|Characteristic or Environmental Condition||Odds Ratios and Statistical Significance|
|For Becoming a Cycler Versus Becoming a Short-Term Recipient||For Becoming a Cycler Versus Becoming a Long-Term Recipient|
|Ongoing recipient at sample intake||0.962||0.853 **|
|Total months of welfare receipt during two years prior to sample intake||1.014 **||0.991 **|
|Average monthly welfare grant during year prior to sample intake||0.946 **||0.945 **|
|Total months of food stamp receipt during two years prior to sample intake||1.027 **||0.996 **|
|Youngest child is less than 6 years old||1.285 **||1.009|
|Had a child as a teenager||1.026||1.293 **|
|Number of children||1.173 **||0.881 **|
|Female||1.218 **||0.801 **|
|Age 18 to 24 years||3.320 **||1.961 **|
|Age 25 to 34||2.028 **||1.902 **|
|Age 35 to 44 years||1.469 **||1.366 **|
|Black||1.642 **||0.901 **|
|No high school diploma or GED(c)||1.375 **||0.848 **|
|Missing high school diploma or GED variable||1.134||1.276 **|
|Total employment in county or region during month of sample intake (in 10,000's)||1.088 **||1.186 **|
|Total earnings during year prior to sample intake (in $1,000's)||0.998||1.032 **|
|Average unemployment rate during year 1||0.970||0.935 **|
|Percentage change in unemployment rate from month of sample intake to month 48||1.001||1.001|
|Welfare caseload in year 1 (in 10,000's)||1.069||0.974|
|Percentage change in welfare caseload from year 1 to year 4||1.012||1.010|
|Total employment in county or region during month of sample intake||0.990 **||0.983 **|
|Percentage change in number employed from month of sample intake to month 48||1.001||1.004|
|In the Connecticut Jobs First study||2.008||0.101 **|
|In the Vermont WRP study||5.948||0.124 **|
|In the UC-Cleveland study||1.626 **||0.725 **|
|In the UC-Philadelphia study||1.048||0.196 **|
|Randomly assigned in quarter 2, 1994||0.936||0.702 **|
|Randomly assigned in quarter 3, 1994||1.052||1.249 **|
|Randomly assigned in quarter 4, 1994||0.989||1.120|
|Randomly assigned in quarter 1, 1995||1.180||1.326 **|
|Randomly assigned in quarter 2, 1995||1.071||1.113|
|Randomly assigned in quarter 3, 1995||1.058||1.188|
|Randomly assigned in quarter 4, 1995||1.009||1.268|
|Randomly assigned in quarter 1, 1996||1.114||1.531 **|
|Randomly assigned in quarter 2, 1996||1.115||1.436 **|
|Randomly assigned in quarter 3, 1996||1.482 **||2.376 **|
|Randomly assigned in quarter 4, 1996||1.051||1.458 **|
|Randomly assigned in quarter 1, 1997||1.543||1.686 **|
|Randomly assigned in quarter 2, 1997||1.680 **||1.985 **|
|Randomly assigned in quarter 3, 1997||2.191 **||2.689 **|
|Randomly assigned in quarter 4, 1997||2.293 **||2.783 **|
| Sources: MDRC calculations from state and county administrative records and Background Information Forms.
Notes: The full sample includes 10,393 cyclers, 62,388 short-term recipients, and 88,226 long-term recipients. Effects were estimated with logistic regression with "Cycler" as the dependent variable. Separate regressions were run for samples of cyclers and short-term recipients and for samples of cyclers and long-term recipients. The samples were equally weighted by site.
"**" indicates statistical significance at the 0.05 level or smaller.
(a): Estimation of the effect is unreliable because fewer than 5 sample members in Florida FTP and Vermont WRP are cyclers and Hispanic.
(b): Estimation of the effect is unreliable because fewer than 10 sample members in Connecticut Jobs First, Florida FTP, and Vermont WRP are cyclers and belong to the "other" ethnicity category.
(c): Calculations are for sample members in Connecticut Jobs First, Florida FTP, and Vermont WRP evaluations only.
Key Terms and Sample Definitions Used in This Report
PRWORA. The 1996 federal Personal Responsibility and Work Opportunity Reconciliation Act, which re-placed Aid to Families with Dependent Children (AFDC) with a flexible, state-directed block grant pro-gram, Temporary Assistance for Needy Families (TANF); set lifetime limits on eligibility to receive TANF payments; and created financial incentives for states to run mandatory work-focused welfare-to-work pro-grams.
Evaluation Sites. Connecticut Jobs First, Florida FTP, and Vermont WRP. In these localities, MDRC con-ducted an experimental study of a welfare reform initiative, based on random assignment of sample mem-bers to program and control groups.
Urban Change Sites. Cleveland (Cuyahoga County) and Philadelphia (Philadelphia County). These locali-ties are included in MDRC's ongoing Project on Devolution and Urban Change, a non-experimental analy-sis of the effects of PRWORA on welfare caseloads and recipient behavior in four major urban centers.
Welfare Payment. An AFDC or TANF payment.
Sample Intake Period. For evaluation sites, the months during which sample members were randomly as-signed to a program or control group. For Urban Change sites, a designated range of months, during which each sample member's first welfare payment was recorded.
Month of Sample Intake. Varies by sample member. For evaluation sites, a sample member's month of random assignment. For Urban Change sites, a sample member's first month of welfare receipt during the sample intake period.
Welfare Spell. A series of months of welfare payments-consecutive, or with interruptions that lasted for only one month. A new welfare spell was recorded when a sample member began receiving welfare after at least two consecutive months with no payments.
First Welfare Spell. The welfare spell that included the month of sample intake, or which began one or two months later. Individuals with no welfare payments or whose first welfare spell began during a later month were excluded from the sample for this report.
Observation Period. A four-year (48 month) period that began with each sample member's month of sam-ple intake, during which her welfare spells and other outcomes were recorded. The observation period in-cludes different calendar months for each sample member.
Fifth-Year Follow-Up Period. Months 49 through 60, following each sample member's month of sample intake. Data on welfare receipt and other outcomes were available for most, but not all, sample members during these months.
Cyclers. Sample members with three or more welfare spells during the (four-year) observation period.
Short-Term Recipients. Sample members with one or two welfare spells and a total of 1 to 24 months of welfare receipt during the (four-year) observation period.
Long-Term Recipients. Sample members with one or two welfare spells and a total of 25 to 48 months of welfare receipt during the (four-year) observation period.
Ongoing Recipients. Sample members whose first welfare spell began at least four months prior to their month of sample intake.
New Recipients. Sample members whose first welfare spell began no earlier than three months prior to their month of sample intake.
We conducted a multivariate analysis of the probability of becoming a cycler to better understand the factors that affect cycling. Table 6 presents the results from this analysis. The first column of the table provides the odds ratios of the likelihood that a recipient becomes a cycler within the four-year observation period as opposed to becoming a short-term recipient.(26) To interpret the odds rations, note that ratios greater than 1 indicate a positive effect on cycling, and ratios less than one indicate a negative effect. For example, the 1.014 in column one indicates that each additional month of welfare receipt during the two years prior to sample intake increases the likelihood that a respondent will be a cycler as opposed to a short-term recipient by 1.4 percent. In contrast, the 0.946 coefficient in column one indicates that as the average monthly welfare grant increases by one dollar, the likelihood that a respondent will be a cycler as opposed to a short-term recipient decreases by 5.4 percent. Several other variables are significant in predicting the likelihood of cycling. In addition to prior welfare receipt, the number of children, being female, and the lack of a high-school diploma or GED(27) are all positively related to a recipient cycling on and off of welfare during the observation period as opposed to becoming a short-term recipient. In addition, younger recipients are significantly more likely to become cyclers than recipients older than age 45 at sample intake and black recipients are more likely to become cyclers than their white, non-Hispanic counterparts.
The second column of Table 6 shows the multivariate results for the probability of a recipient becoming a cycler as opposed to becoming a long-term recipient within the four-year observation period.(28) The factors that affect the likelihood of cycling versus short-term recipiency are the same factors that affect the likelihood of cycling versus long-term receipt. That is, many of the variables in the second column are statistically significant, although the implication differs. For example, the number of months of welfare receipt during the two years prior to sample intake is a negative predictor of cycling versus long-term receipt. Specifically, the likelihood of cycling decreases by about 1 percent for each additional month of welfare receipt. This is in contrast to the findings reported above, in which the likelihood of cycling increased about 1.4 percent for each additional month of welfare receipt when the likelihood of cycling was compared with short-term welfare receipt. Similarly, as the number of children increases, a respondent is less likely to be a cycler and more likely to be a long-term recipient.
The differences in the predictors of cycling are interesting and noteworthy. Being on welfare during sample intake is a predictor of cycling as compared with long-term recipiency, but this factor does not matter in determining the likelihood of cycling versus becoming a short-term recipient. In other words, past welfare receipt is a better predictor of whether a recipient is likely to cycle versus become a persistent user of welfare, but is not particularly helpful in distinguishing a potential cycler from a short-term recipient. Also, the results suggest that having a child less than 6 years old increases the likelihood of cycling versus short-term recipiency (by 28.5 percent), but the presence of such a young child does not affect the likelihood of cycling versus long-term recipiency. This indicates that there may be different barriers that affect cyclers and short-term recipients.