Bergstrom et al. (2024) study the Brazilian anti-poverty program Bolsa Familia, which provided cash transfers to poor households. Bolsa Familia was one of the world’s largest cash transfer programs with around 14 million households receiving benefits in 2014. Eligibility for the Bolsa Familia program was based on reported household per-capita income: between July 2009 and June 2014 households without children were eligible for a benefit of 70 Brazilian reais per month if their monthly per-capita income was below 70 Brazilian reais per month. In June 2014, both the eligibility threshold and the benefit were raised by 10% so that households without children were eligible for a benefit of 77 Brazilian reais per month if their monthly per-capita income was below 77 Brazilian reais per month.

The paper uses administrative data on all Bolsa Familia beneficiaries to study the impact of this June 2014 reform. We focus primarily on one-adult households without children as other groups face slightly more complex reforms. The paper applies a difference-in-difference strategy to estimate the causal impact of this reform on the number of eligible households. The paper calculates the number of households reporting incomes in various bins of the income distribution over time using portions of the income distribution which were not impacted by the reform as control groups for portions of the income distribution which were impacted by the reform (allowing for flexible group specific time trends).

MVPF = 0.9

The net cost of the Bolsa Familia program for single adult households is equal to the number of eligible households multiplied by the cost of benefit provision. The cost of the reform is therefore the difference in this net cost caused by the reform. There are three components: a mechanical effect, a “threshold effect”, and a “jumping effect”.

The mechanical effect is that all households reporting an income under 70 reais per month (1,902,490 households) get an increased benefit of 7 extra reais per month. This mechanical effect has a total cost of 13,317,430.

The “threshold effect” is that households reporting an income between 70 and 77 reais per month became eligible as a result of the notch change. This effect is also mechanical and comes at a total cost of 77*14,368=1,106,372.

Finally, there is a behavioral “jumping effect”, which the paper estimates using a difference-in-difference strategy. The jumping effect consists of households who change their reported income to jump below the eligibility threshold as a result of the reform: the paper estimates that there are 21,794 such households, leading to a fiscal externality of 77*21,794=1,678,170.

Summing these costs yields our estimate of the net cost: 16,101,971 Brazilian reais per month.

R$16.1M
Net Cost

16.7M Upper Margin
15.5M Lower Margin

For households who receive additional benefits and do not change their reported income in response to the reform, their WTP for the reform is just equal to the additional cash benefits they receive.

For households who do change their reported income in response to the reform, the paper argues that it is difficult to estimate their WTP exactly without making strong functional form assumptions on the utility function because we consider large (non-infinitesimal) reforms for which the envelope theorem cannot be applied.

The lower bound derived in the paper assumes that all behavioral effects are not valued at all (i.e., uses envelope theorem logic). This lower bound WTP equals 14,423,801 with a 95% confidence interval of [14,328,650, 14,518,952]. The upper bound for WTP derived in the paper is equal to 18,023,618.

R$14.4M
WTP

14.5M Upper Margin
14.3M Lower Margin

The MVPF is then 14,423,801/16,101,971 = 0.9 with a confidence interval of [0.86,0.93]. This calculation uses the lower bound estimate for WTP reported in the paper, which assumes that individuals do not value the increase in benefits due to behavioral changes (i.e., applying envelope theorem logic). The confidence interval reported herein is the statistical confidence interval for this lower bound MVPF estimate computed using the delta method.

The paper also provides an upper bound for the WTP and a corresponding upper bound for the MVPF which equals 1.12 with a 95% confidence interval of [1.08,1.16].

0.9
MVPF

0.9 Upper Margin
0.9 Lower Margin

Bergstrom, Katy, William Dodds, and Juan Rios (2014). “Welfare Analysis of Changing Notches: Evidence from Bolsa Família.” American Economic Journal (forthcoming). https://www.aeaweb.org/articles?id=10.1257/pol.20230024

- Category
- Taxes
- Sub-Category
- Cash Transfers
- Beneficiary Type(s)
- Adults
- Average Age
- 46
- Average Income
- 480
- Country of Implementation
- Brazil
- Year of Implementation
- 2014
- Empirical Method
- Difference in Differences
- Research Type
- Primary
- Peer Reviewed
- Yes
- MVPF Publication Link
- www.aeaweb.org/articles?id=10.1257/pol.20230024