The Use of TANF Work-Oriented Sanctions in Illinois, New Jersey, and South Carolina:

Appendix A:
Methodology for Multivariate Analysis

[ Main page of Report | Contents of Report ]

Contents

• Endnotes

To examine the affect of individual characteristics on being sanctioned in each of the three study states in a multivariate context, we use a series of logit models to estimate whether specific characteristics affects whether a TANF case head is sanctioned. The analysis is based on the single-parent TANF cases from each state who had no missing data on sanction status and the selected characteristics for study.⁽⁸⁾ A list of variables included in the models and their means and standard deviations for each state are presented in Table A.1. The logit estimation results for the states are presented in Tables A.2, A.3, and A.4.

We estimate equation (1) to determine how each individual characteristic affects sanction status. This equation expresses sanction status as a function of select characteristics including gender, age, educational level, race/ethnicity, marital status, number of children on the TANF case, age of the youngest child on the TANF case, duration of the current TANF spell, and current earnings as represented in a series of 22 dummy variables.

(1) SancStat_i = ₀ + _j L_ji + _k + _i
where:

• SancStat_i = 1 if sanctioned; 0 otherwise
  L_ji = 1 if specific characteristic j is present; 0 otherwise; j = 1,…, 22
  _i = random disturbance term
  ₀, _j, _k = parameters to be estimated
  i = index for study population, i = 1,…, 33,478(IL); 51,545(NJ); 10,852(SC)

The variable SancStat represents different sanction status' throughout the models. For Illinois--results presented in Table A.2--we examine being partially sanctioned within 12 months (Model 1) and being fully sanctioned within 12 months (Model 2). We do the same for New Jersey, with results for partial sanction within 12 months (Model 3) and full sanction within 12 months (Model 4) presented in Table A.3. Due to more recent study month selected for South Carolina, we have fewer follow-up months for study. For this reason, we examine we examine being fully sanctioned within 10 months in South Carolina. Results of this model (Model 5) are presented in Table A.4.

We then turn to the use of survey data from Illinois to estimate whether more specific personal liabilities or the number of such liabilities affects whether a TANF case head is sanctioned (partially or fully) within 12 months. The analysis sample includes the 375 survey respondents who had no missing data on sanction status, selected background characteristics, or any of the 15 personal liability measures. A list of variables included in these models and their means and standard deviations are presented in Table A.5. The logit estimation results are presented in Tables A.6 and A.7.

First, we estimate equation (2) to determine how each individual personal liability affects sanction status. This equation expresses sanction status as a function of 15 personal liabilities, and the same set of background characteristics as in equation (1).

(2) ASANC12_i = ₀ + _j L_ji + _k X_ki + _i
where:

• ASANC12_i = 1 if sanctioned (partially or fully) within 12 months; 0 otherwise
  L_ji = 1 if specific liability j is present; 0 otherwise; j = 1,…, 16
  X_ki = set of background control variables, k = 1,…, K
  _i = random disturbance term
  ₀, _j, _k = parameters to be estimated
  i = index for survey respondents, i = 1,…, 375

Next, we estimate equations (3) and (4) to determine whether the number of personal liabilities affects sanction status. These equations express sanction status as a function of the number of employment liabilities and a set of background characteristics. We specify the number of barriers as a series of seven dummy variables in equation (3) and as a series of three dummy variables in equation (4).

(3) ASANC12_i = ₀ + _j N1_ji + _k X_ki + _i
where:

• ASANC12_i = 1 if sanctioned (partially or fully) within 12 months; 0 otherwise
  N1_ji = 1 if the number of personal liabilities is j; 0 otherwise; j = 1,…, 6
  N1_7i = 1 if the number of personal liabilities is 7 or more; 0 otherwise
  X_ki = set of background control variables,⁽⁹⁾ k = 1,…, K
  _i = random disturbance term
  ₀, _j, _k = parameters to be estimated
  i =index for survey respondents, i = 1,…, 375

(4) ASANC12_i = ₀ + _j N2_ji + _k X_ki + _i
where:

• ASANC12_i = 1 if sanctioned (partially or fully) within 12 months; 0 otherwise
  N2_1i = 1 if the number of personal liabilities is j; 0 otherwise; j = 1,…, 6
  N2_2i = 1 if the number of personal liabilities is 7 or more; 0 otherwise
  X_ki = set of background control variables, k = 1,…, K
  _i = random disturbance term
  ₀, _j, _k = parameters to be estimated
  i =index for survey respondents, i = 1,…, 375

Table A.1.
Variable Descriptions

Variable	Illinois		New Jersey		South Carolina
	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.
Dependent Variable
Case head has been partially sanctioned within 12 months	.23	.42	.32	.47
Case head has been fully sanctioned within 12 months	.11	.31	.14	.34
Case head has been fully sanctioned within 10 months					.05	.21
Independent Variables
Female	.98	.13	.96	.20	.98	.16
Age 20-24	.27	.44	.24	.43	.32	.47
Age 25-29	.21	.41	.19	.39	.20	.40
Age 30-39	.30	.46	.31	.46	.25	.44
Age 40 and over	.13	.34	.17	.37	.12	.33
High school diploma or GED	.40	.49	.41	.49	.50	.50
More than high school	.11	.31	.10	.30	.14	.34
White, Non-Hispanic	.12	.33	.14	.35	.26	.44
Hispanic	.06	.23	.27	.44	.01	.08
Other, Non-Hispanic	.01	.08	.02	.13	.00	.06
Separated, divorced, widowed	.13	.33	.19	.39	.26	.44
Married	.04	.19	.03	.18	.03	.17
Two children on TANF case	.28	.45	.25	.43	.34	.47
Three children on TANF case	.20	.40	.11	.31	.18	.38
Four or more children on TANF case	.21	.41	.06	.23	.10	.30
Youngest child is age 1-2	.25	.43	.18	.38	.37	.48
Youngest child is age 3-5	.17	.38	.19	.39	.22	.41
Youngest child is 6 or older	.29	.45	.45	.50	.31	.46
Current TANF spell of 6-11 months	.18	.38	.10	.30	.28	.45
Current TANF spell of 12-24 months	.21	.41	.12	.32	.15	.35
Current TANF spell of 25 months or longer	.39	.49	.27	.45	.06	.23
Earnings during quarter	.41	.49	.09	.28	.41	.49
Source: State administrative data from Illinois (N=33,478), New Jersey (N=51,545), and South Carolina (N=10,852).

Table A.2.
Effects of Specific Characteristics on the Probability that a TANF Case Head
Is Sanctioned Within 12 Months in Illinois

	Model 1 (Partial Sanction)				Model 2 (Full Sanction)
	Coefficient	Std Error	Z	P>|z|	Coefficient	Std Error	z	P>|z|
Female	-0.28	0.11	-2.61	0.01	-0.12	0.15	-0.80	0.43
Age 20-24	-0.19	0.05	-3.78	0.00	-0.15	0.07	-2.14	0.03
Age 25-29	-0.27	0.06	-4.61	0.00	-0.25	0.08	-3.01	0.00
Age 30-39	-0.29	0.06	-4.73	0.00	-0.28	0.08	-3.30	0.00
Age 40 and over	-0.37	0.07	-5.33	0.00	-0.17	0.09	-1.84	0.07
High school diploma or GED					-0.44	0.04	-10.86	0.00
More than high school					-0.51	0.07	-7.25	0.00
White, Non-Hispanic	-0.22	0.05	-4.90	0.00	-0.16	0.06	-2.54	0.01
Hispanic	-0.42	0.06	-6.71	0.00	-0.42	0.09	-4.70	0.00
Other, Non-Hispanic	-0.85	0.22	-3.93	0.00	-0.78	0.30	-2.59	0.01
Separated, divorced, widowed	-0.20	0.05	-4.17	0.00	-0.28	0.07	-4.15	0.00
Married	-0.23	0.08	-2.89	0.00	-0.15	0.11	-1.43	0.15
Two children on TANF case	-0.02	0.04	-0.47	0.64	0.08	0.05	1.57	0.12
Three children on TANF case	-0.03	0.04	-0.81	0.42	0.09	0.06	1.60	0.11
Four or more children on TANF case	-0.06	0.04	-1.31	0.19	0.02	0.06	0.39	0.70
Youngest child is age 1-2	0.09	0.04	2.50	0.01	0.39	0.05	7.60	0.00
Youngest child is age 3-5	-0.04	0.04	-1.02	0.31	0.27	0.06	4.49	0.00
Youngest child is 6 or older	-0.14	0.05	-3.01	0.00	0.22	0.06	3.43	0.00
Current TANF spell of 6-11 months	0.21	0.04	4.88	0.00	0.39	0.06	6.17	0.00
Current TANF spell of 12-24 months	0.17	0.04	4.10	0.00	0.53	0.06	8.76	0.00
Current TANF spell of 25 months or longer	-0.00	0.04	-0.12	0.91	0.48	0.06	8.25	0.00
Earnings during quarter	-0.75	0.03	-25.59	0.00	-0.59	0.04	-14.86	0.00
Constant	-0.20	0.12	-1.65	0.10	-1.97	0.17	-11.70	0.00
Chi-Square	1354.25				706.63
Prob > Chi-Square	0.0000				0.0000
Number of Observations	32,703				32,703
Source: Results of multinomial logit models predicting the probability of being sanctioned (partially or fully) within 12 months using administrative data on single-parent TANF cases in Illinois in November 2001.

Table A.3.
Effects of Specific Characteristics on the Probability that a TANF Case Head
is Sanctioned Within 12 Months in New Jersey

	Model 3 (Partial Sanction)				Model 4 (Full Sanction)
	Coefficient	Std Error	Z	P>|z|	Coefficient	Std Error	z	P>|z|
Female	-0.06	0.05	1.15	0.28	-0.10	0.07	1.89	0.17
Age 20-24	-0.13	0.04	12.22	0.00	-0.18	0.05	13.99	0.00
Age 25-29	-0.24	0.04	31.81	0.00	-0.26	0.06	22.36	0.00
Age 30-39	-0.33	0.04	54.73	0.00	-0.37	0.06	42.41	0.00
Age 40 and over	-0.63	0.05	163.06	0.00	-0.63	0.07	91.65	0.00
High school diploma or GED	-0.22	0.02	103.15	0.00	-0.22	0.03	62.50	0.00
More than high school	-0.38	0.04	110.19	0.00	-0.44	0.05	75.22	0.00
Missing education information	-0.35	0.06	32.13	0.00	-0.38	0.09	18.50	0.00
White, Non-Hispanic	-0.42	0.03	178.64	0.00	-0.51	0.05	127.39	0.00
Hispanic	-0.49	0.02	414.92	0.00	-0.41	0.03	160.79	0.00
Other, Non-Hispanic	-0.78	0.09	67.93	0.00	-0.93	0.15	39.92	0.00
Separated, divorced, widowed	-0.24	0.03	70.32	0.00	-0.35	0.04	65.77	0.00
Married	-0.31	0.06	25.67	0.00	-0.30	0.09	11.53	0.00
No children on TANF case	-0.24	0.23	1.07	0.30	-0.21	0.32	0.44	0.51
Two children on TANF case	-0.02	0.02	0.62	0.43	-0.09	0.03	7.49	0.01
Three children on TANF case	-0.02	0.03	0.35	0.55	-0.14	0.05	7.97	0.01
Four or more children on TANF case	-0.02	0.05	0.28	0.60	-0.11	0.06	2.96	0.09
Youngest child is age 1-2	-0.02	0.03	0.33	0.56	0.04	0.05	0.78	0.38
Youngest child is age 3-5	0.08	0.04	5.04	0.02	0.16	0.05	11.66	0.00
Youngest child is 6 or older	0.12	0.04	11.11	0.00	0.11	0.05	5.06	0.03
Youngest child age is missing	-0.39	0.24	2.75	0.10	0.05	0.33	0.02	0.89
Current TANF spell of 6-11 months	0.07	0.03	4.22	0.04	0.22	0.04	27.82	0.00
Current TANF spell of 12-24 months	-0.01	0.03	0.14	0.71	0.18	0.04	18.23	0.00
Current TANF spell of 25 months or longer	0.03	0.03	1.19	0.27	0.10	0.03	7.77	0.01
Earnings during month	-1.10	0.04	636.34	0.00	-0.74	0.06	156.47	0.00
Constant	-0.02	0.06	0.11	0.74	-1.13	0.09	175.33	0.00
Chi-Square	2546.57				1266.37
Prob > Chi-Square	0.0000				0.0000
Number of Observations	51,539				51,539
Source: Results of multinomial logit models predicting the probability of being sanctioned (partially or fully) within 12 months using administrative data on single-parent TANF cases in New Jersey from July 2000 to June 2001.

Table A.4.
Effects of Specific Characteristics on the Probability That a TANF Case Head is
Sanctioned Within 10 Months in South Carolina

	Model 5 (Full Sanction)
	Coefficient	Std Error	Z	P>|z|
Female	0.02	0.37	0.05	0.96
Age 20-24	-0.21	0.13	-1.59	0.11
Age 25-29	-0.55	0.16	-3.44	0.00
Age 30-39	-0.84	0.17	-4.90	0.00
Age 40 and over	-1.39	0.25	-5.52	0.00
High school diploma or GED	-0.61	0.10	-6.02	0.00
More than high school	-0.70	0.17	-4.07	0.00
White, Non-Hispanic	-0.18	0.12	-1.48	0.14
Hispanic	-1.14	1.01	-1.12	0.26
Other, Non-Hispanic	-0.67	1.02	-0.65	0.51
Separated, divorced, widowed	0.01	0.15	0.05	0.96
Married	-0.27	0.40	-0.68	0.50
Two children on TANF case	0.05	0.11	0.40	0.69
Three children on TANF case	0.14	0.13	1.05	0.30
Four or more children on TANF case	0.05	0.17	0.32	0.75
Youngest child is age 1-2	-0.05	0.16	-0.31	0.76
Youngest child is age 3-5	-0.14	0.18	-0.81	0.42
Youngest child is 6 or older	0.01	0.17	0.01	0.99
Current TANF spell of 6-11 months	0.16	0.11	1.44	0.15
Current TANF spell of 12-24 months	0.06	0.14	0.42	0.67
Current TANF spell of 25 months or longer	-0.18	0.22	-0.82	0.41
Earnings during quarter	0.04	0.10	0.44	0.66
Constant	-2.25	0.42	-5.35	0.00
Chi-Square	141.69
Prob > Chi-Square	0.0000
Number of Observations	10,789
Source: Results of multinomial logit models predicting the probability of being fully sanctioned within 10 months using administrative data on single-parent TANF cases in South Carolina in June 2002.

Table A.5.
Variable Descriptions for Illinois Survey-Based Models

Variable	Mean	Standard Deviation
Dependent Variable
Case head has been sanctioned (partially or fully) within 12 months	.27	.45
Independent Variables
Human Capital Liabilities
No high school diploma or GED	.44	.50
Fewer than four quarters of recent work experience	.59	.49
Performed fewer than four common job tasks	.28	.45
Personal Challenges
Physical health problem	.21	.41
Mental health problem	.25	.43
Chemical dependence	.03	.17
Severe physical domestic violence in past year	.13	.33
Signs of learning disability	.12	.33
Multiple arrests	.16	.37
Difficulty with English language	.02	.38
Logistical and Situational Challenges
Child/other family member/friend w/health problem or special need	.35	.48
Pregnant	.08	.80
Child under age one in household	.28	.45
Transportation barrier	.21	.54
Child care	.31	.46
Unstable housing	.23	.42
Counts of Liabilities
One	.12	.33
Two	.16	.37
Three	.21	.41
Four	.17	.37
Five	.13	.33
Six	.07	.26
Seven or more	.10	.30
Source: 2001-02 survey of Illinois TANF cases, N=416.

[ Go to Contents ]

Endnote

(8) In New Jersey, we included cases with missing data on education status and the age of the youngest child on the TANF case with appropriate dummy variables indicating missing data in these areas.

(9) The background control variables are: age, race, marital status, number of children, presence of young children, percent of time on welfare in past 25 months, county unemployment rate, neighborhood racial concentration (i.e., 80 percent or more African-American).

Where to?

Top of Page | Contents

Main Page of Report | Contents of Report

Home Pages:
Human Services Policy (HSP)
Assistant Secretary for Planning and Evaluation (ASPE)
U.S. Department of Health and Human Services (HHS)