What can (or should) we expect from public works? Part 2


This page in:

Back in July, we blogged about the conditions under which we would expect medium-term effects to materialize from public works in low-income country contexts, with a focus on productivity. In their forthcoming AER paper, Franklin, Imbert, Abebe, and Mejia-Mantilla 2023 (gated and ungated) provide an important contribution to the evidence on the impact of public works, with a well-designed at scale experiment. And they do so in some notably unique ways, including in terms of the setting (an urban context), the scale (the program employed nearly 4% of all adults in the city, and about one-in-five households in treated areas), and the way they account for indirect effects in the evaluation of anti-poverty programs (which in this urban setting takes the form of commuting and wage spatial spillovers).

There is a lot to like in this paper, and we will not provide an exhaustive summary of the rich set of results. We encourage you to read it. But, rather, we will summarize what they find (but not detail how they come to their findings). And, along the way, we highlight a few points of interest (to us anyway).

Five findings that speak in part to the ‘standard’ outcomes of interest from public works programs

(i) take-up was basically universal, and mostly done by women --especially older adults, and especially women, and partly reflecting the community targeting to poor “female headed households”, with presumably no resident adult male—though the average household has 3.5 working-age adults;

(ii) public works did bring some people into the labor market (on the extensive margin, though the boundaries between inactivity and unemployment and employment in this unskilled urban labor market are porous);

(iii) public works crowded out both wage employment and self-employment in terms of participation and hours (so total labor time did not increase; true for both men and women);

(iv) household income increased (even with no change in total labor supply thanks to a generous program wage rate relative to private wage rates);

(v) households saved most of the income gains rather than spend them (possibly in light of the short-term nature of the intervention).

A spatial model-based approach to identify and quantify the role of indirect effects

The impact on private sector wages: The scale of the program (reaching 18% of households in treated areas) implies that the reduction on private labor in each destination is a large negative labor supply shock to private employers. This is because more than 80% of workers in their sample at baseline (that are poor households in Addis) are wage employed (rather than self-employed). By collecting data on the neighborhoods where workers work, the authors build a spatial equilibrium model to estimate spillovers. Private sector wages increased overall, and also in control neighborhoods that employed commuters from treated areas. Ignoring labor market spillovers would significantly underestimate the impact of the program.


The indirect impact of the program on local amenities: The authors introduce a subjective assessment of neighborhood quality that might have been affected by the program (drainage, cleanliness stress, public toilets, odors from sewerage and trash) and estimated a sizeable improvement. It is great to see this effect assessed in the context of public works evaluations. Older literature (building on the seminal work by Martin Ravallion and co-authors) emphasized that the cost/benefit comparison with cash transfers should consider the value of assets created. These subjective estimates are likely to be a lower bound of the potentially sizeable health externalities from infrastructure upgrading in sanitation and waste management urban settings.

The substantial role that the spatial wage spillovers play in the welfare effects from the program opens up a new and important differentiation with cash. We hope this gets studied in other settings too, with a focus on identifying under what condition the indirect wage effects arise. When they model both the wage impact as well as the value of the local amenity, the welfare gains quadruple the direct benefits!

Some final reflections about the context and extrapolating the results to other urban labor markets in the region

The extent of wage employment in urban Addis caught our eye. Our priors would be that a much (much) larger share of those working in this sample would be self-employed. Indeed, other sources report a much low share of urban employment in wage jobs in Ethiopia (such as 49% from the Global Jobs Indicator Database; and almost undoubtedly much lower when looking at the poorest households with lowest education). Maybe Addis has much higher wage employment among poor households than the rest of the country (and compared to other big cities in the region). For settings with a larger share of self-employment, it is not clear if indirect benefits from higher wages would emerge.

This had us thinking about four other traits of their study and this context which were perhaps different than our priors about a public works program in a large African city, and how these differences relate to the findings of this study.

Consider that most of the public works workers were women, often older. It is well-documented that women bear extra burdens due to child-care demands. But Addis has notably lower fertility (a total fertility rate of 1.8, below replacement rate), compared to, for example 2.6 in Nairobi, 3.2 Abidjan, and 2.9 in Accra (stats from latest DHSs). So this might drive up take-up among women in this setting (among other factors such as norms around mobility, in the case of comparisons to India, for example).

Also, three quarters of these households live in government housing. It is rationed and comes with very low rent- likely driving down mobility as well as perhaps being a disincentive to invest extra earnings into dwellings. This strikes us as a very unique housing situation for poor families in urban Africa. This almost-free public housing relates to commuting since it disincentivizes moving closer to jobs. For commuting workers (more than half of all workers in their sample travel to other neighborhoods for work), commuting costs are 6% of earnings. As public works is offered locally, it also increases take-home pay through savings on paying for transport (larger direct effect) and changes the supply of labor from other neighborhoods (larger indirect effect).

Lastly, another surprising finding about their sample is that their population is not characterized by a labor surplus. The authors carefully acknowledge that the size of the indirect effect depend crucially on the extent of crowding out of the labor supply to the private sector. The finding that household labor supply is inelastic really stood out to us. Recall that these are poor household – they live in 1.2 rooms on average, 80% do not have an improve toilet, and 5% of them have a head with secondary schooling. Just under half of the working age adults in these households have any employment, and average working hours per week is about 21 hours. Would we expect to find inelastic labor supply? Probably not. Yet, on the other hand, the beneficiary sample (which is identified through local committees) is over represented by households with female-headed (likely often widows we presume) maybe pointing to this inelasticity relates to the specific of needs and vulnerabilities identified by the local committees selecting beneficiaries: sixty percent have no resident prime-age male, a third have a child under 5, the average age of the head is 56 years, and 17% have a disabled household member. It leaves us wondering if such inelastic labor supply among the poor is unique to this specific profile of households in Addis, or true in general, as compared to other cities in Africa. Notably, low elasticity is consistent with other findings of low take-up and high turnover in (relatively) well-paid factory jobs in Ethiopia, not specifically targeted to this profile of households.


Kathleen Beegle

Research Manager and Lead Economist, Human Development, Development Economics

Emanuela Galasso

Senior Economist, The World Bank

Join the Conversation