Post on 13-Jan-2016
description
1
The 1993 EITC Expansion and Low-Skilled Single
Mothers’ Welfare Use Decision
Hau Chyi
WISE, Xiamen University
2
Motivation• Total welfare caseloads decreased to
2.3 million in 1999 from 4.1 million in 1990
3
• Previous literature attribute it to the success of welfare reform during the 1990s
• This paper intends to investigate the relative effectiveness of the EITC and the initiatives of the welfare reform
4
EITC Expansion since 1993
Source: Hotz and Scholz (2001)
5
Modeling Strategy
• Estimate a single mother’s joint probability of welfare use and work– General impression is that welfare participants
don’t work. So why is estimating joint distribution necessary?
6
Distribution of Single Mothers’ (monthly) Four States in NLSY 79
7
• The ratio increased since 1993, indicating policies may have not only decreased welfare use, but also promoted work among those who are still on welfare– Looking at marginal probability alone mask th
e compositional change– Also, estimating joint probability of work and w
elfare deals with the simultaneous decisions issue
8
Basic Estimation Framework
• Dependent Variables: – Welfare (W) and Work (H)– Four possible states: D
– We want to know how EITC and welfare reform affect a single mother’s decisions of choosing each state
Welfare
Work
No Yes
No D= 1 3
Yes 2 4
9
Estimation Strategy
• D-D estimator of the effect of the EITC program:– 1993 expansion applies only on families with
more than 2 kids (treatment group)– Difference between families with one child an
d those with more than one is substantial– Difference in behaviors between families with
2 or more kids and those with one kid is used to estimate the EITC effect
10
Source: Hotz and Scholz (2001)
11
Estimation Strategy
• Welfare reform: focus on the five year time limits– Use Grogger and Michalopoulos (2003)’s strat
egy, families with their youngest children older than 13 are not bound by this restriction
– Use the exact year and month to construct the time limit dummy
• Year- and state-fixed effects are also added to control for the business cycle and long-run state-specific factors
12
The latent variable H* can be defined similarly.
13
Assume
• Joint probability distribution can be estimated by a bivariate probit model.
• Work and welfare decisions are identified by the effective tax rate of welfare on earnings
14
Data
• National Longitudinal Survey of Youth 1979 (NLSY 79)– Follows young adults who were between 14 to
23 in 1979 in the U.S.– It has weekly information of hours of work and
monthly welfare use information since 1978– Use information from 1985 to 2000
15
• Detailed information is crucial in examining joint decision– Eligibility of the welfare program is checked ev
ery month– Annual data, such as March Current Populatio
n Survey (CPS), can make it appear that work and welfare decisions are made jointly
16
17
Parallel trend assumption
18
(Correlation Coefficient = -.711)
19
The Average Effect
20
21
22
23
24
To sum up:
1. the EITC expansion explains roughly about 38% of the welfare-caseload reduction among this group of mothers in the 1990s.
2. 42% of those who leave welfare due to the EITC expansion were already working before the expansion.
3. 37% of the increase in the (no welfare, work) alternative due to EITC expansion is from workers leaving welfare.
25
Future Studies
• How does welfare reform and EITC expansion affect single mothers’ offsprings?