Multiplier effects of social protection: a SVAR approach for Brazil

ABSTRACT

Based on a Structural VAR approach, we estimated fiscal multipliers for social benefits in Brazil for 1997–2018. Our results suggest that social benefits have relatively large multiplier effects, even when compared to public investment. The multipliers are also larger in the full sample, which includes the country’s 2014–16 economic crisis than in the period 1997–2014. In particular, our results show that spending one unit on social expenditures generates a final change in GDP of almost three after two years. The higher estimated multipliers in the full sample appear in the response of household consumption and private investment to shocks in total social expenditures and for different types of social benefits (e.g. cash transfers, unemployment insurance, and pensions). In a context in which the expansion of social protection became prominent as a response to structural changes in the labor market and the Covid-19 pandemic, our paper reinforces its potential role in the short-run economic recovery.

KEYWORDS:

• Fiscal multipliers
• social benefits
• business cycle
• social protection
• VAR

JEL CLASSIFICATION:

• E62
• H53
• I38

Acknowledgement

The authors thank the National Council for Scientific and Technological Development (CNPq) of Brazil for Marina’s masters scholarship and Laura’s research productivity fellowship at the University of São Paulo. The authors are also grateful for the helpful comments by Gilberto Tadeu Lima, Fernando Monteiro Rugitsky, Julia de Medeiros Braga and Esther Dweck.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. See, for instance, Ocampo and Arteaga (2016) and Ulu (2018).

2. See Tenhofen, Wolff, and Heppke-Falk (2010); Burriel et al. (2010).

3. Silva, Carvalho, and Ribeiro (2013); Furceri and Zdzienicka (2012); Reeves et al. (2013); Ilzetzki, Mendoza, and Vegh (2013); Izquierdo et al. (2019); Deleidi, Iafrate, and Levrero (2019).

4. See, for example, the “test of endogeneity” proposed by Deleidi, Iafrate, and Levrero (2019): “[…] we use quarterly data to test whether annual government investment is exogenous by evaluating whether the rate of growth of public investment responds to the rate of growth of GDP within the year” (Deleidi, Iafrate, and Levrero 2019, 14). Similar tests have been carried out in our study and confirmed that we can consider social benefits exogenous within the quarterly/monthly VAR framework.

5. The variables used in this paper are not stationary and hence we use their first difference as indicated by diagnostic tests (Dickey-Fuller, Phillips and Perron, KPSS). The number of lags is chosen based on information criteria and autocorrelation LM test (Deleidi et al. 2018) (most models use two lags). All models show stability. The tests for autocorrelation and heteroscedasticity pointed to the absence of these problems in most models.

6. Private investment data are obtained by subtracting public investment (from Orair and Gobetti 2017) from total investment (IBGE). Investment is available from 1997 to 2017.

7. For instance, Blanchard and Perotti (2002), Fatas and Mihov (2001), Tenhofen, Wolff, and Heppke-Falk (2010), Perotti (2007), Burriel et al. (2010), Izquierdo et al. (2019) adopt a one standard-deviation band. Ramey (2011) (p.11–12) claims: “Although this is a common practice in the government spending literature, it has no theoretical justification”.

8. The baseline specification used time dummy variables: 1) dumdate99: assumes value from 1999 Q1, onwards (government’s budgetary dynamics have changed); 2) dum08 assumes 1 in 2008Q3-Q4 and 2009Q1 (international crisis), 3) dum09, introduced to capture the post-crisis recovery period, assumes the value 1 in 2009Q3-Q4; 4) dum10 and dum67 (capture outliers in 2010 and 2006/2007). Besides showing at least 10% statistical significance, the inclusion of time dummy variables improved the general model significance. The removal of these variables only changed results marginally.

9. The cumulative multiplier is estimated considering its persistence: “The long-run multiplier is defined as the cumulative multiplier when J->∞, but in practice is used the number of periods needed for the multiplier to stabilize at its long-run value” (Garcia, Lemus, and Mrkaic 2013, 11).

10. See Carvalho (2018).

11. Some studies – such as Gáldon (2013), Adams and Wong (2018), Pereira and Wemans (2013), Baum and Koester (2011), Hollmayr and Kuckuck (2018) - estimate the SVAR considering unemployment insurance as an automatic stabilizer expenditure. They estimate the weighted elasticity of this type of expenditure relative to output. By making this adjustment, however, our results only change marginally.

12. We obtained the following estimations of the elasticity of primary revenue to GDP, household consumption, and private investment, respectively: 1.5; 1.4; 0.3 (full sample) and 1.3; 0.96 and 0.2 (pre-crisis sample).

13. In most models estimated in this work, this correlation is generally smaller than 0.1 (or close to), in modulus.

14. Models have also been estimated assuming [${\beta }_{tg}=0$ equation] and proved to be robust, as usual in the literature.

Additional information

Funding

Authors do not have any relevant financial or non-financial competing interests.