VoxEU Column Development Environment Labour Markets Migration

New insights on distributive politics from a new way to measure aridity

Economic studies of drought episodes typically use rainfall as a proxy for water scarcity, yet this is likely to lead to an endogeneity problem due to the correlation of variables that affect water scarcity (such as temperature and soil moisture) with relevant economic choices (such as which crops to grow or where to settle). This column presents an index that overcomes endogeneity by combining precipitation and potential water loss from the soil. The index can be used to improve our understanding of distributive politics and, in particular, of how political bias affects the assignment of drought-motivated aid relief

Climate change and natural disaster mitigation are at the top of the global policy agenda.1 In recent years, there have been several calls for a more thorough economic analysis of their causes, consequences, and possible remedies (Fernando 2012, Azmat et al. 2020). Some progress in this direction has been made. The Journal of Economic Geography, for instance, devoted a special issue in 2021 to the “Economic Geography of Climate Change”. As pointed out by the editors (Peri and Robert-Nicoud 2021), the special issue stresses the heterogeneous impacts across space of climate change and natural disasters, as well as their effects on the reallocation of people and economic activity. These aspects are also discussed in several Vox columns (Kocornik-Mina et al. 2016, Peri and Sasahara 2019, Albert et al. 2021, Conte et al. 2021, Di Falco et al. 2022). Macours et al. (2012), instead, discuss how providing financial training and access to capital can help to buffer the shocks of severe weather and, hence, adapt to natural disasters.

In this column, we restrict our attention to droughts. We first discuss an important methodological aspect related to its measurement and conclude that an alternative index – the Standardised Precipitation-Evapotranspiration Index (SPEI) – is superior to precipitation-based indices of aridity. Then, we present the main results of Boffa et al. (2022), where we use the SPEI to contrast the optimal assignment of drought-motivated relief with the actual assignment, providing a deeper understanding of how distributive politics works.

The Standardised Precipitation-Evapotranspiration Index 

Water shortages are more difficult to measure than other natural calamities. Indeed, the notion is more elusive, both geographically and temporally, than it is for other natural disasters (e.g. tornados or floods). Until now, the economic literature has mostly used precipitation as a measure of water shortages, defining droughts as a long period of abnormally low rainfall. Such measures appear to be unreliable, as they lead to measurement error, inducing potentially important endogeneity issues. 

What mostly matters – indeed, both for the ecosystem and for economic activities – is the degree of moisture (i.e. the humidity of the soil), which clearly depends on rainfalls but also on factors such as ground composition and average temperatures. Importantly, those same factors are crucial for the endogenous decision to settle in a given area or to cultivate a given crop. Therefore, when precipitations are used as a proxy for aridity, the measurement error likely correlates with important economic variables that are studied in combination with droughts, such as migration and cultivation decisions. A more compelling measurement of water shortage requires the combination of rainfalls and evapotranspiration (that is, the amount of water released by the soil).2

The meteorology literature (Vicente-Serrano et al. 2010) solved this problem over a decade ago with the Standardised Precipitation-Evapotranspiration Index (SPEI), which combines precipitations and evapotranspiration (which measures loss of water from the soil, due to both evaporation and plant transpiration), thereby allowing us to obtain a proxy for the actual moisture of the ground at any point in time. 

In particular, the SPEI is computed as the difference between precipitation and potential evapotranspiration, net of their historical mean and standardised using the historical standard deviation. A dedicated website ( describes the index in detail and presents its comparative advantages vis-à-vis the most commonly used precipitation-based index: the Standardised Precipitation Index (SPI).3

The impact of evapotranspiration on ground moisture cannot be neglected both because its impact on soil moisture is significant and because it varies substantially over time. Hence, SPEI performs better than SPI both in cross-section and cross-time analyses. In particular, given that evapotranspiration depends on temperature, SPEI’s ability to account for time variability is extremely relevant in the context of global warming.

Despite its superiority compared to standard precipitation indices, only recently has it gained traction among economists. Only a few (in progress) works have incorporated it to the best of our knowledge: Cavalcanti (2018) studies the relationship between aridity and corruption; Albert et al. (2021) study labour and capital reallocation as a consequence of aridity; and Cavalcanti et al. (2022) study the impact of aridity on agricultural productivity.

Droughts and distributive politics

In a new paper (Boffa et al. 2022), we use the SPEI index to measure aridity in Brazil. Using a regression discontinuity design, we study how aridity only partially explains the assignment of drought-motivated aid reliefs, while another important role is played by political alignment.

The previous literature on distributive politics has convincingly shown that central government allocates resources to regional governments based on criteria that diverge from societal needs. Data suggest a possible bias in favour either of swing districts (Johansson 2003) or of the ruling party's strongholds (Brollo and Nannicini 2012, Bracco et al. 2015, Curto et al. 2018, Catalinac et al. 2020). 

We innovate compared to the previous literature by distinguishing between local and federal elections. We show that politicians distort the allocation of transfers only before local elections. The mechanism in our theoretical setting explains this result based on the different electoral mechanisms in place: in mayoral elections, a separate contest takes place in each municipality and what matters for the federal government is the number of municipalities won by each party. Instead, in federal elections, what matters is the total number of votes across the country (irrespectively of the distribution of votes across municipalities). Consequently, before local elections, the political strategy aims at flipping the result in swing municipalities. Before federal elections, instead, political campaigns target single voters that are easier (or cheaper) to flip. 

Importantly, by using the SPEI, we have an objective and reliable criterion for the need-based assignment of aid relief. By their nature, the discretionary transfers used in previous studies did not allow for easily building a benchmark by ranking districts in terms of their marginal need for aid. Instead, the SPEI allows us to rank municipalities by aridity and compare the need-based assignment of transfers with the implemented one. Figure 1 provides a flavour of our findings, confirmed by our regression discontinuity design. Drought-related aid relief is assigned in a non-distorted way in the period that precedes presidential elections (right chart). 

Figure 1 Political alignment and aridity


Notes: The vertical axis represents the difference between the ‘share of aligned municipalities that received aid’ and the same share for the non-aligned municipalities: positive values are observed if the share of municipalities receiving aid is larger among aligned than among non-aligned municipalities. Each dot corresponds to a different degree of aridity, measured by the SPEI. The two dashed vertical lines delimit the area defined as moderate aridity.

Interestingly, the SPEI allows us to have a better understanding of what happens before mayoral elections, when party alignment instead matters for the allocation of reliefs. It appears that aligned municipalities are favoured when the SPEI takes intermediate values, while there is no clear discrimination between aligned and misaligned municipalities when the SPEI takes extreme values (indicating either very high or very low levels of aridity). 

In Figure 2, we focus on close elections and show the difference in predicted values between municipalities that are non-aligned (blue dots) and aligned (red dots). We run separate regressions based on aridity (low, moderate, severe) and the type of election:  the top graphs refer to the two years before mayoral elections, while the bottom ones to the two years before presidential elections. Consistent with the model prediction, we only observe a discontinuity in case of moderate aridity at the time preceding municipal elections.

Figure 2 Regression discontinuity design


Notes: Graphs represent predicted values of the regression discontinuity design. The dependent variable is aid relief, the forcing variable is the margin of victory of the candidate from the party of the incumbent president in the previous mayoral election. The top three graphs show the predicted values separately for the municipalities at each aridity level in the years preceding a municipal election. The bottom three graphs represent predicted values for the years leading up to a presidential election. Circles represent the local mean and dashed lines represent 95% confidence intervals.

Concluding remarks

We have observed a growing interest in the consequences of water scarcity. However, droughts are tough to measure. We stress that precipitation are commonly used in the economic literature yet are an unreliable proxy for aridity. They are imprecise and likely to lead to biases, for temperatures and soil moisture are co-factors that explain relevant endogenous choices, such as which crops to grow or where to settle.

The SPEI (Vicente-Serrano et al. 2010) overcomes the limitations of the precipitation-based indices and resolves the endogeneity issues. However, it is not yet the standard in economics and is employed only in a few (as of now unpublished) papers. In our paper, a reliable measure of droughts is crucial to show that distributive politics is a very relevant phenomenon in specific cases (i.e. before mayoral elections, for intermediate levels of aridity), but aridity alone explains aid reliefs very well both before presidential elections and when aridity is extreme (high or low). 

Exploiting the opportunities to improve measurement, drawing quickly from developments in other disciplines, would allow economists to provide sharper policy advice. This will benefit both the profession and the policy-making activity in general, especially in delicate and important fields such as climate change and disaster mitigation.


Albert, C, P Bustos and J Ponticelli (2021), “How climate change reallocates capital and labour within countries: New evidence from Brazil”,, 26 November. 

Azmat, G, J Hassler, A Ichino, P Krusell, T Monacelli and M Schularick (2020), “A call to impact: Special Issue of Economic Policy on the economics of climate change”,,  17 January. 

Bracco, E, B Lockwood, F Porcelli and M Redoano (2015), “Intergovernmental grants as signals and the alignment effect: Theory and evidence”, Journal of Public Economics 123: 78–91.

Brollo, F and T Nannicini (2012), “Tying your enemy’s hands in close races: The politics of federal transfers in Brazil”, The American Political Science Review 106: 742–761.

Catalinac, A, B B de Mesquita and A Smith (2020), “A tournament theory of pork-barrel politics: The case of Japan”, Comparative Political Studies 53: 1619–1655.

Cavalcanti, F (2018), “Drought and Corruption”, Mimeo.

Cavalcanti, F, S Helfand and A Moreira (2022), “Climate Change, Drought and Agricultural Productivity”, Mimeo.

Conte, B, K Desmet, D K Nagy and E Rossi-Hansberg (2021), “Trade to adapt: Changing specialisation to cope with climate change”,, 4 May. 

Curto-Grau, M, A Solé-Ollé and P Sorribas-Navarro (2018), “Does electoral competition curb party favoritism?”, American Economic Journal: Applied Economics 10: 378–407.

Dell, M, B F Jones and B A Olken (2014), “What do we learn from the weather? The new climate-economy literature”, Journal of Economic Literature 52(3): 740-98.

Di Falco, S, A B Kis and M Viarengo (2022), “Climate change, weather shocks, and migratory flows”,, 30 April. 

Fernando, M (2012), “An Advisory Paper for Group 8 and Group 20 Humanitarian Healthcare Diplomacy. Humanitarian Relief to Assist Sub-Saharan Nations with Drought Relief”,, 24 January. 

Johansson, E (2003), “Intergovernmental grants as a tactical instrument: empirical evidence from Swedish municipalities”, Journal of Public Economics 87: 883–915.

Kocornik-Mina, A, T McDermott, G Michaels and F Rauch (2016), “Do floods shift economic activity to safer areas?”,, 21 January.

Macours, K, P Premand and R Vakis (2012), “Helping poor farmers help themselves against drought”,, 12 September. 

McKee, T B N, J Doesken and J Kleist (1993), “The relationship of drought frequency and duration to time scales”, Proceedings of the 8th Conference on Applied Climatology 17(22): 179-183.

Peri, G and F Robert-Nicoud (2021), “On the economic geography of climate change”,, 11 October. 

Peri, G and A Sasahara (2019), “The effects of global warming on rural-urban migrations”,, 15 July. 

Vicente-Serrano, S M, S Beguería and J I López-Moreno (2010), “A multiscalar drought index sensitive to global warming: the standardized precipitation evapotranspiration index”, Journal of Climate 23(7): 1696–1718.

Young, M (2010), “Water economics and management: lessons from Australia’s drought”,, 22 January.


1 Australia may have been one of the first countries to implement large-scale reforms to face the effects of droughts (Young 2010), although many countries have been dealing with them at the local level for many years already.

2 This fact was already stressed in Dell et al. (2014), footnote 24.

3 The SPI is based on a precipitation probabilistic approach introduced by McKee et al. (1993).

420 Reads