High Frequency Engel and Supply Curves in General Equilibrium: Experimental Evidence from Large Universal Cash Transfers in Malawi

How to cite

Egger, Dennis; Kastrau, Philipp; Siddiqi, Bilal; Stein, Daniel; Graff, Tilman; Walker, John; Mughogho, Winnie. 2023. "High Frequency Engel and Supply Curves in General Equilibrium: Experimental Evidence from Large Universal Cash Transfers in Malawi." AEA RCT Registry. AEARCTR-0012550. https://doi.org/10.1257/rct.12550-2.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Dennis Egger* , Philipp Kastrau , Bilal Siddiqi , Daniel Stein , Tilman Graff , John Walker , Winnie Mughogho

* Primary investigator

Registry ID
AEARCTR-0012550
DOI
https://doi.org/10.1257/rct.12550-2.0
First registered
Last updated
Status
On going
Trial dates
2023-10-13 – 2028-12-31
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/12550

Abstract

Understanding how firms and markets shape local supply curves in general equilibrium, to which degree economies can absorb (and change in response to) large local demand shocks, and how consumers shopping patterns and competitive pressures respond is crucial for understanding long-term development and the effects of macroeconomic policies. We estimate the high-frequency causal impacts of universal USD 550 cash transfers to every adult in the Khongoni (Phase I) and Chiradzulu Districts (Phase II) in Malawi on household outcomes, including expenditure, enterprise outcomes, and on market activity and prices. The rollout of the approx. 100 million in cash transfers to approx. 200,000 recipients by the NGO GiveDirectly is randomized, and treatment intensity varies across three geographical levels: a) the traditional authority, b) Group Village Headman (GVHs), an administrative unit corresponding approximately to a market catchment area, and c) villages.

Using random variation in treatment timing, together with high-frequency price data for over 100 products (monthly or bi-weekly), and monthly enterprise surveys in all markets within the study area and a matched sample of markets outside the study area, as well as monthly expenditure surveys with eligible households, we document high-frequency marginal propensities to consume by product (Engel curves), inflationary impacts, expenditure switching of households across markets, and changes in resulting market activity and the productive environment. This will allow us to trace out short- and medium run supply curves at the firm- and market level.

In Phase II, we additionally inform a random subset of traders in treated market ex-ante about upcoming demand increases (the IRFs estimated in Phase I). Predicted demand shocks naturally vary in size, and we additionally vary the salience of the temporary or permanent nature of the shock. This allows us to test whether demand predictability and variability affect firm and market-level supply dynamics, and draw inference on the competitive environment within which firms set prices.

Experimental design

not public until completion of the intervention to maintain the integrity of the study.

Design details

Randomization method
The order was randomized by a random number generator on a computer (STATA)
Randomization unit
The Khongoni District in Malawi is divided into Group Village Headman (GVH) administrative units. GVHs have a population of 2000 adults, or approximately 800 households on average. Treatment is randomized at the level of the GVH, and all households within a GVH. Enrolment for all GVH included in a treatment wave will happen within the same calendar month, and transfers will typically be initiated the last Tuesday of the month after enrolment which we term experimental start date (though there may be some delays due to verification, etc.)
Clusters
Phase I: Treatment is assigned at the GVH level, randomly across 25 GVHs (and 9 additional GVHs which were treated non-randomly before the start of the experiment). Phase II: Treatment is assigned across 10 TAs, 113 GVHs, and approx. 900 villages as described above. Information treatments are randomized at the individual enterprise level (though for later expansions, we may consider clustering treatment intensity at the market-level).
Observations
Phase I: Households: We will survey 2500 households (100 per GVH) for a total of 9 rounds, or a total of 22,500 hosuehold-by-month observations Prices: We will collect up to 2 price quotes for 82 goods and services at 51 markets for a total of 24 bi-weekly rounds, or a total of over 200,736 price quotes. Market sellers: We will survey 20 sellers at each of the 51 markets, monthly for a total of 12 rounds, or a total of 12,240 seller-by-month observations. Phase II: This is split into two stages, 344 villages in stage I, and approx. 600 villages in stage II. Households: We will census all households at baseline, and survey 8 households per village, 2752 in stage I, and approximately 8,000 households across both stages. Households will be surveyed in person at baseline, as well as 1-2 months after transfers, 5-6 months after transfers, and 11-12 months after transfers. Enterprises: We will census all enterprises operating in all approx. 80-100 weekly markets within Chiradzulu, monthly during the period of most intense transfers (approx. 2-3 months prior, and up to 6 months after the large transfers), and bi-monthly or less frequently thereafter. We will survey approx. 50 sellers per market in a panel of over 4,000 sellers to be followed over time.
Arms
Phase I: Details in PAP Phase II: Details in PAP
Primary outcomes
- household expenditure (by product group, by location/market) - output price inflation (by product group) - market activity (total revenue, profits, number of customers, labor supply) - imports (intermediate input purchases from within vs. outside the study area) - child mortality - other household outcomes, including assets, labor supply, food security, mental health, female empowerment, etc.

Cited by

No citation data available for this registration yet.

Keywords