Disruptions in agricultural production often lead to detrimental effects for many households in developing countries, with serious repercussions for young children. This amplifies the negative impact of these shocks since conditions experienced early in life have long-lasting effects on various socioeconomic outcomes. In particular, it is well established that harsh conditions experienced in utero can have detrimental and persistent effects on health throughout the whole life cycle (Lavy et al., 2020, Almond and Currie, 2011, Maccini and Yang, 2009, Stein et al., 1975).

In this paper, we study the impact, on child health, of exposure to an agricultural shock that is indirectly linked to extreme weather conditions and climate change: the damage caused by desert locust plagues. Locust swarm invasions are destructive events that recurrently put food supply in many developing countries at risk (especially in sub-Saharan Africa, see Brader et al., 2006). They are caused by a specific species of grasshoppers that usually live, in their solitary phase, around the Sahara desert. Under favorable breeding conditions (excess rainfall), these grasshoppers go through a gregarization process with substantial changes in their behavior, morphology, and physiology. They become more voracious and can grow into huge swarms that travel to less arid areas to feed and reproduce.

We focus on how this shock operates in areas with a single harvest per year.1 We argue that agricultural shocks due to pest invasion can affect households living in this type of economy through two main channels. The first channel is a speculative/anticipatory effect that kicks in immediately during the growing season in which the plague is occurring, in anticipation of the upcoming harvest failure. At this point, households and markets are still relying on the harvest from the previous agricultural season. Yet, crop destruction by the pest in the ongoing season could lead to an anticipation of a future crop production decline. This “bad news” may affect the supply and demand of crops on the markets before the actual shock.2 The second channel is the actual crop failure effect that would constitute an income shock for farmers and a supply shock for markets after the harvest of affected crops. This effect should last at least until the following harvest.3 We can separately identify these two effects because they kick in at different points in time.

We explore in detail these two effects, by making use of a sharp episode of locust plague invasion and tracing out how it affected the health of children exposed to it in utero. For that, we rely on the timing and location of locust swarm events that occurred between June and December 2004 in Mali to identify the temporal and spatial variation in the exposure of different cohorts of children to the plague. Data on locust swarm invasions come from the locust monitoring system run by the Food and Agriculture Organization (FAO) Desert Locust Information Service (DLIS). Geocoded household survey data with detailed information on the timing of birth and health outcomes of children is taken from the Demographic and Health Surveys (DHS). We also link this data to local agricultural crop price data from the Malian Agricultural Market Observatory (Observatoire du Marche Agricole - OMA) to tease out potential local price effects.

In the first part of our empirical analysis, we start by showing that the plague led to significant inflation of crop prices (average increase of 8.4%) in affected areas, compared to non-affected areas, during the growing season of 2004.4 We take this result as empirical evidence of the speculative price effect in treated markets: during that period, local markets were still relying on the previous harvest that has not been affected by the plague. Therefore, we can infer that the estimated price inflation is caused by the expectations of agents about a potential failure of the upcoming harvest. The magnitude of this price inflation is consistent with the impact of market interventions on crop prices. Gross et al. (2020) show that the random scaling-up of a food security program that buys grain from outside sources and sells it locally in poor and isolated areas of Burkina Faso decreased crop prices by 6% on average. The magnitude of our price inflation is also consistent with the impact of other agricultural shocks such as droughts on crop prices. Kudamatsu et al. (2016) shows for instance that a severe lack of growing-season rainfall raises staple crop prices by 7 to 10% in Sub-Saharan Africa.

We then turn to the core of our analysis and explore the impact of locust invasion on child health. First, using a Difference-in-Differences identification strategy, we show that children exposed in utero to the adverse effects of the plague suffered major health setbacks. They had, on average, a height-for-age Z-score 0.42 points lower than non-exposed children. This represents around 30% of the average height-for-age Z-score. Our estimates suggest an increase in the average stunting rate by more than 20%. We find no impact on cohorts of children exposed to the shock after birth.

In the second part of this analysis, we study the timing of the estimated effect by quarters of birth. Our results show that cohorts of children that were subject only to the speculative/anticipatory effect in utero suffer as much as those exposed to the actual crop failure effect.5 Importantly, we argue that the health effects of the exposure (in utero) to the speculative/anticipatory effect go beyond the documented local crop price inflation effect for two main reasons. First, it includes (potentially) any precautionary consumption smoothing effect: a decrease in grain consumption driven by households smoothing the impact of the anticipated consumption shock over time. Second, it also captures the impact of in-utero exposure to the stress/anxiety that the news of an imminent shock (harvest failure in our case) can bring to households (Torche, 2011, Talge et al., 2007, Tapsoba, 2023). We are not able to tease out these mechanisms with our data (this is beyond the scope of this paper). We argue nonetheless that the documented significant inflation effect in anticipation of the future crop failure cannot be explained by any of these two alternative mechanisms or by other alternative factors such as increased use of pesticides, a devotion of local resources to controlling the invasion, limited access to health care, etc.6 Moreover, we show that the estimated treatment effect (on child health) of exposure to the speculative/anticipatory effect is partly absorbed when we account for the price variation.

The extent to which local markets are isolated from other sources of agricultural supply also plays a crucial role in this context. In particular, we find that exposed children born in isolated areas, with limited access to crops from other areas, suffer more compared to those born in well-connected areas. This pattern is driven by the crop failure effect. We found no differential effect of treatment by the level of isolation of local markets for children exposed only to the speculative/ anticipatory effect.7

Our results are robust to specifications that include region-specific time trends, household and mother characteristics, climate shocks, household fixed effects, and mother fixed effects. We also argue that they are not likely to be biased by potential migration, pre-existing differences in trends between treated and non-treated areas, and exposure to anti-locust pesticide spraying. Results are also robust to restricting the analysis to male–female and rural–urban sub-samples.

The distinction between the speculative/anticipatory effect and the crop failure effect is crucial when we consider how this shock affects time-sensitive health investments such as those that households make when they have children in utero. The anticipation of future negative shocks such as a crop failure at the next harvest may lead to persistent health effects for entire cohorts of children that would otherwise not suffer any detectable effect. The locust plague episode studied in this paper provides a setup that allows us to document this speculative/anticipatory effect separately from the actual crop failure effect. This effect is also potentially present in other types of agricultural shocks such as droughts or floods since agents often form strong beliefs about the next harvest outcome early in the rainy season.8 A common illustration of such behavior is when farmers decide to replant some seeds in the middle of the season hoping to receive the right amount of rainfall for the new crops to produce as well and help limit the damage of the anticipated shock. This legitimate anticipation of potential negative shock may however lead to persistent health consequences for vulnerable household members such as children in utero.9

The findings of this paper also have important policy implications. In particular, we provide evidence of the existence of a strong speculative/anticipatory effect that operates differently than the actual crop failure effect when agricultural shocks such as locust plagues occur. This calls therefore for different types of policy reactions. Fighting the speculative behavior of intermediaries is crucial during the growing season when the overall crop supply on markets is at its lowest level. Conversely, after the harvest period, policy action should focus on coping with the local crop failure shocks. Our findings also suggest that easy and diversified access to agricultural production from non-affected areas can effectively mitigate this effect.

This paper contributes to two main strands of the economic literature. First, it contributes to the literature on the importance of early-life conditions (Lavy et al., 2020, Maluccio et al., 2009, Black et al., 2007, Behrman and Rosenzweig, 2004, Stein et al., 1975). A substantial part of this literature focused on identifying the effect on child health of exposure to weather shocks (Adhvaryu et al., 2019, Maccini and Yang, 2009), violence and civil wars (Tapsoba, 2023, Dagnelie et al., 2018, Quintana-Domeque and Ródenas-Serrano, 2017, Koppensteiner and Manacorda, 2016, Valente, 2015) or adverse institutional setup (Kudamatsu, 2012). The health impact of desert locust swarms has also been investigated more recently in this literature.10 (Le and Nguyen, 2022) analyze the impact of prenatal exposure to locust events in 39 countries in Africa and Asia between 1990 and 2018. Linnros (2017) estimates the aggregate effect of locust infestations on child health using data from Senegal, Burkina Faso, Mali, and Niger. Importantly, our contribution goes beyond documenting the average impact of locust swarms: it lies in the careful investigation of the underlying channels through which pest invasions affect household welfare in developing countries. Our paper is the first one to document evidence of a speculative/anticipatory effect on top of the crop failure effect for this type of shock.

Second, this paper is related to the literature on the effects of news about the future production of seasonally produced commodities on competitive storage behavior and prices (Osborne, 2004, Chambers and Bailey, 1996, Deaton and Laroque, 1992, Williams et al., 1991). We borrow our theoretical framework from this literature. Deaton and Laroque (1992) present a supply and demand model for commodities with competitive speculators who hold inventories in the expectation of making extra profits when selling in the future. Osborne (2004) uses a structural model to explain dramatic seasonal price swings and a high degree of serial correlation in commodity price data. They show that the fact that markets incorporate news about future production lowers variation in prices without substantially increasing the mean price. We contribute to this literature by showing how news about a one-off shock to future production can affect current consumption, storage behavior, and prices. Moreover, irrespective of efficiency concerns in the market behavior, this type of anticipation may still affect time-sensitive investments and lead to long-term damages. In our case, this effect comes from the fact that in-utero conditions get worse for an entire cohort of children that would otherwise be exposed to the shock only after birth and hence not suffer any detectable effect in the medium/long run.

The paper proceeds as follows. Section 1 provides some background on locust plagues. Section 2 discusses the conceptual framework and the channels through which locust invasions can affect the well-being of households. Section 3 presents the data. In Section 4 we study the impact of locust plague on local crop price, while in Section 5 we analyze its impact on child health. Section 6 concludes.