Economic mobility along the business cycle. The case of Peru

ABSTRACT

The performance of Latin American countries in reducing poverty and expanding the middle class has been remarkable. By taking a close look at the Peruvian experience, we examine how this aggregate behavior relates to business cycle conditions and whether different population groups share this behavior. We find that social mobility is cyclical; it decreases in recessions but increases with strong economic growth. The reduction in poverty in Peru appears to be the result of a sustained increase in the poverty exit rate together with a prolonged decrease in the poverty entry rate. These results hold among heterogeneous groups and are particularly marked for households regarded as initially disadvantaged.

KEYWORDS:

• Poverty dynamics
• pro-poor growth
• pseudo-panel data
• Latin America

JEL CLASSIFICATION:

• C23
• C53
• I32
• O54

Ackowledgement

We are indebted to an anonymous referee, Roger Asencios, Nelson Ramírez-Rondán and participants at the XXXIV Economists Meeting of the Central Reserve Bank of Peru for their useful comments. We also want to thank Bruno Escobar for his outstanding research assistance. The financial support of the Research Center of Universidad del Pacífico (CIUP) is gratefully acknowledged. We alone are responsible for the views expressed in this paper and for any remaining errors.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ In 1998 two important shocks struck the Peruvian economy; namely the Russian financial crisis and a strong El Niño weather phenomenon, that led the economy into a long recession. The recovery was interrupted by political instability in the years 2000 and 2001. Later, in the mid-2000s and after deepening important reforms to promote stability and growth, the country benefited from extraordinarily favorable external conditions (high terms of trade and low international interest rates) that were only briefly interrupted by the global financial crisis in 2008–2009. See Rossini and Santos (2015) for a comprehensive narrative of the Peruvian experience.

² Since 2007, a specialized Poverty Advisory Committee – made up of international organizations’ representatives, national government agencies, and academic and research community representatives – participates in the review of poverty indicators. In particular, the Committee verifies the quality of the survey and the measurement of poverty. In 2010 the World Bank recognized the ENAHO survey as the most innovative survey at the Second Regional Statistical Innovation Competition.

³ In 2011 there was an update in the procedures and variables used in the calculation of poverty. For instance, several items in the food basket were updated. Until then, they were based on information collected in the early to mid-90s. From that point on the information comes from the consumption patterns observed in the ENAHO 2010. Likewise, energy requirements were based on 1985 international standards. They were updated to the new requirements published in 2004. Additionally, the imputation procedure for missing values on goods acquired through self-consumption, self-production, payment in kind and donations, changed from the average to the median by using more precise item categorization. Furthermore, the new calculation of household expenditure excludes expenses on housing improvements, expenses on patron saint festivities and out-of-home expenses paid by others (invitations) but includes the value of usage of durable goods.

⁴ Cruces et al. (2015) also use data on the ownership of various assets available in ENAHO, which also contain some retrospective information. In a preliminary examination, we find the accuracy of the backcasts of these regressors to be much lower than what is obtained for the variables included in our regressions (e.g., household composition). Yet, we believe this is an interesting extension for future work.

⁵ These values tend to be slightly smaller than those reported in the literature: about 0.6 in Dang and Lanjouw (2013), 0.58 in Dang et al. (2014) and 0.57 in Cruces et al. (2015). The reason is that the error terms can be rationalized as $e_t=\alpha +v_t$, that captures the neglected time-invariant heterogeneity, and $v_t$ is an independent noise; then, given the higher fit of our equations, the contribution of $\alpha$ (which is the source of the correlation between $e_0$ and $e_1$) is likely to be less important in our results.

⁶ Other characteristics displayed similar dynamics as they are related with these attributes. For instance, the education of the head’s parents (either the father or the mother) correlate strongly with the education of the head him of herself, and thus their transition probabilities resemble those shown in Figure 4(e) and 4(f).

⁷ For instance, Peru signed a free trade agreements with Chile (2006), the United States (2006), Canada (2008), China (2011), Mexico (2011) and The European Union (2012); and it is currently in negotiations with India and Turkey.