.0 13.5
Joplin, Missouri 10.8 7.3 3.5 7.5 3.3 7.3 3.5 0.0 10.8
Southeast Missouri 9.3 5.4 3.9 5.2 4.0 5.6 3.7 0.0 9.3
Jamestown, New York 1.9 0.6 1.3 0.4 1.4 0.8 1.1 0.0 1.9
North Country, New York -0.3 -0.2 0.0 -0.2 0.0 -0.2 -0.1 0.0 -0.3
Medford-Ashland, Oregon 12.8 7.0 5.8 6.7 6.1 7.5 5.3 0.0 12.8
Central Oregon 14.0 7.4 6.7 7.0 7.1 8.0 6.1 0.0 14.0
Florence, South Carolina 8.1 4.6 3.6 4.4 3.7 4.8 3.3 0.0 8.1
Vermont 7.4 3.9 3.5 3.7 3.8 4.2 3.2 0.0 7.4
Eau Claire, Wisconsin 15.5 8.6 6.9 8.3 7.2 9.1 6.4 0.0 15.5
Wausau, Wisconsin 8.6 5.1 3.5 5.0 3.6 5.3 3.3 0.0 8.6
Average 9.2 5.2 4.0 5.1 4.2 5.5 3.7 0.0 9.2
United States 8.7 5.0 3.8 4.8 3.9 5.2 3.5 0.0 8.7
Source: Lewin calculations using ES-202, NISP, OES, and BLS education and training requirements data.

Exhibit D.2
Percent Change in Employment, 1996-1998
Region Overall Demand=0.3
Supply=0.4
Demand=0.4
Supply=0.5
Demand=0.2
Supply=0.3
Demand=0.3
Supply=0
Demand Supply Demand Supply Demand Supply Demand Supply
Decatur and Florence, Alabama 2.5 1.6 0.9 1.6 0.9 1.6 0.9 0.0 2.5
Rural Mississippi 6.8 4.2 2.6 4.2 2.6 4.3 2.5 0.0 6.8
Joplin, Missouri 7.9 4.8 3.2 4.7 3.2 4.9 3.0 0.0 7.9
Southeast Missouri 4.8 2.9 1.9 2.8 1.9 3.0 1.8 0.0 4.8
Jamestown, New York 2.5 1.9 0.6 2.0 0.4 1.8 0.6 0.0 2.5
North Country, New York 4.5 3.0 1.6 3.0 1.5 3.0 1.5 0.0 4.5
Medford-Ashland, Oregon 8.1 5.0 3.1 4.9 3.1 5.1 3.0 0.0 8.1
Central Oregon 7.5 4.6 2.9 4.5 3.0 4.7 2.8 0.0 7.5
Florence, South Carolina 6.4 3.8 2.6 3.7 2.7 3.9 2.5 0.0 6.4
Vermont 5.0 3.3 1.7 3.3 1.6 3.3 1.7 0.0 5.0
Eau Claire, Wisconsin 4.4 3.6 0.8 3.8 0.6 3.4 1.0 0.0 4.4
Wausau, Wisconsin 7.4 4.5 2.8 4.5 2.9 4.7 2.7 0.0 7.4
Average 5.6 3.6 2.1 3.6 2.0 3.6 2.0 0.0 5.6
United States 7.1 4.7 2.4 4.8 2.3 4.7 2.4 0.0 7.1
Source: Lewin calculations using ES-202, NISP, OES, and BLS education and training requirements data.

Exhibit D.3
Percent Change in Wages, 1993-1996
Region Overall Demand=0.3
Supply=0.4
Demand=0.4
Supply=0.5
Demand=0.2
Supply=0.3
Demand=0.3
Supply=0
Demand Supply Demand Supply Demand Supply Demand Supply
Decatur and Florence, Alabama -0.8 12.8 -13.7 9.9 -10.7 18.1 -19.0 22.5 -23.1
Rural Mississippi 1.2 19.7 -18.6 15.5 -14.3 27.4 -26.2 34.5 -33.7
Joplin, Missouri 6.6 18.2 -11.6 14.9 -8.3 24.2 -17.6 31.9 -26.9
Southeast Missouri 0.4 13.4 -13.0 10.5 -10.1 18.7 -18.3 23.5 -23.2
Jamestown, New York -2.7 1.5 -4.2 0.9 -3.6 2.6 -5.4 2.6 -4.7
North Country, New York -0.4 -0.6 0.1 -0.5 0.1 -0.7 0.3 -1.0 0.7
Medford-Ashland, Oregon -1.8 17.5 -19.3 13.4 -15.2 24.8 -26.7 30.6 -32.0
Central Oregon -3.8 18.4 -22.2 13.9 -17.7 26.6 -30.3 32.3 -35.1
Florence, South Carolina -0.5 11.4 -11.9 8.8 -9.3 16.1 -16.5 20.0 -20.3
Vermont -2.0 9.8 -11.8 7.4 -9.4 14.1 -16.1 17.1 -18.6
Eau Claire, Wisconsin -1.5 21.5 -22.9 16.5 -18.0 30.3 -31.8 37.5 -38.6
Wausau, Wisconsin 1.0 12.8 -11.8 10.0 -9.0 17.7 -16.7 22.3 -21.6
Average -0.4 13.0 -13.4 10.1 -10.5 18.3 -18.7 22.8 -23.1
United States -0.1 12.5 -12.5 9.7 -9.8 17.5 -17.5 21.8 -21.9
Source: Lewin calculations using ES-202, NISP, OES, and BLS education and training requirements data.

Exhibit D.4
Percent Change in Wages, 1996-1998
Region Overall Demand=0.3
Supply=0.4
Demand=0.4
Supply=0.5
Demand=0.2
Supply=0.3
Demand=0.3
Supply=0
Demand Supply Demand Supply Demand Supply Demand Supply
Decatur and Florence, Alabama 1.1 4.1 -3.0 3.3 -2.2 5.5 -4.4 7.1 -6.3
Rural Mississippi 1.9 10.6 -8.7 8.5 -6.5 14.4 -12.5 18.5 -17.1
Joplin, Missouri 1.4 11.9 -10.5 9.4 -8.0 16.4 -15.0 20.9 -19.8
Southeast Missouri 0.8 7.2 -6.4 5.7 -4.9 9.9 -9.1 12.5 -11.9
Jamestown, New York 3.0 4.8 -1.8 4.1 -1.1 6.1 -3.2 8.4 -6.2
North Country, New York 2.3 7.5 -5.2 6.1 -3.8 10.0 -7.7 13.0 -11.3
Medford-Ashland, Oregon 2.1 12.4 -10.3 9.9 -7.8 17.0 -14.9 21.8 -20.2
Central Oregon 1.7 11.4 -9.8 9.1 -7.4 15.7 -14.0 20.0 -18.7
Florence, South Carolina 0.6 9.4 -8.7 7.4 -6.7 13.0 -12.4 16.4 -15.9
Vermont 2.5 8.2 -5.7 6.6 -4.1 10.9 -8.4 14.3 -12.4
Eau Claire, Wisconsin 6.2 9.0 -2.8 7.7 -1.5 11.3 -5.1 15.7 -11.1
Wausau, Wisconsin 1.9 11.3 -9.5 9.0 -7.2 15.5 -13.6 19.9 -18.5
Average 2.1 9.0 -6.9 7.2 -5.1 12.1 -10.0 15.7 -14.1
United States 3.8 11.8 -7.9 9.6 -5.7 15.7 -11.9 20.6 -17.7
Source: Lewin calculations using ES-202, NISP, OES, and BLS education and training requirements data.

Employment and Wages By Skill LeveL

Exhibit E.1
Employment by Skill Level
Region Low-Skill Medium-Skill High-Skill
1993 1996 1998 1993 1996 1998 1993 1996 1998
Decatur and Florence, Alabama 40,062 43,939 45,057 43,576 45,386 44,529 18,580 19,823 20,289
Rural Mississippi 268,548 307,264 328,993 273,053 289,710 294,511 121,396 133,554 133,796
Joplin, Missouri 28,757 32,028 34,672 25,127 26,826 28,291 10,082 11,259 11,958
Southeast Missouri 69,670 76,436 80,174 67,339 70,713 72,219 29,248 32,666 34,007
Jamestown, New York 23,145 23,581 24,171 21,974 21,702 21,487 10,342 10,456 10,632
North Country, New York 58,991 58,836 61,568 52,070 51,960 52,304 29,621 30,868 31,941
Medford-Ashland, Oregon 26,625 30,259 32,803 20,749 22,934 24,085 10,248 11,755 12,494
Central Oregon 20,909 24,061 25,935 16,969 19,142 20,062 8,539 9,963 10,844
Florence, South Carolina 22,023 23,888 25,461 23,236 23,361 24,140 10,383 11,779 12,762
Vermont 106,349 114,562 120,398 94,097 99,640 102,003 51,935 56,048 57,887
Eau Claire, Wisconsin 26,080 30,440 31,818 22,715 25,871 27,714 11,262 12,980 13,685
Wausau, Wisconsin 22,685 24,730 26,625 22,552 24,905 25,657 9,835 10,978 11,442
United States 44,774,785 48,865,069 52,450,640 42,163,600 44,657,465 45,974,071 22,484,157 24,440,577 25,758,811
Source: Lewin calculations using ES-202, NISP, OES, and BLS education and training requirements data.

Exhibit E.2
Annual Wages by Skill Level
Region Low-Skill Medium-Skill High-Skill
1993 1996 1998 1993 1996 1998 1993 1996 1998
Decatur and Florence, Alabama 16,355 16,223 16,405 26,006 26,771 27,488 42,018 41,691 42,261
Rural Mississippi 14,453 14,623 14,910 21,067 22,220 23,737 35,258 35,872 37,991
Joplin, Missouri 15,447 16,508 16,743 22,510 23,732 24,614 37,743 38,458 39,979
Southeast Missouri 13,454 13,511 13,617 19,833 20,974 21,694 33,241 33,695 34,336
Jamestown, New York 15,638 15,218 15,676 25,170 25,258 26,615 40,542 39,642 41,128
North Country, New York 15,519 15,455 15,810 25,532 26,092 27,928 41,167 40,057 41,428
Medford-Ashland, Oregon 16,046 15,754 16,087 25,209 25,456 26,494 40,365 40,034 41,586
Central Oregon 16,954 16,328 16,603 25,434 25,185 26,578 38,803 38,487 40,200
Florence, South Carolina 15,344 15,273 15,373 24,671 25,302 26,687 41,158 40,951 42,332
Vermont 16,446 16,115 16,528 26,046 26,414 28,371 43,522 42,802 44,503
Eau Claire, Wisconsin 14,728 14,513 15,442 24,359 24,543 26,109 40,783 39,958 42,915
Wausau, Wisconsin 16,435 16,599 16,912 26,695 27,270 28,587 43,517 44,301 46,668
United States 18,668 18,654 19,380 30,064 30,918 33,324 51,078 51,438 55,069
Source: Lewin calculations using ES-202, NISP, OES, and BLS education and training requirements data.
Note: Wages are in constant 1998 dollars.

Results for Jackson, Tennessee

Initially, we planned on including Jackson, Tennessee as another region in our study. However, data quality issues precluded our use of Jackson as a region. In examining ES-202 data for Jackson, we found a sharp increase in total employment, especially between 1996 and 1998 (19.4 percent). Other sources (Census Bureau and the Tennessee labor department) reported smaller employment increases during these years. For example, the state reported an increase of 4.3 percent for this region. Given this discrepancy, we excluded the region from our analysis.

Exhibits F.1 and F.2 present the results of our economic model using BLS data for Jackson. Between 1993 and 1996, according to the data, both employment and wages increased in Jackson. Employment increased by 10.4 percent, compared to an average of 9.2 percent for the 12 regions. Wages increased by 1.5 percent compared to an average decrease of 0.4 percent for the 12 regions and 0.1 percent for the U.S. The maximum impact of welfare reform on employment was negative in this period, because Jackson experienced an increase in welfare caseloads during this time.

Exhibit F.1
Percent Change in Employment and Wages in Jackson, Tennessee, 1993-1996
  Jackson, Tennessee 12 Region Average U.S. Average
Demand Shift 10.9 9.1 8.7
Supply Shift 9.8 9.4 8.8
Change in Employment 10.4 9.2 8.7
  • Due to Demand Shift
6.2 5.2 5.0
  • Due to Supply Shift
4.2 4.0 3.7
Max Impact of Welfare Reform -0.1 0.4 0.2
Change in Wages 1.5 -0.4 -0.1
  • Due to Demand Shift
15.6 13.0 12.5
  • Due to Supply Shift
-14.1 -13.4 -12.5
Max Impact of Welfare Reform 0.4 -1.4 -0.8
Source: Lewin calculations using ES-202, NISP, OES, and BLS education and training requirements data.

Between 1996 and 1998, analysis of the BLS data found that both employment and wages increased in Jackson, Tennessee and the increases were much higher than the average increase in the 12 regions and in the U.S. Employment increased 19.4 percent compared to an average of 5.6 percent in the 12 regions. Wages increased 6.6 percent compared to an average of 2.1 percent in the 12 regions. The magnitude of the supply and demand shifts are substantially larger than in the other regions and the U.S. We believe these findings are unreliable for the 1996 to 1998 period given the unreliability of the employment data. Therefore, we decided not to report the findings in the full report.

Exhibit F.2
Percent Change in Employment and Wages in Jackson, Tennessee, 1996-1998
  Jackson, Tennessee 12 Region Average U.S. Average
Demand Shift 21.4 6.3 8.2
Supply Shift 16.8 4.8 5.6
Change in Employment 19.4 5.6 7.1
  • Due to Demand Shift
12.2 3.6 4.7
  • Due to Supply Shift
7.2 2.1 2.4
Max Impact of Welfare Reform 1.0 0.6 0.7
Change in Wages 6.6 2.1 3.8
  • Due to Demand Shift
30.6 9.0 11.8
  • Due to Supply Shift
-23.9 -6.9 -7.9
Max Impact of Welfare Reform -3.2 -2.0 -2.5
Source: Lewin calculations using ES-202, NISP, OES, and BLS education and training requirements data

Endnotes

(1) We use the year 1994 because CPS made significant changes in design starting in 1994 and we do not feel estimates from 1993 are comparable to estimates from later years. [ Back to text ]

Location- & Geography-Based Data
Urban Communities | Rural Communities