The “Miracle” Tumbled with a Pandemic: Poverty and COVID-19 in Peru

From 2004 to 2019, Peru experienced some of the highest growth rates in its history; over the same period of time, its poverty rate dropped from 58.7% to 20.2%. This phenomenon was called the “Peruvian miracle,” yet Peru has been one of the COVID-19 pandemic’s greatest victims. By August 2019, the COVID transparency portal in Peru showed that the country had the highest number of deaths per million people in the world. Moreover, forecasts expected it to experience the largest GDP reduction in the LAC region and an increase of 10 percentage points in its poverty rate (Lavado, 2020). In this article, I will focus on the Peruvian social policy framework, the role it played during the pandemic, and some proposals for future improvement.

The Peruvian government reacted early (with only six cases detected) and applied most policies recommended by international organizations: monetary transfers, border closures, lockdowns, and the closure of commercial activities that involve gatherings. However, prior to COVID-19, it was already known that the precariousness of the health system and widespread labor informality (close to 70%) would constitute vulnerable flanks in the battle against the virus. In the trade-off between saving lives and saving the economy, Peru lost on both fronts.

The pandemic showed the many shortcomings of Peru’s social policy system. In the face of requests from support from citizens due to lockdowns, bureaucratic bottlenecks created difficulties for the implementation of emergency policies. Regarding this, I would like to elaborate on three main elements: (i) changes in poverty patterns, (ii) the unattended vulnerable population, and (iii) social program targeting.

First, poverty patterns at the time the virus arrived were considerably different from the prevailing patterns at the time the national social policies were designed. Data from the National Household Survey (Enaho) show that poverty in urban areas imposes new patterns of hardship. Although rural poverty rates continue to be much higher than urban ones, the absolute number of urban poor households has been larger than the number of rural poor households since 2013. This is because poverty reduction has occurred at different rates in different areas and regions, which makes it crucial to implement specially tailored social policies that address the specific determinants of urban poverty: precarious jobs, income instability, high bankruptcy risk of small family businesses, household overcrowding, and homelessness. This scenario is particularly challenging for the country, given that the COVID-19 pandemic has had a major impact on urban populations.

A second element to consider is the issue of vulnerability. In a previous study (Herrera & Cozzubo, 2016) we found that (i) the proportion of vulnerable households, understood as those with a high probability of falling into poverty, had been growing consistently during the “Peruvian miracle” and (ii) that in 2019 about a third of the Peruvian population had a high probability of becoming poor. In this work, we proposed not only the methodology for identifying vulnerable households but the need to design tailored policies that address concerns such as food safety, insurance mechanisms, job opportunities, and so on. We also found that these households were very susceptible to falling into poverty in the face of adverse shocks of great magnitude or when they occurred cumulatively. Many Peruvian households have experienced such shocks during the pandemic: they face, amongst other things, the death of income earners, out-of-pocket health expenses, unemployment, and small business bankruptcy. This scenario has led us to project that nearly 1 million households could fall into poverty (an increase of approximately 10 percentage points in the poverty rate) during the pandemic.

Finally, it is important to highlight the limitations of the National Targeting System (SINAFO) and the algorithm used for conditional cash transfers (JUNTOS) or non-contributory pensions (Pension 65) in fighting the pandemic’s effects. In 2017, the El Niño phenomenon affected 2 million people and destroyed an estimated $3 billion of property (EM-DAT). During this disaster, the lack of protocols for social policy in emergency situations, such as natural disasters and other large shocks (such as health-related ones), became clear. When the COVID-19 pandemic struck the country, the targeting algorithm had not yet been adapted to include the urban poor, did not prevent vulnerable households from falling into poverty, and did not include mechanisms for massive emergency cash transfers.

In addition to health measures, the government approved a series of targeted and universal cash transfers to combat poverty. These vouchers were an unprecedented effort as they served a population almost five times larger than that of the JUNTOS program, which had thus far been the largest social program in Peru. While their intentions were good, the implementation of the bonds was highly criticized and met with significant logistical obstacles, such as access to the web portal and agglomeration of beneficiaries in banking agencies where they cashed the bonds. Moreover, there were non-trivial problems for vulnerable households who were not selected to receive the emergency programs. For example, administrative records used for targeting were up to five years old and designed for policies with a rural focus; therefore, many poor urban households did not receive the cash transfers.

What can we learn from the pandemic response for current and future social policy in Peru? How can we include in the social policy framework these 3 million mostly urban households that will fall into deprivation? I humbly outline some ideas for discussion:

1. Prevention. Define protocols for social policy in emergency situations, differentiating between types of shocks (ex. Natural disasters versus health crises)

2. Adaptation. In complementarity with the fight against rural poverty, design and implement strategies for the different dimensions of urban poverty.

3. Inclusion. Consider vulnerable populations, not only populations already living in poverty, when designing social policy. Preventing the fall of households into poverty not only prevents households from losing opportunities and skills due to episodes of deprivation, but can also be a cost-effective strategy

4. Updated and useful information. Improve the quality and updating of administrative records. Ensure that it is interconnected with census and household survey databases. Use geolocation tools and poverty and income maps.

5. Knowledge. Improve the targeting algorithm that determines which households should receive benefits, considering the research from academia and the implementation experiences in other countries. It is now possible to exploit benefits from big data and machine learning techniques. For example, Blumenstock et al. (2016) and Jean et al. (2016) show how satellite images and cellphones usage data can be used for generating detailed poverty maps, while work from McBride and Nichols (2018) and Kshirsagar et al. (2017) are examples of how the targeting of social programs can benefit from applying machine learning techniques.

6. Evidence. In addition to the traditional difficulties of targeting, it is now crucial to act immediately. We must not stop making public policy decisions based on evidence.

---

Blumenstock, J. E. (2016). Fighting poverty with data. Science, 353 (6301), 753-754.

Herrera, J., & Cozzubo, A. (2016). La vulnerabilidad de los hogares a la pobreza en el Perú, 2004–2014 (No. 2016-429). Departamento de Economía-Pontificia Universidad Católica del Perú.

INEI (2020). Informe técnico Perú: Estimación de la Vulnerabilidad Económica a la Pobreza Monetaria. Metodología de cálculo y perfi­l sociodemográ­fico. Instituto Nacional de Estadística e Informática.

Jean, N., Burke, M., Xie, M., Davis, W. M., Lobell, D. B., & Ermon, S. (2016). Combining satellite imagery and machine learning to predict poverty. Science, 353(6301), 790-794.

Kshirsagar, V., Wieczorek, J., Ramanathan, S., & Wells, R. (2017). Household poverty classification in data-scarce environments: a machine learning approach. arXiv preprint arXiv:1711.06813.

Lavado, P.  (2020) Estimación del impacto de la COVID-19 en la pobreza, empleo y desigualdad. Presentación para PNUD-Perú. Junio, 2020.

McBride, L., & Nichols, A. (2018). Retooling poverty targeting using out-of-sample validation and machine learning. The World Bank Economic Review, 32(3), 531-550.