Public employment programs have a long history as a counter-cyclical policy instrument to reach and protect the poor in times of crisis or during the lean season. Simply put (and perhaps this is World Bank-speak) public works have traditionally been about protection, and not promotion. In his review of the evidence from the early experience of guarantee schemes in South Asia that preceded the NGREGA program in India, Ravallion 1991 noted that “rural public works have had a long and generally successful history as an instrument of seasonal stabilization and famine relief in India,” and that there is “ample evidence from South Asia that relief work can save the rural poor from potentially disastrous effects of a sudden contraction in real incomes from other sources”.
But might they do more than provide relief in crisis? Is there evidence that they can result in sustained poverty reduction? These are questions that Bagga et al (2023) put forth in their new working paper and its meta-analysis of 11 recent experimental studies of public works. They find that the short-run effects on employment and earnings from public works fade in the medium term.
Their paper got us thinking about what were the conditions under which we would expect medium-term effects to materialize from public works. A good starting point for this is Gehrke and Hartwig (2018), who discuss four mechanisms through which such programs might increase livelihoods of the poor in the medium run (“productive effects”), and critically assess if public works might generate gains to the poor that go beyond those of cash transfers programs. Let us review them one at a time:
The income and insurance effect is when households save or invest their public works earnings in livelihoods (be it farms, household enterprises, or human capital increases). Toward this possibility, do we see where the public works income goes in terms of spending and saving? Sort of. Bagga et al find that food consumption increases for only two out of six programs in their review. But this might not be evidence that households do not spend their public works income on food. High liquidity constrained households may spend the cash earned soon after payment is received, evading detection in a survey administered several weeks or months later, as reported in qualitative work in the Comoros and Malawi studies in the meta study.
If not spent on food, maybe this income is saved or used to buy assets. This pathway is unlikely to be quantitatively important, as income transfers are generally small in workfare programs and of short duration. Part of the story here is the opportunity cost of public works: the net gain to public works beneficiaries is lower that the potential gain if there are foregone earnings. Bagga et al find that the earnings impact is, on average, 46 percent of potential program earnings, pointing to substantial forgone earnings. Interestingly Ravallion 1991 also found that the net income from public works was about half of potential program earnings. A shift from short-term toward an employment guarantee scheme might enable these programs to have more of an investment and/or insurance role. The ex-ante insurance effect, for example, has been documented in India, where access to guaranteed employment (NREGA) allowed farmers to shift their production towards riskier and more profitable crops (Gehrke 2019).
The indirect wage floor effect is a second pathway for a productive impact of public works. This operates when the option value of having access to public works drive up wages in the private sector (especially if the program offers higher-than-market wages). The increase in private wages documented in India depends critically on the lack of rationing: that everyone who is willing to work at the offered wage can get it. These general equilibrium effects have been shown to be sizeable (as in Muralidharan et al 2022), but require strong government capacity (and financing) to make work available and with good implementation quality. However, offering higher wages may run against the notion of pro-poor self-selection into the program (and thus avoiding implementing targeting schemes to reach poor households).
The skill acquisition effect operates through on-the-job training. Most public works programs entail low-skilled work for a fairly short duration, suggesting scant likelihood of a transfer of skills. The instances in which the skill transfers seem to persist is when public works programs are complemented by specific training programs that help transition beneficiaries into the labor market (as in Argentina and Côte d’Ivoire). A variation of this pathway, not explicit in Gehrke and Hartwig, is labor market activation – where the program provides labor market experience to traditionally excluded populations, such as women or unskilled youth, who stand to benefit the most from workfare programs given their lower opportunity cost. The Djibouti public works program in the meta-analysis targeted young mothers as an excluded group, successfully offering them this short-run employment but this effect did not sustain after the program ended. So public works alone might not be sufficient to tackle challenging constraints, such as social norms, inhibiting activation into the labor force.
Lastly, indirect benefits from the value of the assets created (such as improved market access through road rehabilitation), could go beyond the income benefits to beneficiaries. Often there is a tradeoff between labor intensive public works and such indirect benefits. Here the evidence is scant so far and estimated to be zero or rather small on aggregate agricultural productivity. But benefits might be thought of as much broader. It is encouraging to see studies using auxiliary data, such as remote sensing data and large-scale administrative data, to measure the environmental impact of public works assets that specifically target mitigation, such as in Hirvonen et al 2022.
From their review of these mechanisms, Gehrke and Hartwig conclude that “standard short-term PW interventions are insufficient for fostering productive investments among participants”. Indeed, most of the interventions in Bagga et al, who do not find medium-term impacts, are short-term programs of limited scale. To get medium-term effects, these programs likely need to be ‘bigger’ in some fashion. This could be pivoting towards an employment guarantee scheme (where accounting for the indirect effects might tilt the cost-effectiveness comparison with cash transfers), offering complementary measures such as training (hard or soft skills) and/or financial inclusion interventions (“public works plus”), paying higher wages for more days of work, shifting to less labor intensive public assets…. But, of course, these options come with tradeoffs. So we are left with the question of not only how, but if, one should build off public works programs to gear them towards promoting medium-term poverty reduction at the risk of compromising the traditional intent of public works to protect in times of crisis, versus investing in other approaches to productive inclusion/graduation.