ATTACHMENT 8 Rules of Thumb Using 2012 QC Data.pdf

ATTACHMENT 8

INTENTIONALLY BLANK

MEMORANDUM INTENTIONALLY BLANK

TO: INTENTIONALLY BLANK

FROM: INTENTIONALLY BLANK DATE: 2/13/2014

SUBJECT: Rules of Thumb Update: Final Report

Contract No.: AG-3198-K-14-0007

Memo No.: 001

The Supplemental Nutrition Assistance Program (SNAP) provides millions of Americans with the means to purchase food for a nutritious diet and is the largest domestic food- and nutrition-assistance program administered by the U.S. Department of Agriculture’s Food and Nutrition

Service (FNS). According to FNS’ administrative records, SNAP distributed benefits to 47.6 million participants in an average month in fiscal year 2013, for a total of $76.1 billion in benefits over the entire fiscal year.1

FNS must routinely respond to questions from policymakers about the effects of proposed changes to SNAP policy on the program’s total amount of benefits paid to its participants. These questions often can be answered using microsimulation models that simulate the effect of a proposed policy change on each household in the model’s database. All eligibility and benefit determination rules modeled by the QC Minimodel and the MATH SIPP+ model can be adjusted to simulate the effects of the proposed change on SNAP caseload and benefit levels relative to the current (baselaw) program specifications. Given its flexibility and detail, microsimulation is generally the optimal method to estimate the effect of a proposed policy change.

In some instances, however, a policymaker needs to quickly obtain a general idea of the magnitude of a proposed change involving only one aspect of eligibility or level of SNAP benefits.

In these cases, it is not imperative to have the most precise estimate. Instead, the response needs to be quick and relatively close to the microsimulation model’s estimate. For this reason, we have created several “rules of thumb,” derived by (1) simulating a change in total SNAP benefits over a limited change in one SNAP eligibility or benefit parameter and (2) averaging the simulated changes in total SNAP benefits.

1 http://www.fns.usda.gov/pd/34SNAPmonthly.htm

An Affirmative Action/Equal Opportunity Employer http://www.fns.usda.gov/pd/34SNAPmonthly.htm

MEMO TO: INTENTIONALLY BLANK

FROM:

DATE:

PAGE:

INTENTIONALLY BLANK

1/24/2014

In this update of the Rules of Thumb report, we provide approximate rates of change in the cost of SNAP with respect to:

• The following eligibility or benefit parameters: minimum benefits, minimum benefits for elderly households only, minimum benefits for disabled households, maximum benefits, standard deductions, earnings deductions, maximum excess shelter deductions, and gross- and net-income screens

• Changes in sources of income that interact with the eligibility of SNAP

In general, an analyst uses a rule of thumb to answer the following question: “If a certain parameter were to change by so many dollars (or percentage points), what would be the approximate change in total SNAP benefits?” For instance, given our current rule of thumb for maximum benefit, we would expect that a $15 increase in the maximum benefit would result in a monthly increase of $174 million for the program.

We have based this update on simulations from the 2012 QC Minimodel and the 2011 MATH

SIPP+ model. In the next sections, we describe the methods used to calculate the rules of thumb and then provide examples of our calculation of rule of thumb estimates. In the last section, we discuss additional issues related to the rules of thumb for net- and gross-income screens.

A. METHODOLOGY

1. Changes to SNAP Parameters

As mentioned in the previous section, the SNAP rules of thumb are approximate cost estimates. They are derived from microsimulations that estimate the effect of a proposed policy change by evaluating the change in eligibility status and benefit level for each household within a micro-level data set. The QC Minimodel is based on a sample of SNAP households (Filion et al., 2013), and the MATH SIPP+ model is based on a sample of households drawn from the non-institutionalized U.S. population (Leftin et al., 2013). The microsimulation model determines the eligibility status and benefit amount of each household in the model’s database, according to the baselaw and reform SNAP eligibility rules in place in the simulated state of residence and given the households’s economic and demographic characteristics. Once the microsimulation model has simulated baselaw and reform outcomes for each household, it sums the household-level changes in eligibility, participation and benefits to produce an estimate of the change in the program’s outcomes (that is, the change in the number of SNAP participants, households, and benefits) due to the reform. This one simulation of a reform can then answer any number of specific questions.

For example, it can tell us how a reform affects the average benefit of households without children, or how many households, by poverty level, lose eligibility due to the reform.

An Affirmative Action/Equal Opportunity Employer

DATE:

PAGE:

INTENTIONALLY BLANK

When using a rule of thumb, the principal outcome of interest is the total change in SNAP benefits due to the reform. Since the rule of thumb must provide a general indication of the cost change due to a change in a SNAP parameter, we average the results of the microsimulation over a range of incremental changes to the same parameter (see Table 1). For instance, we derive the rule of thumb estimates for changes to the maximum benefit from the simulation results of changes to the maximum benefit ranging from $1 to $25. We divide the simulated changes to the total

SNAP costs by the change in the parameter values for the maximum benefit, and then we average these changes in SNAP benefits per dollar change in the parameter values. For example, we take the following steps to derive the rule of thumb for the maximum benefit:

1. Simulate 20 reforms that range incrementally from a $1 to $25 decrease in the maximum benefit.

2. Divide the change in total benefits for each reform by the change in the maximum benefit relative to the baselaw value.

3. Calculate the average rate of change in total benefits relative to the change in the maximum benefit. This average rate is the rule of thumb for the maximum benefit.

Because the QC data set contains the most complete sample of SNAP households, we can usually best approximate a reform using a rule of thumb estimated with restrictive reforms using the QC Minimodel. However, the QC Minimodel is usually not the most appropriate

Microsimulation model for expansive reforms because the QC data set only contains information on participating households, and the model is unable to simulate any increases in eligibility or participation.

The rules of thumb for parameters that control eligibility only, such as the gross- and net-income screens, may be substantially different when estimated over restrictive and expansive reforms. For example, a $10 decrease in the gross-income screen will result in a drop in the number of participants whose gross income was within $10 of the income screen, whereas a $10 increase will result in a smaller increase in the number of participants whose gross income was over the income screen by $10, because not all those who become newly eligible are expected to participate.

Also, on average, the benefits of the households losing eligibility will be higher than those who gain eligibility. For this reason, we should not expect that the income-screen rules of thumb will be similar when estimated under restrictive and expansive reforms. Given this difference, FNS has requested that the gross- and net-income screens also be estimated over expansive reforms using the 2011 MATH SIPP+ model (see Table 1).

The limited range of reforms used to calculate rules of thumb (see Table 1, fourth column) indicate the range of parameter changes for which a rule of thumb should be used. Reforms that involve larger changes from baselaw should be estimated with the microsimulation model. We expect small changes from baselaw to be fairly stable in terms of the changes in benefits, while large changes from baselaw may result in benefit changes that differ substantially from those within the limited range.

DATE:

PAGE:

INTENTIONALLY BLANK

2. Changes to Income Sources Interacting with SNAP

Because low-income households often participate in more than one Public-Assistance program, changes to programs that interact with SNAP can cause changes in the number of SNAP participants and benefits. For instance, because income from Temporary Assistance to Needy

Families (TANF) is part of a household’s gross income, changes in the amount of TANF benefits will affect the gross income of SNAP households—and therefore each household’s SNAP eligibility status and benefit level. For example, an increase in TANF benefits will cause an increase in SNAP household income, which, in turn, will lower the household’s SNAP benefit if eligibility is retained.

We estimate rules of thumb for eight types of income sources that interact with SNAP: TANF, Supplemental Security Income (SSI), Social Security, Unemployment Compensation, General

Assistance (GA), Child Support, earned income, and non-Public Assistance unearned income.2

TANF, SSI, and GA have a direct effect on the SNAP benefit because receipt of these program benefits increases countable gross income in SNAP households. They do not, however, affect the benefit level of other programs.

Social Security, Unemployment Insurance, Child Support, Earnings, and non-Public

Assistance Unearned Income have a direct effect on the SNAP benefit and an additional indirect effect when households also receive income from TANF, SSI or GA. This is because TANF, SSI, or GA benefit amounts change when other household income sources change.

We use a two-step process to estimate rules of thumb for these interacting income sources.

First, we estimate the direct influence on SNAP benefits of an additional dollar of program benefits from the interacting source. Next, we estimate the indirect effect of a decrease in TANF, SSI, and

GA benefits. For example, an additional dollar of TANF benefits will increase net income either one-for-one (if the household has no remaining deductions) or at a rate less than one (if other deductions apply). The decrease in SNAP benefits will equal the increase in net income multiplied by the SNAP benefit reduction rate parameter. We estimate this decrease by simulating an increase of $10 for every household in the QC-based model’s database and then calculating the difference in SNAP benefits in households that also receive TANF. On average, the decrease in SNAP benefits in households that receive TANF is approximately 26 cents (see Table 2). To estimate a

TANF rule of thumb for all SNAP households, we then apply to this

2 Non-Public Assistance unearned income includes Social Security, Unemployment Compensation, Child-

Support payments, Pension Income, Workers’ Compensation, Veterans’ Benefits, Alimony, Educational Loans, Financial Assistance, Rental Income, and income from Interest or Dividends. It excludes TANF, SSI and GA benefits.

DATE:

PAGE:

INTENTIONALLY BLANK

decrease the fraction of TANF income that is paid to SNAP households, 74.13 percent (see Table

3).3 The decrease weighted by the fraction represents the average change in SNAP benefits due to a dollar increase in TANF benefits: 19 cents.4

Next, we estimate the indirect effects caused by the interaction between Social Security, Unemployment Compensation, Child Support, Earned Income, or Unearned income and TANF, SSI, and GA benefits. Assuming that $1 from these interacting income sources will decrease the benefits from TANF, SSI, and GA by $1, we calculate the average indirect effect in a way similar to the TANF example. The average indirect effect equals the average increase in SNAP benefits in households in the transfer program receiving the interacting income, weighted by the fraction of interacting income paid to these households. For example, the rule of thumb for Social Security equals the sum of the average direct decrease in SNAP benefits from an increase in Social Security and the three average indirect increases in SNAP benefits resulting from TANF, SSI, and GA. In other words, it equals the sum of the direct decrease (0.3032 x 6.69%) and the indirect increase due to TANF (-0.3744 x 0.13%), SSI (-0.3162 x 1.51%), and GA (- 0.3310 x 0.06%). Summing the values gives us the rule of thumb for Social Security: $0.015.

3. Parameters and Interacting Income Sources

We developed rules of thumb for estimating the benefit impact of changes in the following

SNAP parameters and interacting income sources:

• Maximum Benefit. The maximum SNAP benefit amounts are proportional to the

Thrifty Food Plan values. The 2009 American Reinvestment and Recovery Act

(ARRA) increased the maximum benefit by 13.6 percent. For every reform, we applied the same percent change (the BENMULT parameter) to the set of baselaw maximum-benefit parameters. The dollar amount used to calculate a reform’s change in total

SNAP benefits relative to the change in maximum benefit equals the reform’s parameter percent change multiplied by the maximum benefit for a U.S. household of four.

• Minimum benefit (all households). The 2008 Food Conservation and Energy Act

(2008 Farm Bill) changed the minimum benefit to eight percent of the maximum

3 The fraction of TANF income that is paid to SNAP households is used to measure the level of interaction between SNAP and TANF. We do not use the fraction of interacting TANF recipients or households that also receive

SNAP because that would assume that the additional TANF dollar has the same effect on the SNAP benefit over all of the SNAP units that receive TANF.

4 In other words, the decrease in SNAP benefits arising from a $1 increase in gross income multiplied by the fraction of TANF income received by SNAP household, or (0.2557) x (0.7413) = 0.1896.

DATE:

PAGE:

INTENTIONALLY BLANK

benefit for a single-person SNAP household, and ARRA raised the minimum benefit for a one- or two-person household to $16 in the contiguous United States, $19 in

Alaska, and $25 in Hawaii in FY 2012. For each reform, we applied the same percent change to each of the minimum-benefit values in each geographic area. The dollar amount used to calculate a reform’s change in total SNAP benefits relative to the change in minimum benefit equals the reform’s parameter percent change multiplied by the minimum benefit of U.S. households of one or two persons.

- Minimum benefit for elderly households only. For this rule of thumb, all one-and two-person households that do not include an elderly member receive their

FY 2012 minimum benefit, and all one- and two-person households that contain an elderly member receive their reform minimum benefit.

- Minimum benefit for disabled households only. This rule of thumb is similar to the minimum benefit for elderly households only, except one- and two-person households that contain a disabled member receive their reform minimum benefit.

• Standard deduction. The 2008 Farm Bill raised the standard deduction for households of three or fewer people, and in the fiscal years following FY 2008, the standard deduction is indexed to inflation. The standard deduction for households with four or more people was adjusted for inflation under previous law. The reforms for the rules of thumb reduce the set of standard deductions by a specific percentage. For instance, for every reform, we applied the same percent reduction—from 1 to 40 percent— to the set of standard deductions. The dollar amount used to calculate a reform’s change in total SNAP benefits relative to the change in the standard deduction equals the reform’s parameter percent change times the standard deduction for four-person SNAP households.

• Earnings deduction. The 1985 Food Security Act set the current earnings deduction to 20 percent of earnings. These rules of thumb involve simple increases and decreases of the 20 percent rate.5

• Shelter deduction cap. The FY 2001 Appropriations Act included provisions to raise the cap to $340 in FY 2001 and then index the cap in FY 2002 and beyond. In FY 2012, the shelter deduction cap for the contiguous United States was $459. We developed rules of thumb by varying the value of the shelter-cap multiplier, which we then convert to dollars.

5 MFIP earnings deductions are unchanged for the Earnings Deduction Rule of Thumb.

DATE:

PAGE:

INTENTIONALLY BLANK

• Gross and net income screens. The baselaw parameter values for gross and net income screens have remained relatively unchanged over time except for cost-of- living adjustments. For each reform, we applied the same proportional change to each of the baselaw parameter values.

- Restrictive reforms. We estimated these rules of thumb over a restrictive range of reforms (a decrease of 1 to 20 percent of the baselaw values) using the 2012

QC Minimodel.

- Expansive reforms. We estimated these rules of thumb over an expansive range of reforms (an increase of 1 to 20 percent of the baselaw values) using the 2011

MATH SIPP+ model.

• TANF, SSI, GA, Social Security, Unemployment Compensation, Child Support, Earned Income, Unearned income. We estimated these rules of thumb using a modified 2012 QC Minimodel. The reform used to calculate the rules of thumb was a simulation of SNAP outcomes given that each household had received an extra $10 of gross income.

B. EXAMPLES OF USING RULES OF THUMB

Table 1 presents the average rules of thumb for SNAP benefit and eligibility parameters and for interacting income sources. The examples below explain how to use the rules of thumb to produce an approximate cost impact estimate.

1. Rules of Thumb for Benefit and Eligibility Parameters

a. Increase Maximum Benefit by $15

The rule of thumb for the maximum benefit indicates that a $1 increase in the maximum benefit represents a $0.2522 increase per participant (see Table 1). The rule of thumb cost estimate is calculated by multiplying the dollar change in the parameter by the corresponding rule of thumb and the number of participants in SNAP. So, $15 x $0.2522 per $1 change per participant x

46,021,511 SNAP participants = $174 million increase in monthly SNAP costs. Annual value estimates equal the monthly estimate multiplied by 12. Estimating SNAP costs for changes greater than $25 is not recommended.

b. Increase the Minimum Benefit by $10 for Elderly Households Only

To estimate the costs of a reform that raises the minimum benefit for elderly households only, use the rule of thumb labeled “minimum benefit—elderly only” (see Table 1). This rule of thumb indicates that a $1 increase in the minimum benefit of elderly households represents an increase of

$0.1551 per participant. So, similar to the calculation in the first example, $10 x $0.1551 per $1 change per participant x 46,021,511 SNAP participants = $71 million increase in monthly SNAP

DATE:

PAGE:

INTENTIONALLY BLANK

costs. Annual value estimates equal the monthly estimate multiplied by 12. Estimating SNAP costs for changes greater than $10 is not recommended.

c. Increase the Earnings Deduction by 10 Percent

The rule of thumb for the earnings deduction indicates that a one percentage-point increase in the earnings deduction represents an increase of $0.5179 in benefits per participant (see Table 1).

So, 10 x $0.5179 per 1 percent change per participant x 46,021,511 SNAP participants =

$238 million increase in monthly SNAP costs. Annual value estimates equal the monthly estimate multiplied by 12. Estimating SNAP costs for changes greater than 10 percent is not recommended.

2. Rule of Thumb for Interacting Income Sources—Increase Social Security Income by

$10

The rule of thumb for interacting income sources indicates that a $1 increase in Social Security reduces SNAP benefits by 1.4 cents. The rule of thumb cost estimate for a $1 increase in Social

Security income is calculated by multiplying the dollar change in the source of income by the corresponding rule of thumb and the number of SNAP participants. So, $10 x -$0.014 per

$1 change per participant x 46,021,511 SNAP participants = $6 million decrease in monthly SNAP costs. Annual value estimates equal the monthly estimate multiplied by 12.

C. DISCUSSION OF RULES OF THUMB AND RECOMMENDATIONS

As shown in the “income screens” panel of Table 1, the rules of thumb for the income screen differ by model and type of reform (restrictive or expansive). In general, the rules of thumb differences by type of reform can be explained by differences in characteristics and behavior of

SNAP households that are currently eligible versus those that become eligible after the expansive reform.

We recommend that FNS use the restrictive rules of thumb based on the 2012 QC Minimodel when estimating impacts of restrictive reforms and use the expansive rules of thumb based on the

2011 MATH SIPP+ model when estimating impacts of expansive reforms

Please let us know if you have any questions or concerns.

cc: INTENTIONALLY BLANK

DATE:

PAGE:

INTENTIONALLY BLANK

1/24/2014

Table 1. Average Rule-Of-Thumb Benefit Impact Estimates for Changes to SNAP

Average Rule of Thumb

(Monthly Dollar Change in Total SNAP Benefits Range of Parameter Per Baselaw Participant Changes for which

Parameter Unit of Parameter

Changea Per Unit of Parameter

Change ($)) Rule of Thumb is

Accurate

Benefit Computation

Maximum Benefit (U.S. Household of 4) ±$1 0.2522 ±$25 Minimum Benefit ±$1 0.0332 ±$10 Minimum Benefit - Elderly Only ±$1 0.1551 ±$10

Minimum Benefit - Disabled Only ±$1 0.0544 ±$10

Income Deductions

Standard Deduction ±$1 0.1143 ±$25 Earnings Deduction ±1 Percentage Point 0.5179 ±8 Percentage Points

Shelter Deduction Cap (US) ±1 Percentage Point 0.0144 ±8 Percentage Points

Income Screensb

Gross, 2012 QC, Restrictive ±1% of Baselaw 0.0207 ±20% of Baselaw Net, 2012 QC, Restrictive ±1% of Baselaw 0.0075 ±20% of Baselaw Gross, 2011 SIPP, Expansive ±1% of Baselaw 0.0303 ±20% of Baselaw

Net, 2011 SIPP, Expansive ±1% of Baselaw 0.0027 ±20% of Baselaw

Interacting Income Sources

TANF ±$1 0.1895 n.a.

SSI ±$1 0.1687 n.a.

GA ±$1 0.1879 n.a.

Social Security ±$1 0.0148 n.a.

Unemployment Compensation ±$1 0.0384 n.a.

Child Support ±$1 0.0389 n.a.

Earned Income ±$1 0.0032 n.a.

Non-PA Unearned Income ±$1 0.0092 n.a.

Source: Based on 2012 QC Minimodel estimates. Estimates reflect the per-person change in SNAP benefits (in 2012 dollars) from changing the parameter by one unit ($1 or 1 percentage point) in 2012.

a The changes to parameters that are expressed as percent changes to baselaw values, i.e. Earnings Deducation, Shelter Deduction Cap, and the Gross and Net Income Screens, should be expressed in the number of percentage points change. For instance, the value entered into the Rule of Thumb formula when estimating the impact of a 10% increase in the Gross income screen should be 10, not 0.10.

b The rows labeled "SIPP, Expansive" are the recommended rules of thumb for an increase in an income screen. The rows labeled "QC, Restrictive" are the recommended rules of thumb for a decrease in an income screen.

DATE:

PAGE:

INTENTIONALLY BLANK

1/24/2014

Table 2. Benefit Reduction Rates

Other Programs in Which SNAP Units Participate Benefit Reduction Ratea

All SNAP Households -0.2204

TANF -0.2557

SSI -0.3162

General Assistance -0.1823

Social Security -0.3032

Unemployment Compensation -0.2996

Child Support Income -0.3029

Earned Income -0.2891

Unearned Incomeb -0.2800

TANF and Social Security -0.3744

TANF and Unemployment Compensation -0.2912

TANF and General Assistance -0.2980

TANF and Child Support Income -0.3148

TANF and Earned Income -0.3120

TANF and Unearned Income -0.3223

SSI and Social Security -0.3162

SSI and Unemployment Compensation -0.3464

SSI and General Assistance -0.3604

SSI and Child Support Income -0.3905

SSI and Earned Income -0.3605

SSI and Unearned Incomeb -0.3298

GA and Social Security -0.3310

GA and Child Support -0.3578

GA and Earned Income -0.2720

GA and Unearned Incomeb -0.2893

Source: 2012 QC Minimodel simulation of increasing gross income by $10.

a The number of units in the 2012 QC Minimodel with both GA and Unemployment Compensation income is too small to calculate a reliable benefit reduction rate; therefore, it is not included here.

b Unearned Income includes Social Security, Unemployment Compensation, child support, pension income, Workers' Compensation, Veterans' benefits, alimony, education loans, financial assistance, rental income, and income from interest or dividends. It excludes TANF, SSI, and GA benefits.

Table 3. CPS-Based Measures of Overlap in Program Benefits

TANF SSI GA Social Security Unemployment

Compensation

Child Support Earned Income Unearned Income

Total Income $740,947,584 $3,916,933,900 $53,497,398 $53,423,807,939 $4,781,840,242 $2,695,062,543 $486,004,000,000 $133,004,621,118

Percentage of

Income Received by SNAP Units 74.13% 53.36% 103.06% 6.69% 14.61% 19.47% 1.34% 7.16%

Percentage of

Income Received by SNAP/TANF Units n.a. n.a. n.a. 0.13% 1.01% 2.90% 0.08% 0.77%

Percentage of

Income Received by SNAP/SSI Units n.a. n.a. n.a. 1.51% 0.69% 2.76% 0.10% 2.45%

Percentage of

Income Received by SNAP/GA Unitsa n.a. n.a. n.a. 0.06% n.a. 0.04% 0.01% 0.11%

Source: 2011 MATH SIPP+ Model n.a. = not applicable a The number of SNAP units with both GA and Unemployment Compensation income is too small to calculate a reliable benefit reduction rate; therefore, GA is not included as an offsetting income source in the Unemployment Compensation rule of thumb.