Cash grants and firm recovery - An RCT following cyclone Idai in Mozambique

In recent years, extreme weather events and natural disasters are increasing in quantity, severity and frequency. This trend is predicted to intensify even further in the near future due to anthropogenic climate change (IPCC, 2014). One of the countries that is most vulnerable to natural disasters because of its long coastline and socioeconomic fragility is Mozambique (Global Climate Risk Index, 2019; USAID, 2018). In March 2019, cyclone Idai hit Mozambique’s second largest port city Beira and surrounding areas. Around two million people were estimated to be affected by what is reported to be the most devastating disaster in the history of Southern Africa. The catastrophe resulted in a death toll of more than 1,000; several thousands of displaced people, washed away main roads that connect Beira and numerous villages with the rest of the country, severe food and water shortages, disease outbreaks and criminal robbery. Similar events are likely to occur on a more regular basis in the upcoming years and, therefore, it is crucial to create scientific evidence on how Mozambique can prepare for and recover from natural disasters.

While it is relatively well known how households in the Global South are affected by, cope with and recover from natural disasters, the situation of businesses after natural disasters is under-researched. There exists little evidence on firms and disasters in high-income countries, and studies on businesses in the Global South are even scarcer. However, micro, small and medium enterprises (MSMEs) employ around 90 percent of all workers in developing countries and are often the only source of income for the poor (Page and Söderbom, 2015). Thus, only if MSMEs get back up on their feet after a disaster, the local community will be able to recover fully (Mendoza et al., 2018).

In this project, we examine the process of how micro firms recover from cyclone Idai, and what are potential constraints to their recovery. By implementing an experiment (randomized controlled trial), we want to understand whether cash grants enhance firms’ recovery. This is similar to a study by de Mel et al (2012) who studied the impact of cash grants on firm recovery after the 2004 Asian tsunami in Sri Lanka. The principal outcome is that firms that obtained grants after a tsunami recovered their profit levels around two years earlier before enterprises that did not receive support. This effect, however, is stronger for the retail than for the manufacturing sector.

In September 2019, we initiated our sampling approach and baseline data collection in the cities of Beira and Chimoio. After completion of the baseline data collection, we randomly selected 130 firms into our treatment group and handed out cash grants of 100 USD in October 2019 (90 percent power). In April 2020, we conducted a follow-up survey. In total, firms were asked to recall five different points in time, i.e. months. In the baseline data collection, enterprises provided information about their pre-cyclone situation in February 2019 (1), post-cyclone condition in April 2019 (2) and baseline state in August 2019 (3). In the follow-up survey, we inquired about their post-treatment situation in February 2020 (4) and one year after the cyclone in March 2020 (5). Hence, we have two survey waves but five different points in time for which the enterprises provided data. The follow-up survey could not take place through personal interviews due to the Covid-19 pandemic and were done by phone.

Firm recovery does not only imply that firms manage to get back to their pre-cyclone income and profit levels but that they reach the level at which they would have been had the cyclone not occurred. In the ideal world, we would be able to observe the same treated firm's outcome affected by the cyclone and in absence of the cyclone. The difference between the affected and unaffected outcome would be the causal impact of the disaster. Unfortunately, it is impossible to know what would have happened to the same enterprise in the absence of Idai. Therefore, we need a reasonable counterfactual enterprise that was not hit by the cyclone but followed the same economic trends as the affected firm. We use the city of Chimoio as a counterfactual location. Chimoio is about 200 km away from Beira and was much less affected by Idai. As Chimoio was affected to a smaller extent, it is unlikely that the city experienced an unobserved positive development.

2019-10-15

2019-11-02

- The cash grants helped firms to recover their pre-cyclone income and profit levels
- Firms in the cyclone-affected city experience stronger effects of the cash grants than firms in the counterfactual location
- The cash grants increased firms' size
- The cash grants helped firms repair damages
- The cash grants had stronger effects for some manufacturing industries than others. For example, it might have stronger effects of firms that build houses than for enterprises that are not involved in the reconstruction processes after the disaster, e.g. tailors

Profit = Income - Expensens
Repair of damaged items such as roof, walls, machinery, raw materials

Using standard power calculation tools (see e.g. Duflo et al., 2006) in combination with knowledge about micro companies from a previous survey called IIM 2017, we aimed at interviewing a sample of around 400 manufacturing companies with less than 10 employees. Assuming the detectable effect of our treatment is relatively low (around 25 percent of a standard deviation in the outcome variable), and that we can explain 50 percent of the variation in the outcome variable by a combination of the treatment variable and covariates, a sample of 400 should yield a power of around 90 percent at a 0.05 significance level. This is a high level of power because 80 percent is acceptable in most of the cases.
Out of the total sample of 400 firms, 130 will be selected randomly for treatment, which is close to the optimal allocation of treated and control observations given the costs of collecting data for one control observation vs one treated observation: P/(1-P)=√(c_c/C_t ) where P is the share of treated firms, cc is the cost of one unit of control data, and ct is the cost of a treated unit. Imputing estimated costs of around $75 for untreated and $175 for treated observations, we arrive at 35 percent treated. The exact number of firms randomly selected via the stratified adaptive cluster sampling method will depend on how many firms from IIM 2017 can be located and interviewed.

The treatment consists of a support of around US$ 100 (~6,000 meticais), whereas untreated firms will be allocated phone credit of US$ 5 for their time. The value of US $100 is chosen to match the median yearly investments made by a micro sized firm in Beira prior to the cyclone (6,700 meticais), and ensures comparability to previous studies in the academic field (i.e. Mel et al., 2012).

Randomization in Stata

Firms

No

2 Cities

400 Firms

130 firms in treatment group (65 in Beira and 65 in counterfactual city Chimoio), 370 firms in control group (185 in Beira and 185 in Chimoio)

See above