Working for your Bread:

class: center, middle, inverse, title-slide

# Working for your Bread:
## The Labor Supply Effects of SNAP
### .hi-green[Marianne Bitler], UC Davis & NBER
### .hi-green[Jason Cook], University of Utah
### .hi-green[Jonathan Rothbaum], US Census Bureau

---

# Disclaimer

> The content is solely the responsibility of the authors and does not necessarily represent the official views of RIDGE, USDA, or the Census Bureau. Results have gone through Census disclosure review under the following release numbers: CBDRB-FY21-POP001-0045,  CBDRB-FY21-POP001-003, and CBDRB-FY21-POP001-0049.

---

# Motivation

- SNAP is a primary component of U.S. safety net

- In 2019, $60 billion for nutrition benefits to over 36 million people

- Program means-tested `$\Rightarrow$` economic theory predicts `$\downarrow$` labor supply ([Hoynes and Schanzenbach, 2015](#bib-Hoynes2015))

- **Qualitative evidence:** Awareness of earnings/benefits trade off

> I was offered a job for $2 more, and then, I had to account for the travel... [I]f I take this job with me spending basically as much money as I’m making, .red[**my SNAP benefits are going to be lowered as well**]. So it basically would’ve been me working backwards. [emphasis added] .red[(Caspi, De Marco, et al., 2020)]

---

# Literature

## Sparse literature relating SNAP to labor supply

### .large[**Early Literature:**]

- Structural Methods ([Moffitt, 2004](#bib-Robert2004))

### .large[**More Recent Literature:**] 
- County-level roll out of the Food Stamp program `$\rightarrow$` modest negative labor supply effects for single mothers ([Hoynes and Schanzenbach, 2012](#bib-Hoynes2012))
  
- SNAP expansions among immigrant populations reduce work hours ([East, 2018](#bib-East2016))
  
- Small literature on natives is mixed (.red[Scholz, et al., 2009; Moffitt, 2016; Ben-Shalom, et al., 2011; Farkhad and Meyerhoefer, 2018])

---

# Literature

## ABAWDs (.red[**A**]ble-.red[**B**]odied .red[**A**]dult .red[**W**]ithout .red[**D**]ependents)
- I.e., individuals 18-49 who are required to work to get SNAP more than 3 months of 36 except in downturns.

- Growing strand of literature assessing effects of work requirements for ABAWDs

- Mixed results (.red[Harris, 2019; Han, 2018a; Stacy, Scherpf, et al., 2016; Cuffey, Mykerezi, et al., 2015; Gray, Leive, et al., 2019; Stacy, Scherpf, et al., 2018])

## No work studying intensive margin effects of SNAP benefits formula

.red[**Why?**] Data requirements are substantial hurdle

---

# Data Quality

- Much of prior research uses self-reported surveys on earnings and SNAP

- Problematic because self-reported SNAP is under-reported in ways likely not innocuous .red[(Harris, 2019; Han, 2018a; Stacy, Scherpf, et al., 2016; Cuffey, Mykerezi, et al., 2015; Gray, Leive, et al., 2019; Stacy, Scherpf, et al., 2018)]

- Further, to study SNAP benefit formula, need to know monthly income and detailed expenses (e.g., shelter costs) which are rarely measured

---

# This Paper
 
First assessment of whether SNAP benefit formula distorts labor supply along intensive margin .red[**+**]

.footnote[ .green[**+**] .red[**Intensive margin**] - how *much* people work as opposed to *whether* they work at all.]

- Uses novel, .green[**detailed administrative program data**] for Colorado and Oregon

- .green[**Can observe net income and disregards**], e.g., dependent care and excess shelter cost

- Accurate SNAP eligibility and benefit levels

---

# Contribution

Fills gap in SNAP-labor supply and bunching literatures

## Bunching

- Growing public finance literature revitalized by ([Saez, 2010](#bib-Saez2010)) studying adjustments to kinks/notches in budget set

- Largely focused on taxes, and papers on means tested programs have focused on tax credits requiring work (see review by ([Kleven, 2016](#bib-Kleven2016))

- No evidence of this systematic behavior for safety net programs

---
# Policy Implications

- SNAP designed to reduce labor supply distortions with benefits that fade out as earned income `$\uparrow$`  .red[(Oliveira, Prell, et al., 2018)]
  
  - Whether there are distortions is an empirical question
  
- Important policy implications for SNAP and for work requirements in the safety net more broadly
  - E.g., Time limits

---
class: inverse, middle

# Understanding SNAP Benefit Formula

---

# Benefit Calculation
- Maximum benefits attached to Dept. of Agriculture Thrifty Food Plan
 
  - Amount to feed family at "minimal cost"

- Families expected to contribute 30 percent of available resources (i.e., .red[**net income**]) for food

**How is net income calculated?**

---

# Benefit Calculation
## **Net Income Deductions**
- **Standard** Deduction - .red[*misc costs*]
  - $167 for 2021, for fewer than 3 people in SNAP case

- **20% Earned Income** Deduction - .red[*work incentive*]

- **Child Care**, **Child Support**, and **Medical** Deductions  - .red[*rarely used*]

- **Shelter** Deduction

- Rent/utility costs exceeding half of net income
  
  - Capped at $586 for 2021, unless disabled or elderly case

---
# Share of Cases using each Disregard

- .red[**"Earner Sample":**] $1+ of earned income and not elderly or disabled case

- .red[**Shelter**] is most common deduction outside of earner and standard

---
# Average Levels of Net Income Inputs

- Earner sample has `$\uparrow$` earnings and `$\downarrow$` unearned inc
- Disregards have similar values, but shelter is most common

---
class:white-slide

**Mathematically...**

First .hi-green["Countable Income"] is calculated

`$$CountableInc = UnErnInc + 0.8ErnInc - StdDepChldMedDed$$`
--

then the .hi-green[Shelter Deduction] is calculated

`$$ShelterDed = min\left[ShelterCap, Shelter Expenses - \frac{CountableInc}{2}\right]$$`
 
--

Now, .hi-green[Net Income] can be calculated

`$$NI = CountableInc - ShelterDed$$`

Finally, we can get .hi-green[Benefits]

`$$Benefits = max\left[MaxBenefit - 0.3*NI, MinBenefit\right]$$`

**Note:** Because earned income enters into shelter deduction, .green[**benefit reduction rate** is **steeper** for cases with **shelter deductions**]

---
class:white-slide

Let's visualize this for a family of 3 with standard deduction and either
- **No Shelter Expenses** or **$600 of Shelter Expenses**

.pull-left[
.center[.hi-green[Net Income]]
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-5-1.png)
.center[**Earned Income**]
]
.pull-right[
.center[.hi-green[Benefit Amount]]
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-6-1.png)
.center[**Earned Income**]
]

---
# Maximize Benefits; Net Income `$=0$`
## Take Away

- With shelter deduction, can have sizeable earnings with full benefits
  
  - Potential scope for bunching behavior

- SNAP benefits taxed away **24 cents** on every dollar of *earned* income for households with **no shelter deduction**

- SNAP benefits taxed away **36 cents** on every dollar of *earned* income for households with **some excess shelter expenses**

- Thus, recipients may .red[**bunch at Net Income = 0**], i.e., where each additional dollar earned is initially taxed

- Possible .red[**larger effects for households with shelter deduction**]

---
# Empirical Design

- Use well-known bunching estimators utilized in other contexts .hi[*]
- Augment .red[Chetty, Friedman, et al. (2011)] and estimate
`\begin{equation*}
 C_j=\sum_{i=0}^{7}\beta_i\cdot\left(Z_j\right)^i+\sum_{i=-200}^{200}\gamma_i\cdot1\left[Z_j=i\right]+\epsilon_j \label{eq:basicbunch}
\end{equation*}`

- `$C_j$`: number of respondents in net income bin `$j$`
	- `$Z_j$`: net income bin for unit ($50 intervals)
	- Excluded region around net income = 0 is $400

- .green[**Intuition:**] approximate shape of counterfactual bunching region by interpolating the shape from surrounding areas with order 7 polynomial

- Parametric bootstrap for standard errors

.footnote[.hi[*] .red[Kleven (2016); Saez (2010); Bertanha, McCallum, et al. (2019)]]

---
# Example

---
# Empirical Design
## Caveat

- Specification selected for its simplicity, but has drawbacks

- Bunching likely `$\Rightarrow$` non-zero taxable income elasticity at kink, but without restrictions on the heterogeneity distribution any sized elasticity is consistent with kinks (but not notches) .red[(Blomquist, Newey, et al., 2019; Bertanha, McCallum, et al., 2019)] 
- In future work, leverage benefit changes (e.g., new Thrifty Food Plan adjustments every October, or categorical eligibility) to better model counterfactual density
---
class: inverse, middle

# Results

---
# Aggregate Gross/Net Income

.pull-left[
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-9-1.png)
]
.pull-right[
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-10-1.png)
]

- .red[**Oregon:**] Large mass at <$150 Gross (683,000 cases)
 - Not centered at NI=0 so not interpreted as caused by kink
 - Not seen in categorically eligible cases via TANF .red[(not pictured)]

---
# Aggregate Gross/Net Income

.pull-left[
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-12-1.png)
]
.pull-right[
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-13-1.png)
]

- .red[**Colorado:**] Relatively smooth gross income

- .red[**Both:**] No noticeable bunching at NI=0

---
# Net Income by Size of Case

- Possible that labor market frictions differ by .hi-green[family size]

- Especially true for .hi-green[single-person units]

`$\Rightarrow$` We might expect heterogeneity in bunching by SNAP case size

---
# Net Income by Size of Case

.pull-left[
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-15-1.png)
- $150 mass are single-unit cases, but not centered at NI=0
]
--

.pull-right[
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-16-1.png)

- Possible bunching for single-unit cases (n.s.)
]

Let's explore the .hi-green[single-unit bunchers in CO], by .hi-green[self employment income]

---
# CO Net Income by Self Employment

- Self employed exhibit most bunching behavior in other contexts (.red[Saez, 2010; Chetty, Friedman, et al., 2011])

- Colorado provides data on self employment income

.pull-left[
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-18-1.png)
]
.pull-right[
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-19-1.png)
]
---
# Colorado Net Income for Case Size = 1
.pull-left[
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-20-1.png)
]
.pull-right[
- .hi-green[Statistically significant bunching] for self-employed single-unit cases

- Not visible in single-unit cases without self-employment income

- .hi-green[Economically small]: 0.3 percent of cases in Colorado earner sample
]

Next, we explore possible bunching heterogeneity by .red[**shelter deduction**]

---
# Net Income by Shelter Deduction

.pull-left[
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-22-1.png)
]
.pull-right[
![](AEAPP_Bitler_Cook_Rothbaum_files/figure-html/unnamed-chunk-23-1.png)
]
 
 No substantial heterogeneity (.red[aside from Oregon low earners])

---
#Conclusion

## Summary

- Explored bunching where SNAP benefits are initially taxed

- Discussed why there could be possible heterogeneity: .hi-pink[case size], .hi-green[self employed], .brown[**shelter deduction**]

- Only evidence of bunching for self-employed single-unit cases, but **economically small effects**
  
## .red[**Main Take Away**] 
Labor distortions from SNAP benefit kink .red[not a first-order concern]
  
---
# Next Steps

- Leverage people's experience with SNAP. Do people with experience do more bunching?

- Explore impact of work requirements (e.g., ABAWD time limits)

- Explore impact of notches from net/gross income tests, particularly for shelter deduction cases

---
class: white-slide, center, middle

.huge[**Thank you**]
---
count: false

# References

<a name=bib-Ben-Shalom2011></a>[Ben-Shalom, Y, R. Moffitt, et
al.](#cite-Ben-Shalom2011) (2011). "An assessment of the
effectiveness of anti-poverty programs in the United States". In:
_Vasa_. URL:
[http://medcontent.metapress.com/index/A65RM03P4874243N.pdf%5Cnhttp://www.nber.org/papers/w17042](http://medcontent.metapress.com/index/A65RM03P4874243N.pdf%5Cnhttp://www.nber.org/papers/w17042).

<a name=bib-Bertanha2019></a>[Bertanha, M, A. H. McCallum, et
al.](#cite-Bertanha2019) (2019). "Better Bunching, Nicer Notching".
In: _SSRN Electronic Journal_. DOI:
[10.2139/ssrn.3144539](https://doi.org/10.2139%2Fssrn.3144539).

<a name=bib-Caspi2020></a>[Caspi, C., M. De Marco, et
al.](#cite-Caspi2020) (2020). "SNAP and work-related policies: An
in-depth analysis of work perceptions and behaviors". In: _RIDGE
Innovation Grantee Conference_.

<a name=bib-Chetty2011></a>[Chetty, R, J. N. Friedman, et
al.](#cite-Chetty2011) (2011). "Adjustment Costs, Firm Responses, and
Micro vs. Macro Labor Supply Elasticities: Evidence from Danish Tax
Records.". In: _The Quarterly Journal of Economics_ 126.2, pp.
749-804.

---
count: false

# References

<a name=bib-Cuffey2015></a>[Cuffey, J., E. Mykerezi, et
al.](#cite-Cuffey2015) (2015). "Food Assistance and Labor Force
Outcomes of Childless Adults: Evidence from the CPS".

<a name=bib-East2016></a>[East, C. N.](#cite-East2016) (2018).
"Immigrants' labor supply response to Food Stamp access". In: _Labour
Economics_ 51.April, pp. 202-226.

<a name=bib-Farkhad2018></a>[Farkhad, B. F. and C. D.
Meyerhoefer](#cite-Farkhad2018) (2018). "The Impact of Participation
in SNAP on Labor Force Decisions".

<a name=bib-Gray2019></a>[Gray, C, A. Leive, et al.](#cite-Gray2019)
(2019). "Employed in a SNAP? The Impact of Work Requirements on
Program Participation and Labor Supply".

<a name=bib-Haider2008></a>[Haider, S. J. and D. S.
Loughran](#cite-Haider2008) (2008). "The effect of the social
security earnings test on male labor supply: New evidence from survey
and administrative data". In: _Journal of Human Resources_. ISSN:
0022166X. DOI:
[10.3368/jhr.43.1.57](https://doi.org/10.3368%2Fjhr.43.1.57).
---
count: false

# References

<a name=bib-Han2018a></a>[Han, J.](#cite-Han2018a) (2018a). "The
Impact of Snap Work Requirements on Labor Supply". In: _SSRN
Electronic Journal_. DOI:
[10.2139/ssrn.3296402](https://doi.org/10.2139%2Fssrn.3296402).

<a name=bib-Harris2019></a>[Harris, T. F.](#cite-Harris2019) (2019).
"Do SNAP Work Requirements Work?". In: _SSRN Electronic Journal_.
DOI: [10.2139/ssrn.3379741](https://doi.org/10.2139%2Fssrn.3379741).

<a name=bib-Hoynes2015></a>[Hoynes, H. W. and D. W.
Schanzenbach](#cite-Hoynes2015) (2015). "U.S. Food and Nutrition
Programs".

<a name=bib-Hoynes2012></a>[Hoynes, H. W. and D. W.
Schanzenbach](#cite-Hoynes2012) (2012). "Work incentives and the Food
Stamp Program". In: _Journal of Public Economics_ 96.1-2, pp.
151-162.

<a name=bib-Kleven2016></a>[Kleven, H. J.](#cite-Kleven2016) (2016).
"Bunching". In: _Annual Review of Economics_ 8.June, pp. 435-464.

---
count: false

# References

<a name=bib-Meyer2019></a>[Meyer, B. D. and N.
Mittag](#cite-Meyer2019) (2019b). "Using Linked Survey and
Administrative Data to Better Measure Income: Implications for
Poverty, Program Effectiveness and Holes in the Safety Net". In:
_American Economic Journal: Applied Economics_ 11.2, pp. 176-204.

<a name=bib-Meyer2018a></a>[Meyer, B. D, N. Mittag, et
al.](#cite-Meyer2018a) (2018). "Errors in Survey Reporting and
Imputation and Their Effects on Estimates of Food Stamp Program
Participation". In: _NBER Working Paper \#25143_.

<a name=bib-Meyer2015a></a>[Meyer, B. D, W. K. C. Mok, et
al.](#cite-Meyer2015a) (2015b). "Household Surveys in Crisis". In:
_Journal of Economic Perspectives_ 29.4, pp. 199-226.

<a name=bib-Meyer2007></a>[Meyer, B. D. and J. X.
Sullivan](#cite-Meyer2007) (2007). "Under-reporting, Take-up, and the
Distributional Effects of the Food Stamp Program".

---
count: false

# References

<a name=bib-Robert2004></a>[Moffitt, R.](#cite-Robert2004) (2004).
"Welfare Programs and Labor Supply". In: _Handbook of Public
Economics_. ISSN: 1098-6596. eprint: arXiv:1011.1669v3.

<a name=bib-Moffitt2016></a>[Moffitt, R. A.](#cite-Moffitt2016)
(2016). "The U.S. Safety Net and Work Incentives: Is There a Problem?
What should be Done?". In: _The US Labor Market: Questions and
Challenges for Public Policy_. . Chap. V, pp. 122-138.

<a name=bib-Oliveira2018></a>[Oliveira, V., M. Prell, et
al.](#cite-Oliveira2018) (2018). _Design Issues in USDA's
Supplemental Nutrition Assistance Program: Looking Ahead by Looking
Back._. 243. Economic Research Service, pp. 1-80.

<a name=bib-Saez2010></a>[Saez, E.](#cite-Saez2010) (2010). "Do
Taxpayers Bunch at Kink Points?". In: _American Economic Journal:
Economic Policy_ 2.3, pp. 180-212.
---
count: false

# References

<a name=bib-Scholz2009></a>[Scholz, J. K, R. A. Moffitt, et
al.](#cite-Scholz2009) (2009). "Trends in Income Support". In:
_Changing poverty, changing policies_. New York: Russell Sage
Foundation, pp. 203-241.

<a name=bib-Stacy2016></a>[Stacy, B., E. Scherpf, et
al.](#cite-Stacy2016) (2016). "New Evidence on Labor Supply and SNAP:
What Are the Roles of Work Requirements, Expanded Eligibility, and
New Program Rules?".

<a name=bib-Stacy2018></a>[Stacy, B, E. Scherpf, et
al.](#cite-Stacy2018) (2018). "The Impact of SNAP Work Requirements".