Estimating the potential effects of poverty reduction policies

Abstract

States require a measure of poverty that captures all family resources net of taxes and nondiscretionary expenses and uses thresholds reflecting current needs in the state to assess the well-being of families under current and alternative policies. This paper describes the implementation of a poverty measure for the State of Connecticut based on the recommendations of the National Academy of Sciences, and it describes the potential antipoverty effects of changes in child care, adult education, and child support policies. The paper concludes with a discussion of the challenges in implementing a modern poverty measure and in simulating policy alternatives. © 2010 by the Association for Public Policy Analysis and Management.

INTRODUCTION

Over the past several years, numerous states have initiated task forces and commissions focused on policies to reduce poverty.141 States' recommendations frequently include expanding income supports, such as the Earned Income Tax Credit (EITC), child care subsidies, and nutritional assistance. Many also include education investments at both the early childhood and postsecondary levels. Some states focus solely on reducing child poverty; others consider options for reducing poverty more generally. States' interest in reducing poverty stems from their recognition of the economic costs of poverty, especially the toll that poverty exerts on children.

Most state commissions quickly recognize the need for a benchmark that would allow them to track progress in reducing poverty and to “test out” the effects of different policy proposals on poverty reduction. Such a benchmark would incorporate all components of family resources as well as an up-to-date measure of family needs in their state. The “official” measure of poverty, based solely on cash income and a national measure of need set back in the 1960s (and subsequently adjusted by changes in prices), is inadequate for these purposes.

The measure of poverty recommended by the National Academy of Sciences (NAS) in 1995 meets states' needs for a benchmark.152 The NAS measure includes all types of income, including that received in kind and through the tax system. The measure accounts for the effects of nondiscretionary work expenses and out-of-pocket health expenses on net family income. The NAS measure also uses an updated measure of the cost of basic needs and captures differences in the cost of living across the states.

This article describes the use of the NAS measure to estimate poverty in Connecticut (CT) and to simulate the effects of various policy options on the NAS poverty rate. We use two years of Current Population Survey (CPS) data representing the population in CT in 2005 and 2006 to improve the statistical reliability of the estimates. We begin by building a baseline estimate of poverty and then simulate the effects of alternative policies on poverty. We use the Urban Institute's Transfer Income Model, Version 3 (TRIM3) to implement the key features of the NAS measure for these simulations.163 The model simulates benefits from key government assistance programs, including Supplemental Security Income, Supplemental Nutritional Assistance (SNAP), Temporary Assistance for Needy Families (TANF), and Medicaid, using specific state rules in the base years to correct for underreporting in the CPS data, and then in the simulations to capture the effects of changes in these benefits on family income and poverty. The model also imputes values for in-kind resources (housing assistance, WIC, and LIHEAP) and calculates taxes (federal, state, and payroll), and child care expenses in the baseline and in the alternative simulations. The model captures the likely effects of policies on participation behavior, and it optionally simulates the potential effects of alternative policies on employment.

This article begins by describing the implementation of the NAS poverty measure for CT. Then we describe the simulation results for a few proposals that would reduce child poverty.174 These include expansions of child care subsidies, education investments that would increase the number of adults with AA degrees, and improved payment of child support. Our final section highlights some of the challenges both in implementing the NAS at the state level and in estimating the effects of policies on poverty.

MEASURING POVERTY IN CT

There is broad agreement that the official measure of poverty in the United States is flawed.185 The official measure, based on cash income, fails to take into account many of the antipoverty policies implemented over the past several decades. Also, the thresholds for measuring whether a family is poor are based on outdated data about food consumption adjusted only for changes in the Consumer Price Index (CPI). In contrast, the NAS measure of poverty includes all family resources, accounts for nondiscretionary expenses, and updates the thresholds used to measure poverty.196 Considerable research has vetted the NAS measure since 1995, and the Census Bureau annually provides estimates of poverty applying the alternative measure to the annual Current Population Survey (CPS). Many experts support the NAS measure, although there is some disagreement about the treatment of specific items such as out-of-pocket medical expenses and the value of home ownership.207

We use TRIM3 to add the information required to produce NAS poverty estimates for CT.218 These procedures, applied to two years of the CPS in CT, generally follow the consensus of the Committee on National Statistics (CNSTAT) developed during their 2004 workshop (Iceland, 2005).229 We briefly describe how we estimate resources and the poverty thresholds used for these estimates as follows.2310

Resources

The NAS poverty measure uses a broad definition of resources that approximates the net income available to a family. The NAS measure begins with cash income, adds capital gains and in-kind benefits [Supplemental Nutrition Assistance Program (SNAP) benefits,2411 housing assistance, and others], deducts federal and state income taxes (including refundable credits), and subtracts nondiscretionary expenses such as the cost of child care and transportation to work. Nondiscretionary out-of-pocket medical expenses can be taken into account either as a deduction from resources or through the thresholds used to measure poverty.

We use simulation and imputation procedures to create a NAS “baseline” poverty estimate for CT (Table 8).2512 These procedures both add information not available on the CPS and correct for underreporting of key government assistance benefits (Wheaton, 2007). TRIM3 first assigns immigrant status (documented, undocumented, and refugee) because it is required to simulate eligibility for most government programs. Then the model creates tax filing units in the CPS (based on family relationships as well as student status and income to determine dependency status) and uses a statistical match with IRS data to add realized capital gains to family income.2613

Table 8. Implementing the National Academy of Sciences (NAS) definition of poverty.

Key Data Elements for NAS Poverty Measure	Sources (Data are either reported or added using simulation and imputation procedures.)
Immigrant Status (required to determine eligibility for government benefits)	Imputations of legal and refugee status from Passel and Cohn (2006).
Cash income	Cash income
Earned income, pensions, retirement income, social security or railroad retirement, investment income, and other cash income	Reported on CPS.
Capital gains (losses)	Imputed using a statistical match with Statistics of Income Data (SOI) from the IRS.
Cash welfare: Supplemental Security Income (SSI) Temporary Assistance for Needy Families (TANF)	Program eligibility and potential benefits simulated using CT rules. Participants selected from the eligible by first choosing those that reported benefits and then stochastically selecting additional participants to match control totals derived from CT administrative data by family characteristics.
+Near-cash elements
SNAP/food stamps	Similar to cash welfare, eligibility is simulated and participants selected to meet targets.
Women Infants and Children (WIC) benefits	Program rules simulated to estimate eligibility.
	Reporters plus a random selection of eligible nonreporters assigned participation to meet CT program data.
Low-Income Home Energy Assistance (LIHEAP)
Public and subsidized housing	Program receipt reported, aligned using CT data, and valued as fair market rent minus the household's required payment.
−Taxes Federal income taxes and EITC State income taxes and EITC Payroll taxes	Form tax filing units using CPS data and tax program rules. Use 2008 federal and state tax rules and integration of state/federal income tax calculation. Itemized deductions imputed through a statistical match with IRS data. Payroll taxes use reported earnings and employment sector.
Expenses (deducted from income) Child care subsidies and expenses	Simulate CCDF subsidies and required copays using CT policies. Estimate child care expenses for nonsubsidized families using regression techniques.
Other work expenses	Use work status and Census Bureau imputations.
Health insurance status (used to select appropriate poverty threshold) Public coverage	Simulate Medicaid/SCHIP eligibility using CT program rules; participants selected from eligible reporters and nonreporters to match state data.

For most government benefit program modules, TRIM3 applies the detailed state program rules to estimate eligibility and then selects program participants. The simulated caseload includes all the eligible families or individuals who reported benefits on the CPS (if that information is available) and a portion of those eligible who did not report benefits. The procedures are designed to match the actual number and characteristics of enrollees in each state. These procedures correct for underreporting of benefits and can simulate expansions (or contractions) in each program. Such procedures are used to simulate cash welfare benefits (Supplemental Security Income and TANF) and near-cash benefits (SNAP, WIC, LIHEAP, and housing assistance). The results from each benefit simulation are verified against administrative data for CT and generally come within 90 to 95 percent of enrollees and benefits.

TRIM3 also uses program rules and household information to calculate payroll taxes and federal and state income taxes. Payroll taxes are based on individuals' work status, reported wages, and industry (since many state government workers do not pay these taxes). The federal and state income tax modules estimate taxes for each filing unit by calculating adjusted gross income, subtracting exemptions and deductions, computing tax liability, and applying both non-refundable and refundable tax credits. The CT tax simulations use the rules in effect in 2008, deflated to the years of survey data.

Finally, the model deducts nondiscretionary expenses for work and accounts for out-of-pocket health spending. As noted in Table 8, we impute child care expenses using fairly detailed procedures that model receipt of subsidies from the Child Care and Development Fund (CCDF), compute the required copayment for families who receive subsidies, and predict child care expenses for those without subsidies using regression equations. We impute other work-related expenses, primarily transportation costs, using the average weekly expense amounts provided by the Census Bureau.2714

As described in more detail below, we use poverty thresholds that include out-of-pocket health spending to take into account these expenses. The NAS recommended subtracting actual out-of-pocket expenses from income to compute poverty. Lacking recent estimates of medical spending for families with different characteristics in CT, we use the Census thresholds that include medical costs. Since the thresholds vary by health insurance coverage, we must assign health insurance coverage status to individuals. While the CPS reports public health insurance coverage, Medicaid and SCHIP are significantly underreported. TRIM3 simulates Medicaid and SCHIP eligibility and participation using the same types of procedures used for other program simulations so that caseloads come close to those reported in the administrative data.

Thresholds

The official measure of poverty uses thresholds based on a subsistence food budget times a factor of three. The measure was developed in 1963, based on spending patterns observed in a 1955 consumption survey (Blank & Greenberg, 2008). The thresh-olds represent nationwide spending averages and are adjusted by the change in the Consumer Price Index (CPI) each year. In contrast, the NAS thresholds are based on an average of the last three years of the Consumer Expenditures (CE) data.2815 We adjust these thresholds for the state of CT using factors from the Census Bureau that estimate living costs for urban and rural areas in each state based on differences in the fair market rent values across the country.

As noted earlier, we use the thresholds from the Census that incorporate out-of-pocket medical expenses. Experts differ in their recommendations for the treatment of out-of-pocket medical expenses in the poverty measure. The 1995 report recommended deducting expenses from income. However, surveys used to measure poverty on an annual basis do not provide estimates of out-of-pocket medical expenses, and the skewness of the spending distribution makes it difficult to accurately estimate these expenses from one data set and impute to another. As a result, many experts recommend including expected expenses in the thresholds rather than subtracting actual expenses from income (Iceland, 2005). The Census includes the impact of expected medical expenses in the thresholds using quarterly data from the 1996 Medicaid Expenditure Panel Survey (MEPS) that reflect differences by elderly or nonelderly, health insurance coverage, and health status (Short, 2001).

Table 9 shows the official and NAS poverty thresholds for a reference family of two adults and two children living in an urban area in CT. The official CY 2006 poverty threshold for a family of two adults and two children is $20,444. The NAS estimate that does not account for geographic differences or medical expenses is $21,818 (6.7 percent higher). Note that the two thresholds are not directly comparable, however, because they apply to two different measures of family resources. The threshold that accounts for geographic differences in living costs is $25,139 for families living in urban areas in CT, about 15 percent higher than the national NAS threshold. The inclusion of expected medical expenses increases the threshold to $27,579 for an urban family in good health with private insurance and $25,572 for an urban family with public coverage. Privately insured families are counted as poor at higher income levels than publicly insured families due to higher expected out-of-pocket medical expenses. The NAS thresholds used to estimate poverty also vary by family size.

Table 9. Official and NAS-based poverty thresholds: Family of two adults and two children (CY 2006).

		Geographic Adjustment for CT
	No Geographic Adjustment	CT–Urban	CT–Rural
a Source:^a U.S. Census Bureau: http://www.census.gov/hhes/www/poverty/threshld/thresh06.html. b ^b The official poverty thresholds do not include geographic adjustments. c ^c Alternative thresholds for a two-adult, two-child reference family are obtained from http://www.census.gov/hhes/www/povmeas/altmeas06/nas_experimentalthresholdsv2.xls and reflect thresholds calculated using the most recently available 12 quarters of Consumer Expenditure survey data. d ^d Following the Census Bureau's methodology, we adjust the threshold by insurance and health status using the “risk factors” in Table A-10 (Short, 2001).
Official Poverty Thresholda	20,444	NAb	NAb
Alternative NAS-Based Thresholdc
Exclude medical expenses from threshold	21,818	25,139	23,503
Medical expenses in threshold: Family hasd
Private insurance, good health	23,935	27,579	25,783
Private insurance, fair/poor health	24,402	28,116	26,286
Public Insurance, good health	22,194	25,572	23,907
Public Insurance, fair/Poor health	22,301	25,696	24,023
Uninsured, good health	23,971	27,620	25,822
Uninsured, fair/poor health	24,079	27,744	25,938

Baseline Poverty Estimates for CT

The NAS poverty definition produces a higher poverty rate for CT in 2006 than the official estimate (Table 10). The NAS poverty rate for all persons is 11.3 percent, compared with 8.5 percent for the official rate. The NAS rate for children is 10.9 percent, compared with the official 10.7 percent rate. The NAS poverty rate is obviously affected by the different threshold and resource definitions. As shown earlier, the NAS thresholds are higher for CT than the nation because of the relatively high cost of housing in CT. The effect of the higher threshold on the child poverty rate is offset by the expanded definition of resources that includes near-cash benefits. Since antipoverty programs often target families with children, they are more likely to receive benefits such as the refundable Earned Income Tax Credit (EITC) and health insurance coverage through the SCHIP program. The EITC increases family resources and higher rates of public health insurance coverage reduce medical expenses. Both effects reduce poverty using the NAS measure.

Table 10. Connecticut population by poverty status using the official and NAS definitions (2005/2006).

	Official		NAS
Poverty Definitiona	All Persons	Children	All Persons	Children
Source: Tabulations using TRIM3 and the 2006 and 2007 CPS/ASEC data. a ^a The standard poverty definition compares cash income to the official poverty thresholds. The NAS poverty definition counts the value of transfer benefits in income and subtracts tax liability and work-related expenses. The NAS poverty thresholds are based on the latest consumer expenditure data and are adjusted for geographic differences in cost of living. b ^b The poverty gap is the amount of money that would be needed to lift all families currently below poverty up to the poverty threshold. The poverty gap for all persons applies to families with children, families without children but with elderly members, and other families. The poverty gap for children applies to families with children.
% deep poor (<50% poverty)	3.4%	3.8%	3.0%	2.4%
% poor (<100% poverty)	8.5%	10.7%	11.3%	10.9%
% low income (<150% poverty)	15.0%	18.1%	26.5%	29.6%
Poverty gap (millions, 2006 $)b	$1,033	$351	$1,349	$372
Weighted sample (000s)	3,475	820	3,475	820

The poverty gap, the estimate of the total amount of money required to bring all families up to the poverty thresholds, also reflects these differences. The NAS definition indicates a higher gap of $316 million in CT for all families ($1,349 million compared with $1,033 million) and $21 million for families with children ($372 million compared with $351 million).

The differences between the two measures in the share of families living in deep poverty (below one-half the poverty threshold) and the share who are low income (below 150 percent of the poverty threshold) provide additional insight into the effects of the NAS measure on poverty. The share of children living in deep poverty drops using the NAS (from 3.8 percent to 2.4 percent), but the share of children in low income families increases dramatically (from 18.1 to 29.6 percent). Most safety net programs target families near the official poverty line, so that most families with incomes above the higher NAS poverty threshold do not receive assistance. The NAS measure indicates that a much larger share of families in CT live near the edge of poverty than the official poverty measure suggests.

ESTIMATING THE EFFECTS OF ALTERNATIVE POLICIES

We use the NAS poverty measure to estimate the effects of a variety of options to reduce child poverty for the Child Poverty Prevention Council (CPPC) in CT. Below we highlight the effects of expanding child care subsidies to cover all low-income families, investments in education that would increase the number of adults with associate of arts (AA) degrees, and fatherhood initiatives that would result in improved payment of child support. We estimate the effects of these initiatives by simulating their effects on income and poverty for each family in the CT sample and compare these results to the baseline. We also show the potential effects of expanding child care subsidies and AA degrees on employment and earnings.

Modeling Procedures

TRIM3 simulates the effects of different policies at the micro level by starting with the baseline NAS estimates and then calculating how different program rules affect families' benefits, taxes, and potentially labor supply to produce new poverty estimates. Capturing the effects of program rule changes is the most straightforward part of the simulations. The new program rules substitute for the baseline rules to estimate benefit eligibility, and new benefits are assigned to families. This scenario requires estimating the number of families that would respond to or “take up” the new benefits, and the TRIM3 program modules all include algorithms to predict participation based on the characteristics of eligible families.2916 The simulation procedures change receipt of the benefits targeted for expansion and then capture interactions across programs by recalculating all other benefits and taxes. For example, an expansion of child care subsidies will lower out-of-pocket child care expenses, and this will reduce SNAP benefits (SNAP disregards these expenses up to a cap) and possibly increase tax liability (by reducing a family's deduction for child care expenses).

Capturing the effects of alternative policies on labor supply is more challenging. Many poverty-reduction policies potentially will alter individuals' work and earnings. For example, policies that expand child care subsidies will increase net take-home pay for program beneficiaries and likely increase the number of single parents and secondary workers that look for work. Policies that increase adult education could increase wage rates and employment. The model uses estimates from the best available economics literature to estimate these effects. Typically, this literature provides estimates of the percent of adults likely to move into the labor market or increase earnings in response to a percentage increase in net income. The model uses these estimates to change labor supply and earnings of individuals likely to be affected by the policy intervention. The model subsequently recalculates all benefits and taxes based on these revised earnings and family income estimates.

We often provide high and low employment effects based on ranges that can be derived from recent empirical results showing the effects of similar options on work and earnings. The employment effects are illustrative, since available estimates typically reflect a different point in time and often a somewhat different intervention and target group. The estimates also assume that the labor market can absorb new workers and pay them the wages that currently accrue to workers with similar education and characteristics. Nonetheless, the illustrative employment effects show state policymakers the “potential” effects of policies likely to affect work incentives. We discuss the specific estimates used to simulate alternative employment scenarios below in the discussion of each policy option.

Increasing Child Care Subsidies

This option assumes that Child Care and Development Fund (CCDF) subsidies are an entitlement for eligible families. The option potentially will reduce poverty by reducing families' out-of-pocket child care expenses–one of the important work-related expenses deducted from income in the NAS poverty measure–and potentially by increasing families' incomes if some parents respond to this work incentive.

The option considered by the CPPC follows CT's current CCDF eligibility limits–families are eligible to begin receiving subsidies if their income is less than or equal to 50 percent of state median income (SMI); once enrolled, they remain eligible as long as income does not exceed 75 percent of SMI. Using the 2008 SMI levels, a family of four gains initial eligibility with income up to $46,908 and remains eligible with income up to $70,368. Family copayments range from 2 percent of income (for families with income up to 20 percent of SMI) to 10 percent (for families with income at 50 percent of SMI or higher). TRIM3 simulates this option by assuming that all families with income under 50 percent of SMI who are not currently receiving a subsidy would apply for and receive subsidies if they have out-of-pocket child care expenses, and that they would continue to receive those subsidies as long as their income remained under 75 percent of SMI.3017

We first simulate the direct effects of this option holding constant families' employment and earnings and then simulate the option assuming that parents' employment and earnings would increase. Schaefer, Kreader, and Collins (2006) review the wide range of estimates of the effect of child care subsidies on employment and report that estimates often vary by study group. Estimates range from an 11 percent increase in the probability of employment for low-income families (not on welfare) for each $1,000 annual increase in subsidies, down to about 4 percent for single parents and secondary earners. We used the lower end of the range (3 percent if unmarried and 8 percent if married), given the uncertainty of the estimates. The employment effect is simulated in TRIM3, so that the targeted effect applies to all single parents and secondary earners who are not working and have children under age 13.

The guarantee of CCDF subsidies to eligible families would reduce child poverty from 10.9 percent to 10.4 percent through the direct change in family net income, and to 9.5 percent assuming a positive effect on employment (Table 11). Increasing subsidized child care also would reduce deep poverty (from 2.4 to 1.6 percent among children), but have a limited effect on the share classified as low income (29.3 percent compared with 29.6 percent). The subsidized child care guarantee would have the largest effect on family resources at the lower end of the income scale. The poverty gap for families with children would drop from $372 million in the baseline to $294 million (about 20 percent) in the scenario with employment effects.

Table 11. Impact of guaranteed child care subsidies on NAS poverty in Connecticut (2005/2006).

	Guaranteed Child Care Subsidies
	Baseline		No Employment Effect		Employment Effect
NAS Poverty Definitiona	All Persons	Children	All Persons	Children	All Persons	Children
Source: Tabulations using TRIM3 and the 2006 and 2007 CPS/ASEC data. a ^a The NAS poverty definition counts the value of transfer benefits in income and subtracts tax liability and work-related expenses. The NAS poverty thresholds are based on the latest consumer expenditure data and are adjusted for geographic differences in cost of living. b ^b The poverty gap is the amount of money that would be needed to lift all families currently below poverty up to the poverty threshold. The figure for all persons applies to families with children, families without children but with elderly members, and other families. The poverty gap for children applies to families with children
% deep poor (<50% poverty)	3.0%	2.4%	2.9%	2.1%	2.7%	1.6%
% poor (<100% poverty)	11.3%	10.9%	11.1%	10.4%	10.7%	9.5%
% low income (<150% poverty)	26.5%	29.6%	26.5%	29.4%	26.4%	29.3%
Poverty gap (millions, 2006 $)b	$1,349	$372	$1,325	$349	$1,271	$294
Weighted sample (000s)	3,475	820	3,475	820	3,475	820

Education Improvements

The CPPC considered numerous options for improving education, including increasing high school retention, expanding access to college, increasing the number of dropouts completing their General Education Degree (GED), and workforce training initiatives for low-wage adults. We highlight the effects of increasing the number of adults with an associate of arts degree (AA) to demonstrate the potential effects of an adult education initiative on poverty.3118

Obviously, the poverty-reduction effects of an education initiative depend on how higher levels of education likely would affect the employment and earnings of those affected. The recent economics literature does not provide a consensus about the ability of increased post-secondary education to raise employment and earnings for disadvantaged youth and adults (Holzer, 2008). However, some approaches hold promise based on recent evaluations. Lerman (2007) summarizes the recent literature and shows that two-year attendance at a community college and completion of the AA could raise the earnings of male graduates by as much as 30 percent and female graduates by 47 percent.

This simulation assumes that one-half of the 600,000 non-disabled adults under age 50 in CT with a high school diploma but no higher degree obtain an AA degree. We illustrate a lower-effect scenario that increases wages for men and women completing the AA degree by 15 percent (with no new employment) and a higher-effect scenario that assumes a 40 percent increase in wages and a 15 percent increase in employment among nonworkers who completed the AA. Individuals who gain a job are assumed to find full-year employment for 35 hours per week at $18 per hour (the 2006 median hourly rate for individuals in CT with an AA degree).

The higher employment effect scenario would reduce the child poverty rate to 9.8 from 10.9 percent (Table 12). The share of children living in low income families would decline from 29.6 to 26.8 percent. The effect on the share of children living in deep poverty is fairly small, because most of these children live with an adult not likely to benefit from the education initiative due to their age, disability, or student status. Also, some of the adults who gain education have children and some do not, and some live in poverty at the baseline and some do not. The poverty rate for all persons drops by nearly 1 percentage point (10.5 percent compared with 11.3 percent), and the share in the low income category declines by almost 2 percentage points (24.8 compared with 26.5 percent). This scenario reduces the poverty gap for families with children from $372 million to $319 million. The new employment and higher wages not only reduce poverty but also reduce the need for government services; for example, the amount of annual SNAP benefits paid to CT families is estimated to fall by $9 million (4 percent), and TANF benefits fall by $6 million (6 percent). Note that the low employment effects scenario produces only small reductions in the poverty rate.

Table 12. Impact of increased attainment of AA degrees on NAS poverty in Connecticut (2005/2006).

	Increased AA Degrees
	Baseline		Lower Employment Effect		Higher Employment Effect
NAS Poverty Definitiona	All Persons	Children	All Persons	Children	All Persons	Children
Source: Tabulations using TRIM3 and the 2006 and 2007 CPS/ASEC data. a ^a The NAS poverty definition counts the value of transfer benefits in income and subtracts tax liability and work-related expenses. The NAS poverty thresholds are based on the latest consumer expenditure data and are adjusted for geographic differences in cost of living. b ^b The poverty gap is the amount of money that would be needed to lift all families currently below poverty up to the poverty threshold. The figure for all persons applies to families with children, families without children but with elderly members, and other families. The poverty gap for children applies to families with children.
% deep poor (<50% poverty)	3.0%	2.4%	3.0%	2.4%	2.7%	2.1%
% poor (<100% poverty)	11.3%	10.9%	11.1%	10.7%	10.5%	9.8%
% low income (<150% poverty)	26.5%	29.6%	25.9%	28.6%	24.8%	26.8%
Poverty gap (millions, 2006 $)b	$1,349	$372	$1,330	$360	$1,273	$319
Weighted sample (000s)	3,475	820	3,475	820	3,475	820

Increase Child Support Payments

The CPPC also considered fatherhood initiatives to increase employment rates for fathers with children living elsewhere and potentially increase child support payments to custodial mothers. Fatherhood initiatives could include investments in their education and training or better connections to employment prospects, thereby yielding higher earnings and enabling regular payment of child support obligations. Lacking experimental evidence on the effects of such initiatives, we demonstrate the potential effect of a fatherhood initiative by closing the entire gap between the amount of child support income due to low-income custodial families in CT and the amount actually received.3219 TRIM3 simulates the change in income for custodial families only and then resimulates all benefits and tax programs to capture the effects of higher incomes.3320 Consequently, custodial parent families that receive higher child support will receive somewhat less in SNAP (the child support disregard will dampen this effect) and potentially fewer families will be eligible for TANF.

The NAS child poverty rate falls from 10.9 to 10.5 percent as a result of higher child support payments (Table 13). In most cases, the amount of the child support award is not sufficient to raise the family above poverty, even when the award is paid in full. However, the full payment of all child support awards would reduce the poverty gap for families with children by 5 percent (from $372 million to $353 million). This option also reduces the share of children living in deep poverty.

Table 13. Impact of full payment of child support awards on NAS poverty in Connecticut (2005/2006).

	Baseline		Full Payment Effect
NAS Poverty Definitiona	All Persons	Children	All Persons	Children
Source: Tabulations using TRIM3 and the 2006 and 2007 CPS/ASEC data. a ^a The NAS poverty definition counts the value of transfer benefits in income and subtracts tax liability and work-related expenses. The NAS poverty thresholds are based on the latest consumer expenditure data and are adjusted for geographic differences in cost of living. b ^b The poverty gap is the amount of money that would be needed to lift all families currently below poverty up to the poverty threshold. The figure for all persons applies to families with children, families without children but with elderly members, and other families. The poverty gap for children applies to families with children.
% deep poor (<50% poverty)	3.0%	2.4%	2.9%	2.1%
% poor (<100% poverty)	11.3%	10.9%	11.2%	10.5%
% low income (<150% poverty)	26.5%	29.6%	26.3%	28.8%
Poverty gap (millions, 2006 $)b	$1,349	$372	$1,329	$353
Weighted sample (000s)	3,475	820	3,475	820

Examining Residual Poverty

This type of simulation exercise can help to inform states about the potential effects of different options for reducing poverty. It also can help states to understand better the characteristics of poor families with children. We simulated, for example, the combined effects of all of the policies that CPPC considered and showed why a significant share of children remained poor even after expanding adult education, child care subsidies, child support, and participation in safety net programs. About one in five of these children lived with an adult working full-time and full-year but at very low wages. The majority of the children who remained poor lived in families with an adult working either part-year or part-time (56 percent). The rest lived in families in which all the adults were elderly, disabled, and/or students (12 percent), or unemployed (10 percent).

SUMMARY

The NAS poverty measure provides an important benchmark for states to assess the potential effects of options that affect the net resources of families. Many options affect in-kind benefits or taxes, and the effects can only be shown by using a poverty definition based on net income. Even options that primarily affect earnings or child support income require a poverty measure that takes into account near-cash benefits since higher earnings can be offset by a loss in benefits such as SNAP or a gain (or loss) in the EITC.3421 The NAS poverty measure also allows states to assess the effects of their current policies on family well-being and allows them to monitor well-being over time.

The NAS poverty measure also takes into account differences in living costs across the country. The NAS measure indicates that significantly more families live in poverty in CT than the number indicated by the official poverty measure. The difference for children between the two poverty rates is fairly small, however, because many government assistance programs target low-income children. Some families with children receive subsidized child care, SCHIP, and WIC, and these benefits increase net income and family well-being, all else equal. Compared with the official measure, the NAS poverty measure also indicates a smaller share of children living in deep poverty in CT and a smaller poverty gap for families with children. On the other hand, the NAS measure indicates that a far larger share of families in CT are low income (below 150 percent of the NAS threshold) in CT. The NAS poverty threshold in CT is relatively high, and most government benefits target those living below the much lower official poverty threshold.

The options demonstrated in this article show that increasing subsidized child care, adult education, and child support would all reduce poverty among children. The effects will be much more robust if they produce the expected effects on parents' employment. Of course, assumptions about increases in labor supply require a strong labor market able to absorb more workers, many of whom may have limited skills and work experience. It is important to consider the sensitivity of the poverty effects using different assumptions about employment. Each of these options would require increased state resources focused on low-income families, and states need to understand the potential cost-benefit trade offs.

While an NAS-style poverty measure is a necessity for measuring the effects of policies on poverty, it is important to recognize the numerous imputations and assumptions required to implement the measure and to estimate the effects of policy on income. Census surveys do not include numerous critical elements for the NAS measure, including the values of many in-kind benefits, the receipt of other in-kind benefits (such as child care subsidies), and federal and state income taxes. In addition, individuals substantially underreport the receipt of current government benefits.3522 It is important to correct for underreporting; otherwise, the baseline resources of families will be significantly underestimated and poverty overestimated. The McDermott 2009 MAP Act recommends investments in data that would provide more direct measurement of some of the elements required, thereby reducing the number of imputations required to implement the NAS in the future.

The assumptions regarding the employment effects of various policies can no doubt be debated by economists since the literature is far from definitive. Simulations can provide high and low estimates to show the range of possibilities. Investments in careful evaluations of small-scale experiments that test out the effects of policy alternatives are also required to improve the behavioral content in models designed to estimate the effects of policies on poverty.

ACKNOWLEDGMENTS

The authors wish to thank Joyce Morton, the Lead Programmer on the TRIM3 Model at the Urban Institute, who provided invaluable assistance in completing the estimates of poverty for this study. Jam Tashi also provided excellent research assistance. The authors also recognize funding of the poverty measure development and simulations from the Connecticut Office of Planning and Management. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders.

• 1

  Levin-Epstein and Gorzelany (2008) describe the poverty-reduction discussions in 14 states and the District of Columbia.

• 2

  The Measuring American Poverty Act of 2009, H.R. 2909, introduced by Rep. Jim McDermott on June 17, 2009, recommends “largely following the recommendations of the NAS to improve and update the current poverty measurement.”

• 3

  This model has been developed and used at the Urban Institute for over 30 years under primary funding from the Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation (HHS/ASPE). The federal government uses the model to understand the coverage and impacts of government programs. Because TRIM3 requires users to input assumptions and interpretations about economic behavior and the rules governing federal programs, the conclusions presented here are attributable only to the authors of this article.

• 4

  This is a subset of proposals considered for the Child Poverty and Prevention Council in Connecticut. The full report, Giannarelli and Zedlewski (2009), can be found at http://www.ct.gov/opm/cwp/view.asp?A=2997&Q=383356#Report.

• 5

  See Blank and Greenberg (2008) for a discussion of the shortcomings in the official measure of poverty.

• 6

  See Citro and Michael (1995) for a complete explanation of the academy's recommendations.

• 7

  Iceland (2005) summarizes much of the research completed to evaluate the new measure of poverty as well as expert opinion on its various elements.

• 8

  The TRIM3 project Web site, http://trim3.urban.org, provides full documentation of the national, CPS-based model. Recently, both the Center on American Progress (CAP) and the Legislative Commission to End Poverty in Minnesota used TRIM3 to analyze recommendations to reduce poverty (CAP, 2007; LCEP, 2009).

• 9

  Committee members did not come to a single recommendation on every element of the measure. For example, many workshop participants favored incorporating the value of housing to homeowners (not included in the measure used here), but there was little consensus on the method that should be adopted. The Census Bureau provides some variations in approaches for implementing the NAS recommendations (Dalaker, 2005).

• 10

  The CT report referenced above includes more complete documentation of these procedures.

• 11

  As of October 1, 2008, Supplemental Nutrition Assistance Program (SNAP) is the new name for the federal Food Stamp Program. We use the new terminology in this report.

• 12

  We make additional adjustments to the baseline to account for policies in effect in CT in 2008, including a higher minimum wage, lower TANF benefit levels (in real terms), expansions in Medicaid/SCHIP eligibility for pregnant women and parents, and slight differences in dollar amounts used in income tax calculations (in real terms). These adjustments only make small differences in poverty rates, but they do provide a set of baseline estimates closer to those in effect in 2008.

• 13

  The statistical match used to impute capital gains (losses) and itemized deductions for federal income tax calculations is based on a national data match with the Statistics of Income (SOI) data. Results, however, were verified against tax reports for CT in 2006 and come close to those totals since other correlates of itemized deductions such as amount and types of taxable income are captured in the match.

• 14

  We obtained the figures used by the Census Bureau from Kathleen Short, personal communication, September 2008.

• 15

  The Consumer Expenditure (CE) Survey is a nationally representative survey that asks respondents to record a diary of many types of expenditures and interviews respondents about other expenses. The CE data are used to obtain national-level spending on food, clothing, shelter, and utilities for families whose spending is at approximately 80 percent of the median amount. Adjustments are made to allow for some spending on other items, and further adjustment is made for medical costs. See Appendix A of Short (2001) for details.

• 16

  As described in the model's documentation, each program typically includes probabilities of participation based on the characteristics of current enrollees shown in the administrative data. A random number is drawn and compared to the participation probability for each family to select participants.

• 17

  Working families with no child care expenses in the baseline simulation are presumably using free child care (for example, from an unpaid relative); we assumed these families would not want CCDF subsidies.

• 18

  The CT report includes the effects of each option separately and together (Giannarelli & Zedlewski, 2009).

• 19

  The amount of child support income received by a family is reported in the CPS data. Regression techniques are used to impute child support award amounts for custodial families.

• 20

  Note that these estimates do not include the potential effects on the families of the noncustodial parents, who are possibly earning more but also paying more in child support. TRIM3 does not have a method for identifying noncustodial parents' full child support awards.

• 21

  Each of these programs has its own phase-in and phase-out rules, and the marginal tax rate, that is, benefits lost as income increases, can be high. See Zedlewski et al. (2006) for a description of the variation in these phase out rules. The TRIM3 simulations capture these effects by recalculating benefit eligibility and receipt after estimating income changes.

• 22

  Of course, individuals also underreport private sources of income, especially income from self-employment and assets.