The Effects of Weather Shocks on Economic Activity: What are the Channels of Impact?1

Sebastian Acevedo Mejia; Mico Mrkaic; Natalija Novta; Evgenia Pugacheva; Petia Topalova

doi:10.5089/9781484363027.001.A001

Jump to Content Jump to Main Navigation

The Effects of Weather Shocks on Economic Activity: What are the Channels of Impact?¹

Author: Mr. Sebastian Acevedo Mejia, Mr. Mico Mrkaic, Natalija Novta, Evgenia Pugacheva¹, and Petia Topalova²

¹ 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund
| ² 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

Author’s E-Mail Address: SAcevedoMejia@imf.org, MMrkaic@imf.org, NNovta@imf.org, EPugacheva@imf.org, PTopalova@imf.org.

Publication Date:: 13 Jun 2018

eISBN:: 9781484363027

ISBN:: 9781484363027

Language:: English

Keywords:: WP; economic activity; dependent variable; growth rate; development centre; temperature projection; production function; Increase-A Model-Based

Download PDF (7 MB)

Global temperatures have increased at an unprecedented pace in the past 40 years. This paper finds that increases in temperature have uneven macroeconomic effects, with adverse consequences concentrated in countries with hot climates, such as most low-income countries. In these countries, a rise in temperature lowers per capita output, in both the short and medium term, through a wide array of channels: reduced agricultural output, suppressed productivity of workers exposed to heat, slower investment, and poorer health. In an unmitigated climate change scenario, and under very conservative assumptions, model simulations suggest the projected rise in temperature would imply a loss of around 9 percent of output for a representative low-income country by 2100.

Abstract

I. Introduction

Since the turn of the 20th century, the Earth’s average surface temperature has increased significantly. The speed at which this increase has taken place in the past 30 to 40 years appears to be unprecedented in the past 20,000 years (Figure 1).² Most scientists agree that global temperatures are set to rise further, at a scale and pace very much dependent on our ability to restrain greenhouse gas (GHG) emissions, the central cause of global warming (IPCC 2014). Although considerable uncertainty surrounds temperature projections, the scientific consensus predicts that without further action to tackle climate change, average temperatures could rise by 4°C or more by the end of the century. Very substantial cuts to current emissions will be needed to limit warming to less than 2°C. Will climate change have significant macroeconomic consequences, especially in low-income countries that tend to be more exposed to the vagaries of the weather? And how can these countries cope with the rises in temperature they are set to experience over the coming decades?

Pinning down the economic consequences of climate change is difficult. Temperature increases of the magnitude that could potentially occur over the next century—and many other aspects of climate change, such as rapid rise in sea levels, ocean acidification and the like—sit well outside recent (and relevant) historical experience and could affect a large number of countries. Extrapolating from historically observed relationship between activity and weather patterns could also be problematic as populations adapt to persistent changes in climate. Yet, studying the macroeconomic effects of annual variation in weather patterns could produce useful insights.³ In an influential study, Dell, Jones, and Olken (2012) find that higher temperatures significantly reduce economic growth in low-income countries. Burke, Hsiang, and Miguel (2015a) provide evidence that productivity peaks at about 13°C and declines strongly at higher temperatures. Since low-income countries are concentrated in geographic areas with hotter climates, the Burke, Hsiang, and Miguel (2015a) findings suggest that a rise in temperature would be particularly harmful for this set of economies.

In this paper, we provide new evidence on the effects of weather shocks on economic activity, the persistence of these effects and the channels through which they operate. Using data from more than 180 economies over 1950–2015 and following the methodology pioneered by Dell, Jones, and Olken (2012) and Burke, Hsiang and Miguel (2015a), we exploit the annual variation in temperature and precipitation to estimate their causal effect on aggregate economic activity and output of various subsectors of the economy in the short and long run. We study how weather shocks affect the key elements of the aggregate production function, namely productivity, capital accumulation and labor supply, to trace the mechanisms through which weather fluctuations shape aggregate output. Finally, we incorporate the empirical estimates of economic loss from weather shocks and projected changes in temperature into a dynamic general equilibrium model to trace the potential long-term effects of climate change for a typical low-income country.

Our findings suggest that the macroeconomic effect of temperature shocks is uneven across and within countries. Confirming the global nonlinear relationship between annual temperature and growth uncovered by Burke, Hsiang, and Miguel (2015a) in our expanded data set, we find that rising temperatures lower per capita output in countries with relatively high annual average temperature, such as most low-income countries. In these economies, the adverse effect is long-lasting and operates through several channels, namely lower agricultural output, depressed labor productivity in sectors more exposed to the weather, reduced capital accumulation, and poorer human health.

Model simulations suggest that the temperature increase projected by 2100 in a scenario of unmitigated climate change implies significant economic losses for most low-income economies. Under the conservative assumption that weather shocks have permanent effects on the level, rather than the growth rate, of per capita output, model simulations suggest that the per capita GDP of a representative low-income country would be 9 percent lower in 2100 than it would have been in the absence of temperature increases, with the present value of output losses amounting to more than 100 percent of current GDP when discounted at the growth-adjusted rate of 1.4 percent.

Our study contributes to several strands of literature. First, it reexamines the findings of Dell, Jones, and Olken (2012) and Burke, Hsiang, and Miguel (2015) with expanded geographic and temporal coverage, using a more flexible empirical specification, and several sources of weather data. We build on their analysis by examining the channels through which temperature variations affect economic outcomes, using sectoral data to identify productivity effects and infant mortality data to study health impacts. Second, we complement the literature on the effects of weather shocks that previously relied on experiments (see, for example, Seppänen, Fisk and Lei 2006, Niemelä et al. 2002 for the effects on productivity) or within-country studies (see, for example, Barreca 2012, Burgess et al. 2014 for the effect of weather on health outcomes, Schlenker and Roberts 2009, Burke and Emerick 2016, and Wang et al. 2017 for evidence on agricultural output, Deryugina and Hsiang 2014, and Somanathan et al. 2017 for evidence on productivity) and demonstrate that many of the effects uncovered for individual countries or in experimentally controlled settings exist across countries and are important for aggregate outcomes. Finally, we contribute to the large literature that uses models to analyze the effects of natural disasters and climate change on macroeconomic outcomes (Füssel 2010).⁴ Rather than an integrated assessment model (IAM), we use a dynamic general equilibrium model to simulate the potential long-run effects of climate change for a typical low-income country. Existing IAMs are typically not geographically granular enough, lumping together economies with different income levels and average temperatures. They also include various feedback loops between emissions, growth, and climate that are less relevant for low-income countries. The model we use overcomes these concerns and can be easily modified to analyze sectoral issues and structural economic transformation.

The rest of the papers is organized as follows. To set the stage, section II presents some key stylized facts about historical patterns of temperature and precipitations as well as scientific projections of future changes. Section III describes the data and lays out the empirical strategy used to assess the macroeconomic effect of weather shocks. Section IV presents the main findings and several robustness checks of the empirical results, while Section V looks at the channels through which aggregate economic output is affected. Section VI discusses the potential long-run impacts of temperature increase. Section VII concludes summarizing the main findings and policy implications.

II. Temperature and Precipitation: Historical Patterns and Projections

Global temperatures have increased by roughly 1°C compared with the 1880–1910 average (Figure 2).⁵ The rise started in earnest in the 1970s, following a large increase in carbon dioxide (CO₂) emissions.⁶

The increase in average annual temperature has occurred in all regions, with the same accelerating trend starting in the 1970s (Figure 3, panels 1, 3, and 5). The median temperature over the first 15 years of this century, compared with the first 15 years of the past century, was 1.4°C higher in advanced economies, 1.3°C higher in emerging market economies, and 0.7°C higher in low-income countries. Even though most of the warming happened in advanced economies, by 2015 the temperature in the median low-income country (25°C) was more than twice that of the median advanced economy (11°C). Trends in precipitation are generally less clear (Figure 3, panels 2, 4, and 6). Precipitation has increased somewhat in the northern hemisphere since the 1950s, and precipitation in the median low-income country has declined since the 1970s.

The Intergovernmental Panel on Climate Change (IPCC) constructed four possible scenarios, called Representative Concentration Pathways (RCP), using alternative GHG concentration assumptions to project likely ranges of temperatures over the 21st century. The rest of the paper focuses on two of these scenarios: an intermediate path (RCP 4.5) and an unmitigated path (RCP 8.5), as shown in Figure 1, panel 2.⁷

Under the RCP 8.5 scenario of unmitigated climate change, the average global temperature by 2081–2100 could rise by 3.7°C (with a projected range of 2.6°C–4.8°C). Warming would occur all over the globe, with larger increases over the northern hemisphere, where some regions could experience temperatures almost 12°C higher than in 2005 (Figure 4). Between 2005 and 2100, the increase in temperature for the median advanced economy is projected to be 4.4°C, and 4.5°C for the median emerging market economy and median low-income country.

Temperature increases are projected to be smaller in absolute terms closer to the equator, but are very significant when set against the historical year-to-year and intrayear variability in temperature observed in those locations. For example, the historically observed natural year-to-year temperature variability for countries located in the tropics is roughly 0.5°C. The projected increase in temperature for these countries between 2005 and 2100 under the extreme unmitigated climate change scenario is 4.1°C—in other words, more than 8.5 times larger than the current natural variability, implying a totally new climatic regime (see also World Bank 2013). Change in precipitation is projected to vary by region, with dry areas generally expected to become drier and wet regions expected to experience an increase in rainfall.

III. Empirical Strategy and Data

In the absence of historical experience with climate change that may be relevant for countries today, we build on existing literature and identify how annual fluctuations in temperature and precipitation affect macroeconomic performance. Using the approach of Dell, Jones, and Olken (2012) and Burke, Hsiang, and Miguel (2015a), we use within-country and across-country year-to-year fluctuations in temperature and precipitation to identify their causal effect on aggregate outcomes, both contemporaneously and over the medium term. We build on these studies by expanding the geographic and temporal coverage of the analysis to more than 180 economies during 1950–2015 (see Annex 1 for a list of countries and territories), examining the effects of weather shocks on a larger set of outcome variables, establishing the robustness of findings to different sources of weather data and alternative, more flexible empirical specifications.

We use Jordà’s (2005) local projection method to trace out the impulse response function of real per capita GDP to a weather shock. This approach was advocated by Stock and Watson (2007), among others, as a flexible alternative that does not impose the dynamic restrictions embedded in vector autoregressions or autoregressive distributed lag specifications and is particularly suited to estimating nonlinearities in the dynamic response. We derive the impulse response by estimating a set of regressions:

\begin{matrix} y_{i, t + h} - y_{i, t - 1} = β_{1}^{h} c_{i, t} + β_{2}^{h} c_{i, t}^{2} + γ_{1}^{h} c_{i, t - 1} + γ_{2}^{h} c_{i, t}^{2} + Σ_{j = 1}^{h - 1} δ_{1}^{h} c_{i, t + h - j} + Σ_{j = 1}^{h - 1} δ_{2}^{h} c_{i, t + h - j}^{2} + φ_{1}^{h} {Δ y}_{i, t - 1} + μ_{i}^{h} + θ_{r, t}^{h} + ϵ_{i, t}^{h}, & (1) \end{matrix}

in which i indexes countries, t indexes years, and h indexes the estimation horizon (from horizon 0, which captures the contemporaneous effect, up to horizon 7, which captures the effect 7 years after the shock). Regressions for each horizon are estimated separately. The dependent variable is the cumulative growth of the outcome of interest between horizons t − 1 and t + h, measured as difference in the natural logarithms (y_i,t). Following Burke, Hsiang, and Miguel (2015a), the estimated regression has a quadratic specification in the weather variables, c_i,t, which comprise average annual temperature and precipitation. The regressions control for one lag of the dependent and weather variables and for forwards of the weather variables, as suggested by Teulings and Zubanov (2014). Country fixed effects $(μ_{i}^{h})$ control for all time-invariant country differences, such as latitude and average growth rates, while time fixed effects interacted with region dummies $(θ_{r, t}^{h})$ control for the common effect of all annual shocks across countries within a region.⁸ The analysis also explores an alternative fixed-effects structure proposed by Burke, Hsiang, and Miguel (2015a), which includes time fixed effects $(τ_{i}^{h})$ and country-specific linear and quadratic time trends $(θ_{i}^{h} t + θ_{i}^{h} t^{2})$ to account for within-country changes over time, such as demographic shifts, instead of the region-year fixed effects $(θ_{r, t}^{h})$ of the baseline specification. Standard errors are clustered at the country level.

To avoid bias associated with “bad controls” (or overcontrolling), the specification is purposefully parsimonious: many of the determinants of growth, typically included in standard growth regressions (for example, institutional quality, educational achievement, policies, and so forth), may themselves be shaped by weather shocks and are thus not part of the baseline estimation. Of course, to the extent that these are time-invariant and country-specific, they are subsumed in the country fixed effects.

Within this estimation framework, the effect of a weather shock, such as a 1°C increase in temperature, on the level of output at horizon h can be obtained by differentiating equation (1) with respect to temperature:

\begin{matrix} \frac{\partial (y_{i, t + h} - y_{i, t - 1})}{\partial T_{i, t}} = β_{1}^{h} + 2 β_{2}^{h} T_{i, t} . & (2) \end{matrix}

Evaluating equation (2) for each horizon separately and using the 2015 annual average temperature, T_i,2015, allows us to obtain the impulse response functions of per capita GDP to a temperature shock for each country. The marginal effect of an increase in precipitation is computed analogously. The threshold temperature at which the effect on the outcome variable switches from positive to negative, that is, the optimal level of temperature for per capita output, can be obtained by setting equation (2) to zero.

As discussed below, we use this empirical framework to examine the effect of weather shocks on per capita GDP, as well as on sectoral output (crop production, agricultural value added, services value added, and industrial value added), and the key elements of the aggregate production function (investment as a proxy for capital stock, infant mortality and the human development index as a proxy for labor supply). We rely on an extended version of this empirical strategy to capture productivity effects using more disaggregated data on industry outcomes.

Our primary data sources for the outcome variables are the IMF World Economic Outlook and the World Bank World Development Indicators databases, from which we construct our measures of per capita GDP, broad sectoral value added and index of agricultural production. We rely on the Groningen Growth and Development Centre 10-sector database for more disaggregated data on sectoral real value added and employment in 40 countries over the period 1950–2012 when analyzing productivity effects.

Historical temperature and precipitation are from the University of East Anglia’s Climate Research Unit (CRU). We construct average annual temperature and precipitation by aggregating weather data at the grid-cell level, provided by CRU at 0.5 × 0.5-degree resolution, to the level of the country using the 1950 population in each cell as weights. This method allows us to account for differences in population density within countries and captures the average weather experienced by a person in the country. Projections of temperature and precipitation are from the NASA Earth Exchange Global Daily Downscaled Projections dataset, which provides projections of the daily maximum and minimum temperature and total precipitation from 2005 and projections for 2015 – 2100 at the 0.25 x 0.25 degree resolution from the 21 models of the Coupled Model Intercomparison Project Phase 5 for the RCP 4.5 and RCP 8.5 scenario. For each of these scenarios, we compute annual temperatures by (i) averaging the maximum and minimum daily temperatures for each model, (ii) averaging across the 21 models, and (iii) averaging across all days of the year. We construct projections for annual precipitation in a similar fashion, but take the sum of the daily precipitation forecast. Annex 1 lists all data sources used in the paper.

IV. RESULTS

A. Short-term Effect on Per Capita Output

The results from estimating equation (1) are presented in Table 1. The main specification is given in column (5), while the other columns present results of robustness checks to using alternative sources of weather data; alternative population weights; and alternative samples, controls, and estimation approaches. Panel A contains the estimated coefficients for the weather variables at horizon 0 (that is, the contemporaneous effects of weather shocks). Panel B shows the effect of a 1°C increase in temperature estimated, following equation (2), at the median 2015 temperature for advanced economies (median T = 11°C), emerging market economies (median T = 22°C), and low-income countries (median T = 25°C) on impact and after seven years. Similarly, Panel C shows the effect of a 100 millimeter increase in precipitation estimated at the median 2015 precipitation for the three groups of economies on impact and after seven years.

Table 1.

Effect of Weather Shocks on Output

Source: IMF stall calculations.Note: The table presents results from estimating equation (1), with separate regressions for each horizon. Panel A reports the estimated coefficients on the weather variables for horizon 0, Panels B and C show the marginal impact of a change in temperature and precipitation computed as per equation (2) at the median temperature (T) and median precipitation (P) of advanced economies (AE), emerging markets (EM), and low-income developing countries (LIDC) contemporaneously (horizon 0) and cumulatively seven years after the shock. The specifications In columns (1)–(8) control for country fixed effects; lags and forwards of temperature, precipitation, and their squared terms; and lag of growth. Column (8) shows results from estimating an autoregressive distributed lag model with seven lags of the weather variables and their squared terms. Column (9) reports the coefficients on temperature and precipitation from a linear specification estimated on a sample of countries with average temperature above 15°C, also including controls for country fixed effects and lag of growth. In all specifications, standard errors are clustered at the country level. CRU = University of East Anglia, Climate Research Unit; mm = millimeter; UDEL = University of Delaware.

^{^*}p < 0.1;

^{^**}p < 0.05;

^{^***}p < 0.01.

Table 1.

Effect of Weather Shocks on Output

A. Real Output per Capita Growth	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Temperature	1.399^***	1.443^***	1.428^***	1.343^***	1.347^***	1.248^***	1.342^***	1.249^***	−1.154^***
Temperature	(0.359)	(0.367)	(0.366)	(0.355)	(0.357)	(0.339)	(0.355)	(0.380)	(0.320)
Temperature²	−0.049^***	−0.049^***	−0.048^***	−0.052^***	−0.051^***	−0.044^***	−0.051^***	−0.044^***
Temperature²	(0.012)	(0.011)	(0.011)	(0.011)	(0.011)	(0.010)	(0.011)	(0.011)
Precipitation	0.056	0.103^*	0.163^*	0.045	0.110	0.127	0.119	0.082	0.005
Precipitation	(0.097)	(0.061)	(0.085)	(0.058)	(0.104)	(0.103)	(0.104)	(0.112)	(0.034)
Precipitation²	−0.002	−0.002^**	−0.004^**	−0.001	−0.003	−0.003	−0.003	−0.002
Precipitation²	(0.002)	(0.001)	(0.002)	(0.001)	(0.002)	(0.002)	(0.002)	(0.002)
Any Disaster							0.406^**
Any Disaster							(0.180)
Threshold Temperature (°C)	14	15	15	13	13	14	13	14
Weather Source	UDFI	CRU	CRU	CRU	CRU	CRU	CRU	CRU	CPU
Population Weight	2010	2010	1950	2010	1950	1950	1950	1950	1950
Year Fixed Effects	Y	Y	Y	N	N	M	N	N	N
Region x Year Fixed Effects	N	N	N	Y	Y	Y	Y	Y	Y
Country Time Trends	Y	Y	Y	N	N	N	M	N	N
At Least 20 Years of Data	N	N	N	N	N	Y	N	M	N
Adjusted R²	0 15	0.15	0.15	0.14	0.14	0 14	0.14	0.11	0 09
Number of Countries	177	198	189	198	189	184	189	189	127
Number of Observations	8,147	9,114	8,815	9,114	8,815	8,756	8,815	8,917	6,135
B. Impact of a 1°C Increase in Temperature on Real Output per Capita Level at Horizon 0
AE(T=11°C)	0.331^*	0.370^*	0.365^*	0.197	0.218	0.280	0.217	0.277
AE(T=11°C)	(0.196)	(0.196)	(0.195)	(0.191)	(0.196)	(0.190)	(0.195)	(0.212)
EM (T=22°C)	−0.736^**	−0.703^***	−0.697^***	−0.949^***	−0.911^***	−0.687^***	−0.907^***	−0.695^***
EM (T=22°C)	(0.309)	(0.223)	(0.223)	(0.266)	(0.264)	(0.228)	(0.263)	(0.243)
LIDC (T=25°C)	−1.027^***	−0.996^***	−0.987^***	−1.261^***	−1.219^***	−0.951^***	−1.214^***	−0.960^***
LIDC (T=25°C)	(0.370)	(0.268)	(0.267)	(0.318)	(0.315)	(0.270)	(0.313)	(0.287)
Impact of a 1°C Increase in Temperature on Real Output per Capita Level at Horizon 7
AE(T=11°C)	0.898	0.889	0.822	0.457	0.558	0.560	0.552	0.023
AE(T=11°C)	(0.705)	(0.701)	(0.697)	(0.744)	(0.752)	(0.744)	(0.751)	(0.478)
EM (T=22°C)	−1.173	−0.957	−1.048	−1.117^*	−1.115^*	−1.088^*	−1.138^*	−0.547
EM (T=22°C)	(0.852)	(0.665)	(0.651)	(0.604)	(0.591)	(0.595)	(0.589)	(0.386)
LIDC (T=25°C)	−1.738^*	−1.461^*	−1.558^**	−1.547^**	−1.571^**	−1.537^**	−1.599^**	−0.702
LIDC (T=25°C)	(1.002)	(0.761)	(0.745)	(0.686)	(0.667)	(0.670)	(0.664)	(0.450)
C. Impact of a 100 mm per Year Increase in Precipitation on Real Output per Capita Level at Horizon 0
AE (P=800 mm per year)	0.018	0.066	0.101^*	0.028	0.066	0.076	0.073	0.050
AE (P=800 mm per year)	(0.067)	(0.046)	(0.059)	(0.046)	(0.071)	(0.070)	(0.071)	(0.077)
EM (P=900 mm per year)	0.013	0.061	0.093^*	0.026	0.060	0.070	0.067	0.046
EM (P=900 mm per year)	(0.063)	(0.045)	(0.056)	(0.045)	(0.067)	(0.066)	(0.067)	(0.072)
LIDC (P=1,100 mm per year)	0.004	0.052	0.078	0.022	0.049	0.057	0.056	0.038
LIDC (P=1,100 mm per year)	(0.057)	(0.041)	(0.050)	(0.042)	(0.059)	(0.058)	(0.059)	(0.064)
Impact of a 100 mm per Year Increase in Precipitation on Real Output per Capita Level at Horizon 7
AE (P=800 mm per year)	0.304	0.171	0.179	−0.173	−0.187	−0.207	−0.209	−0.287
AE (P=800 mm per year)	(0.198)	(0.216)	(0.227)	(0.214)	(0.223)	(0.225)	(0.224)	(0.229)
EM (P=900 mm per year)	0.295	0.166	0.174	−0.156	−0.166	−0.187	−0.188	−0.267
EM (P=900 mm per year)	(0.188)	(0.205)	(0.215)	(0.200)	(0.209)	(0.210)	(0.210)	(0.216)
LIDC (P=1,100 mm per year)	0.278	0.155	0.164	−0.121	−0.126	−0.148	−0.146	−0.227
LIDC (P=1,100 mm per year)	(0.169)	(0.185)	(0.192)	(0.174)	(0.182)	(0.182)	(0.183)	(0.191)

^{^*}p < 0.1;

^{^**}p < 0.05;

^{^***}p < 0.01.

Table 1.

Effect of Weather Shocks on Output

A. Real Output per Capita Growth	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Temperature	1.399^***	1.443^***	1.428^***	1.343^***	1.347^***	1.248^***	1.342^***	1.249^***	−1.154^***
Temperature	(0.359)	(0.367)	(0.366)	(0.355)	(0.357)	(0.339)	(0.355)	(0.380)	(0.320)
Temperature²	−0.049^***	−0.049^***	−0.048^***	−0.052^***	−0.051^***	−0.044^***	−0.051^***	−0.044^***
Temperature²	(0.012)	(0.011)	(0.011)	(0.011)	(0.011)	(0.010)	(0.011)	(0.011)
Precipitation	0.056	0.103^*	0.163^*	0.045	0.110	0.127	0.119	0.082	0.005
Precipitation	(0.097)	(0.061)	(0.085)	(0.058)	(0.104)	(0.103)	(0.104)	(0.112)	(0.034)
Precipitation²	−0.002	−0.002^**	−0.004^**	−0.001	−0.003	−0.003	−0.003	−0.002
Precipitation²	(0.002)	(0.001)	(0.002)	(0.001)	(0.002)	(0.002)	(0.002)	(0.002)
Any Disaster							0.406^**
Any Disaster							(0.180)
Threshold Temperature (°C)	14	15	15	13	13	14	13	14
Weather Source	UDFI	CRU	CRU	CRU	CRU	CRU	CRU	CRU	CPU
Population Weight	2010	2010	1950	2010	1950	1950	1950	1950	1950
Year Fixed Effects	Y	Y	Y	N	N	M	N	N	N
Region x Year Fixed Effects	N	N	N	Y	Y	Y	Y	Y	Y
Country Time Trends	Y	Y	Y	N	N	N	M	N	N
At Least 20 Years of Data	N	N	N	N	N	Y	N	M	N
Adjusted R²	0 15	0.15	0.15	0.14	0.14	0 14	0.14	0.11	0 09
Number of Countries	177	198	189	198	189	184	189	189	127
Number of Observations	8,147	9,114	8,815	9,114	8,815	8,756	8,815	8,917	6,135
B. Impact of a 1°C Increase in Temperature on Real Output per Capita Level at Horizon 0
AE(T=11°C)	0.331^*	0.370^*	0.365^*	0.197	0.218	0.280	0.217	0.277
AE(T=11°C)	(0.196)	(0.196)	(0.195)	(0.191)	(0.196)	(0.190)	(0.195)	(0.212)
EM (T=22°C)	−0.736^**	−0.703^***	−0.697^***	−0.949^***	−0.911^***	−0.687^***	−0.907^***	−0.695^***
EM (T=22°C)	(0.309)	(0.223)	(0.223)	(0.266)	(0.264)	(0.228)	(0.263)	(0.243)
LIDC (T=25°C)	−1.027^***	−0.996^***	−0.987^***	−1.261^***	−1.219^***	−0.951^***	−1.214^***	−0.960^***
LIDC (T=25°C)	(0.370)	(0.268)	(0.267)	(0.318)	(0.315)	(0.270)	(0.313)	(0.287)
Impact of a 1°C Increase in Temperature on Real Output per Capita Level at Horizon 7
AE(T=11°C)	0.898	0.889	0.822	0.457	0.558	0.560	0.552	0.023
AE(T=11°C)	(0.705)	(0.701)	(0.697)	(0.744)	(0.752)	(0.744)	(0.751)	(0.478)
EM (T=22°C)	−1.173	−0.957	−1.048	−1.117^*	−1.115^*	−1.088^*	−1.138^*	−0.547
EM (T=22°C)	(0.852)	(0.665)	(0.651)	(0.604)	(0.591)	(0.595)	(0.589)	(0.386)
LIDC (T=25°C)	−1.738^*	−1.461^*	−1.558^**	−1.547^**	−1.571^**	−1.537^**	−1.599^**	−0.702
LIDC (T=25°C)	(1.002)	(0.761)	(0.745)	(0.686)	(0.667)	(0.670)	(0.664)	(0.450)
C. Impact of a 100 mm per Year Increase in Precipitation on Real Output per Capita Level at Horizon 0
AE (P=800 mm per year)	0.018	0.066	0.101^*	0.028	0.066	0.076	0.073	0.050
AE (P=800 mm per year)	(0.067)	(0.046)	(0.059)	(0.046)	(0.071)	(0.070)	(0.071)	(0.077)
EM (P=900 mm per year)	0.013	0.061	0.093^*	0.026	0.060	0.070	0.067	0.046
EM (P=900 mm per year)	(0.063)	(0.045)	(0.056)	(0.045)	(0.067)	(0.066)	(0.067)	(0.072)
LIDC (P=1,100 mm per year)	0.004	0.052	0.078	0.022	0.049	0.057	0.056	0.038
LIDC (P=1,100 mm per year)	(0.057)	(0.041)	(0.050)	(0.042)	(0.059)	(0.058)	(0.059)	(0.064)
Impact of a 100 mm per Year Increase in Precipitation on Real Output per Capita Level at Horizon 7
AE (P=800 mm per year)	0.304	0.171	0.179	−0.173	−0.187	−0.207	−0.209	−0.287
AE (P=800 mm per year)	(0.198)	(0.216)	(0.227)	(0.214)	(0.223)	(0.225)	(0.224)	(0.229)
EM (P=900 mm per year)	0.295	0.166	0.174	−0.156	−0.166	−0.187	−0.188	−0.267
EM (P=900 mm per year)	(0.188)	(0.205)	(0.215)	(0.200)	(0.209)	(0.210)	(0.210)	(0.216)
LIDC (P=1,100 mm per year)	0.278	0.155	0.164	−0.121	−0.126	−0.148	−0.146	−0.227
LIDC (P=1,100 mm per year)	(0.169)	(0.185)	(0.192)	(0.174)	(0.182)	(0.182)	(0.183)	(0.191)

^{^*}p < 0.1;

^{^**}p < 0.05;

^{^***}p < 0.01.

Across all specifications, we find that the estimated coefficient on temperature is positive and the coefficient on temperature squared is negative, confirming the nonlinear relationship between growth and temperature shocks uncovered by Burke, Hsiang, and Miguel (2015a). At low temperatures, an increase in temperature boosts growth, whereas at high temperatures, an increase in temperature hurts growth, with the threshold average annual temperature estimated to be about 13°C –15°C. These results suggest highly uneven effects of warming across the globe, depending on the initial climate of a particular location.

Because most advanced economies are situated in colder locations, with annual average temperatures close to the 13°C –15°C threshold, a marginal temperature increase does not materially affect their contemporaneous growth.⁹ Conversely, emerging market economies and particularly low-income countries tend to have much hotter climates, and a rise in temperature significantly lowers per capita GDP.

Figure 5 illustrates these findings, by overlaying the marginal effect of a 1°C increase in temperature on contemporaneous per capita GDP by initial temperature, with the distribution of advanced, emerging market and low-income countries according to their average annual temperature. For the median emerging market economy, a 1°C increase from a temperature of 22°C lowers growth in the same year by 0.9 percentage point. For the median low-income country, with temperature of 25°C, the effect is even larger: growth falls by 1.2 percentage points. And even though countries projected to be significantly affected by an increase in temperature produced only about one-fifth of global GDP in 2016, they are home to close to 60 percent of current global population, as depicted in Figure 6, panel 2, which rescales countries in proportion to their population.

There is also significant variation in the impact of temperature increase within countries. Figure 6, panel 1, shows the estimated impact of a 1°C increase in temperature across all grids in the world, based on their average annual temperature in 2005. As indicated on the map, in a number of economies where the effect on aggregate growth may be statistically indistinguishable from zero, such as the United States, there are areas where a temperature rise would significantly lower output.

Our analysis reveals no consistently significant relationship between precipitation and per capita GDP growth across the various specifications as indicated in Panel C in Table 1. The lack of robust relationship could reflect potentially larger measurement error in the precipitation variable as discussed in Auffhammer et al. (2011). The measurement error may be further amplified by temporal aggregation. For example, if the only channel through which precipitation affects aggregate outcomes is through its effect on agriculture, then only precipitation during crops’ growing period—poorly proxied by annual precipitation—may be relevant.

B. Medium-term Effect on Per Capita Output

The empirical analysis suggests that the effects of temperature increase are long-lasting. Economic activity in countries with warmer climates does not recover quickly after a rise in temperature. The cumulative effects of a 1°C increase in temperature seven years after the shock is shown in the lower half of Panel B in Table 1. Even seven years after a weather shock, per capita output is 1 percent lower for the median emerging market economy and 1.5 percent lower for the median low-income country as depicted in Figure 7.

The estimated persistence may reflect the relatively persistent nature of temperature shocks. Univariate time series regression analysis shows that temperature shocks decay slowly, especially in relatively hot locations, with a 1°C degree increase in annual temperature leading to significantly higher temperatures in the subsequent years. A deepening in the shape of the estimated impulse response of output to a temperature shock hints at the possibility of a growth effect (and consequently much larger economic losses from higher temperatures). However, statistically, we are unable to reject the hypothesis that the contemporaneous and medium-term effects of a temperature shock on per capita output are identical.¹⁰

C. Robustness

To establish the robustness of these findings, we present results from estimating numerous alternative specifications in Table 1. In column (1), we replicate the specification of Burke, Hsiang, and Miguel (2015a), using our substantially larger sample (relative to their study, our paper expands the sample both geographically and temporally by about 25 percent). In this specification, we include country-specific linear and quadratic time trends, we use University of Delaware (UDEL) weather data, and aggregate weather variables using the 1990 population weights. Column (2) uses the same empirical approach as in column (1) but an alternative source of weather data, CRU instead of UDEL, and obtains similar coefficients on the temperature and precipitation variables.

The choice of population weights used to aggregate gridded weather data to the country level could play an important role since migration within and across country borders is one of the potential strategies for coping with adverse weather conditions. Since historical data show an increase in average annual temperatures starting in the 1970s (Figure 3), column (3) presents results with 1950 population weights to account for migration responses that could have already taken place.

Column (4) and column (5) (main specification for the paper) present results for the baseline specification with region-year fixed effects, following Dell, Jones, and Olken 2012, using 2010 and 1950 population weights, respectively. Column (6) limits the sample to countries with at least 20 years of data. Column (7) controls separately for the occurrence of natural disasters since temperature and precipitation fluctuations might affect economic activity through their effect on the incidence of natural disasters. Controlling for natural disasters does not materially alter the estimated coefficients on temperature and precipitation. In columns (1)–(7), impulse responses were estimated using Jordà’s (2005) local projection method. Column (8), however, tests the robustness of the findings to using the autodistributed lag model with seven lags of the weather variables and their squared terms, as in Dell, Jones, and Olken (2012), who tested different models from no lags up to 10 lags and found that across different lag specifications results are broadly consistent in magnitude and statistical significance.

V. Channels of Impact

The weather can influence economic activity through various channels. The most obvious one is agricultural output, given that temperature and precipitation are direct inputs in crop production. However, studies show evidence of broader impacts, including on labor productivity, mortality, health, and conflict.¹¹ The literature so far has often studied these effects within a specific country or through laboratory experiments. We examine whether these channels are also at work in a cross-country setting.

Sectoral Output: Agriculture, Manufacturing, Services

We begin by studying whether weather shocks influence only agricultural production or also affect other economic sectors, by estimating equation (1) but with real value added of agricultural, manufacturing, and services sectors, and crop production as our outcomes of interest. Regression results are presented in Table 2. Figure 8 depicts the impulse response function of the four outcomes considered at the temperatures prevailing in the median low-income country.

Table 2.

Effect of Weather Shocks on Sectoral Output

Source: IMF staff calculations.Note: The table presents results from estimating equation (1) using the same specification as in Table 1, column (5), for different dependent variables, with separate regressions estimated for each horizon. In all specifications, standard errors are clustered at the country level. Panel A reports the estimated coefficients on the weather variables for horizon 0. Panels B and C show the marginal impact of a change In temperature and precipitation computed as per equation (2) at the median temperature (T) and median precipitation (P) of advanced economies (AE), emerging markets (EM), and low-income developing countries (LIDC) contemporaneously (horizon 0) and cumulatively seven years after the shock, mm = millimeter.

^{^*}p < 0,1;

^{^**}p < 0.05;

^{^***}p < 0.01.

Table 2.

Effect of Weather Shocks on Sectoral Output

A. Dependent Variable	Agriculture (1)	Manufacturing (2)	Services (3)	Crop Production (4)
Temperature	0.283	1.281	−0.268	3.860^*
Temperature	(0.871)	(1.035)	(0.585)	(2.085)
Temperature²	−0.043^*	−0.051^*	−0.007	−0.151^***
Temperature²	(0.023)	(0.027)	(0.016)	(0.050)
Precipitation	0.705^***	0.108	−0.000	1.287^***
Precipitation	(0.228)	(0.149)	(0.111)	(0.332)
Precipitation²	−0.015^***	−0.002	−0.001	−0.028^***
Precipitation²	(0.005)	(0.003)	(0.002)	(0.007)
Adjusted R²	0.10	0.13	0.12	0.09
Number of Countries	174	168	174	185
Number of Observations	5,847	5,225	5,730	8,836
B. Impact of a 1°C Increase in Temperature on Dependent Variable Level at Horizon 0
AE (T=11°C)	−0.664	0.152	−0.423	0.547
AE (T=11°C)	(0.464)	(0.532)	(0.303)	(1.077)
EM (T=22°C)	−1.610^***	−0.977^**	−0.578^*	−2.767^***
EM (T=22°C)	(0.431)	(0.439)	(0.298)	(0.664)
LIDC (T=25°C)	−1.868^***	−1.285^**	−0.621^*	−3.671^***
LIDC (T=25°C)	(0.517)	(0.538)	(0.362)	(0.820)
Impact of a 1°C Increase in Temperature on Dependent Variable Level at Horizon 7
AE(T=11°C)	2.070^***	1.642	−0.220	1.177
AE(T=11°C)	(0.753)	(1.798)	(1.445)	(0.889)
EM (T=22°C)	−0,498	−0.926	0.054	−0.509
EM (T=22°C)	(0.654)	(0.939)	(0.734)	(0.812)
LIDC (T=25°C)	−1.198	−1.626	0.129	−0.969
LIDC (T=25°C)	(0.769)	(1.117)	(0.910)	(0.985)
C. Impact of a 100 mm per Year Increase in Precipitation on Dependent Variable Level at Horizon 0
AE (P=800 mm per year)	0.458^***	0.076	−0.013	0.835^***
AE (P=800 mm per year)	(0.149)	(0.105)	(0.075)	(0.223)
EM (P=900 mm per year)	0.428^***	0.072	−0.015	0.778^***
EM (P=900 mm per year)	(0.139)	(0.100)	(0.071)	(0.210)
LIDC (P=1,100 mm per year)	0.366^***	0.065	−0.018	0.665^***
LIDC (P=1,100 mm per year)	(0.121)	(0.090)	(0.063)	(0.185)
Impact of a 100 mm per Year Increase in Precipitation on Dependent Variable Level at Horizon 7
AE (P=800 mm per year)	−0.228	0.024	−0.141	−0.237
AE (P=800 mm per year)	(0.257)	(0.390)	(0.286)	(0.284)
EM (P=900 mm per year)	−0.213	0.030	−0.125	−0.217
EM (P=900 mm per year)	(0.243)	(0.371)	(0.269)	(0.267)
LIDC (P=1,100 mm per year)	−0.184	0.041	−0.094	−0.177
LIDC (P=1,100 mm per year)	(0.217)	(0.332)	(0.235)	(0.235)

^{^*}p < 0,1;

^{^**}p < 0.05;

^{^***}p < 0.01.

Table 2.

Effect of Weather Shocks on Sectoral Output

A. Dependent Variable	Agriculture (1)	Manufacturing (2)	Services (3)	Crop Production (4)
Temperature	0.283	1.281	−0.268	3.860^*
Temperature	(0.871)	(1.035)	(0.585)	(2.085)
Temperature²	−0.043^*	−0.051^*	−0.007	−0.151^***
Temperature²	(0.023)	(0.027)	(0.016)	(0.050)
Precipitation	0.705^***	0.108	−0.000	1.287^***
Precipitation	(0.228)	(0.149)	(0.111)	(0.332)
Precipitation²	−0.015^***	−0.002	−0.001	−0.028^***
Precipitation²	(0.005)	(0.003)	(0.002)	(0.007)
Adjusted R²	0.10	0.13	0.12	0.09
Number of Countries	174	168	174	185
Number of Observations	5,847	5,225	5,730	8,836
B. Impact of a 1°C Increase in Temperature on Dependent Variable Level at Horizon 0
AE (T=11°C)	−0.664	0.152	−0.423	0.547
AE (T=11°C)	(0.464)	(0.532)	(0.303)	(1.077)
EM (T=22°C)	−1.610^***	−0.977^**	−0.578^*	−2.767^***
EM (T=22°C)	(0.431)	(0.439)	(0.298)	(0.664)
LIDC (T=25°C)	−1.868^***	−1.285^**	−0.621^*	−3.671^***
LIDC (T=25°C)	(0.517)	(0.538)	(0.362)	(0.820)
Impact of a 1°C Increase in Temperature on Dependent Variable Level at Horizon 7
AE(T=11°C)	2.070^***	1.642	−0.220	1.177
AE(T=11°C)	(0.753)	(1.798)	(1.445)	(0.889)
EM (T=22°C)	−0,498	−0.926	0.054	−0.509
EM (T=22°C)	(0.654)	(0.939)	(0.734)	(0.812)
LIDC (T=25°C)	−1.198	−1.626	0.129	−0.969
LIDC (T=25°C)	(0.769)	(1.117)	(0.910)	(0.985)
C. Impact of a 100 mm per Year Increase in Precipitation on Dependent Variable Level at Horizon 0
AE (P=800 mm per year)	0.458^***	0.076	−0.013	0.835^***
AE (P=800 mm per year)	(0.149)	(0.105)	(0.075)	(0.223)
EM (P=900 mm per year)	0.428^***	0.072	−0.015	0.778^***
EM (P=900 mm per year)	(0.139)	(0.100)	(0.071)	(0.210)
LIDC (P=1,100 mm per year)	0.366^***	0.065	−0.018	0.665^***
LIDC (P=1,100 mm per year)	(0.121)	(0.090)	(0.063)	(0.185)
Impact of a 100 mm per Year Increase in Precipitation on Dependent Variable Level at Horizon 7
AE (P=800 mm per year)	−0.228	0.024	−0.141	−0.237
AE (P=800 mm per year)	(0.257)	(0.390)	(0.286)	(0.284)
EM (P=900 mm per year)	−0.213	0.030	−0.125	−0.217
EM (P=900 mm per year)	(0.243)	(0.371)	(0.269)	(0.267)
LIDC (P=1,100 mm per year)	−0.184	0.041	−0.094	−0.177
LIDC (P=1,100 mm per year)	(0.217)	(0.332)	(0.235)	(0.235)

^{^*}p < 0,1;

^{^**}p < 0.05;

^{^***}p < 0.01.

We find that agricultural value added and crop production drop with higher temperature, recover somewhat in subsequent years, but generally remain depressed over the medium term—much as expected and as documented in a large body of work.¹² However, the analysis also confirms findings that manufacturing output is similarly hurt as temperatures rise in countries with hot climates, although the estimates are more imprecise (see also Dell, Jones, and Olken 2012; Burke, Hsiang, and Miguel 2015a). Only services sector output appears to be sheltered from the weather.

It is important to note that, unlike aggregate output, agricultural production is significantly affected by precipitation in addition to temperature shocks. Although the results suggest a concave relationship between agricultural output and precipitation, at the typical levels of precipitation of all three country groups, an increase in precipitation unambiguously improves agricultural productivity (Table 2, Panel C). The effects of precipitation are also short-lived; agricultural output seven years down the line is not affected by a precipitation shock today, which is different from the effect of temperature.

To shed light on the reasons why weather shocks affect sectors besides agriculture in such a broad and long-lasting manner, we examine how key elements of the aggregate production function—productivity, and labor and capital inputs—respond to weather shocks. As in other studies, we aim to capture the net reduced-form effects of weather on various outcomes rather than uncover the potentially complex structural relationships that may exist between these variables.

Productivity

Evidence from surveys and other sources shows that exposure to heat above a certain point reduces people’s performance on both cognitive and physical tasks.¹³ We therefore examine whether higher temperatures in parts of the world that are hot decrease labor productivity. If productivity is a channel through which weather shocks affect aggregate GDP, the effect should be significantly larger for sectors in which workers are directly exposed to the weather. We explore this hypothesis using the Groningen Growth and Development Centre 10-sector database, which provides sectoral real value added and employment in 40 countries over the period 1950–2012. We follow Graff Zivin and Neidell (2014) and classify sectors into those that are “heat-exposed” and not¹⁴ to estimate the following specification:

\begin{matrix} y_{i, s, t + h} - y_{i, s, t - 1} = β_{1}^{h} c_{i, t} + β_{2}^{h} c_{i, t}^{2} + γ_{1}^{h} c_{i, t - 1} + γ_{2}^{h} c_{i, t}^{2} + Σ_{j = 1}^{h - 1} δ_{1}^{h} c_{i, t + h - j} + Σ_{j = 1}^{h - 1} δ_{2}^{h} c_{i, t + h - j}^{2} + α_{1}^{h} c_{i, t} \times H_{s} + α_{2}^{h} c_{i, t}^{2} \times H_{s} + ω_{1}^{h} c_{i, t - 1} \times H_{s} + ω_{2}^{h} c_{i, t - 1}^{2} \times H_{s} + Σ_{j = 1}^{h - 1} π_{2}^{h} c_{i, t + h - j} \times H_{s} + Σ_{j = 1}^{h - 1} τ_{2}^{h} c_{i, t + h - j}^{2} \times H_{s} + φ_{1}^{h} Δ y_{i, s, t - 1} + μ_{i, s}^{h} + θ_{r, t}^{h} + ϵ_{i, s, t}^{h}, & (3) \end{matrix}

in which y_i,s,t is the log of real sectoral value added per worker, H_s is an indicator for sectors that are “heat-exposed,” $μ_{i, s}^{h}$ are country-sector fixed effects, and $θ_{r, t}^{h}$ are region-year fixed effects. Standard errors are clustered at the country level.

Specification (5) in Table 3 summarizes the results of this estimation. Panel A contains the estimated coefficients on temperature and precipitation and their squared terms for non-heat-exposed sectors $(β_{1}^{h}, β_{2}^{h})$ and heat-exposed sectors $(β_{1}^{h} + α_{1}^{h}, β_{2}^{h} + α_{2}^{h})$ . Panels B and C present the estimated impacts of temperature and precipitation evaluated for the median advanced, emerging market and low-income country in the year of the shock, as well as seven years later. Panels 1 and 2 in Figure 9 plot the impulse response function of real output per worker to a 1°C increase in temperature for heat-exposed and non-heat-exposed sectors evaluated at the average annual temperature of the median low-income country.

Table 3.

Effect of Weather Shocks on Productivity, Capital, and Labor

Source: IMF staff calculations.Note: Columns (1–4) present results from estimating equation (1) using the same specification as in Table 1, column (5), for different dependent variables. Specification in column (5) presents results from estimating equation (3) where an indicator for heat exposed sectors is interacted with temperature and precipitation, their squared terms, and their lags and forwards; also controlling for country-sector and region-year fixed effects, and lag of growth. Separate regressions are estimated for each horizon. In all specifications, standard errors are clustered at the country level. Panel A reports the estimated coefficients on the weather variables for horizon 0. Panels B and C show the marginal impact of a change In temperature and precipitation computed as per equation (2) at the median temperature (T) and median precipitation (P) of advanced economies (AE), emerging markets (EM), and low-income developing countries (LIDC), contemporaneously (horizon 0) and cumulatively seven years alter the shock. HDI = Human Development Index; mm = millimeter.

^{^*}p < 0.1;

^{^**}p < 0.05;

^{^***}p < 0.01.

Table 3.

Effect of Weather Shocks on Productivity, Capital, and Labor

A. Dependent Variable	Capital Input		Labor Input		Labor Productivity
	Investment	Imports	Infant Mortality	HDI	Non-Heat Exposed	Heat Exposed
	(1)	(2)	(3)	(4)	(5)
Temperature	0.850	0.467	−0.147	0.269^***	0.246	1.902^*
Temperature	(2.042)	(0.943)	(0.117)	(0.078)	(0.681)	(1.002)
Temperature²	−0.045	−0.068^**	0.005^*	−0.008^***	−0.010	−0.087^***
Temperature²	(0.059)	(0.033)	(0.003)	(0.002)	(0.018)	(0.026)
Precipitation	−0.377	−0.654^**	−0.001	0.000	0.047	0.272
Precipitation	(0.398)	(0.271)	(0.024)	(0.018)	(0,201)	(0.195)
Precipitation²	0.003	0.006	0.001	−0.000	−0.003	−0.008^*
Precipitation²	(0.009)	(0.007)	(0.001)	(0.000)	(0.005)	(0.004)
Adjusted R²	0.03	0.08	0.64	0.31	0.03
Number of Countries	169	1/8	182	181	40
Number of Observations	6,093	6,866	8,685	3,864	17,848
B. Impact of a 1°C Increase in Temperature on Dependent Variable Level at Horizon 0
AE(T=11°C)	−0.138	−1.029^**	−0.028	0.094^**	0.030	−0.003
AE(T=11°C)	(0,976)	(0.455)	(0.067)	(0.043)	(0,396)	(0,502)
EM (T=22°C)	−1.126	−2.525^***	0.092^*	−0.082	−0.185	−1.909^***
EM (T=22°C)	(1.064)	(0.753)	(0.055)	(0,056)	(0.412)	(0,363)
LIDC (T=25°C)	−1.395	−2.934^***	0.124^*	−0.129^*	−0.244	−2.428^***
LIDC (T=25°C)	(1.331)	(0.919)	(0.063)	(0.067)	(0.478)	(0.456)
Impact of a 1°C Increase in Temperature on Dependent Variable Level at Horizon 7
AE (T=11°C)	1.812	2.361	−0.364	0.609^**	0.3O5	−1.142
AE (T=11°C)	(2.029)	(1.494)	(0.427)	(0,259)	(1,183)	(0.986)
EM (T=22°C)	−4.225^**	−2.439^*	0.569	−0.237	−0.063	−1.642
EM (T=22°C)	(1.803)	(1.303)	(0.375)	(0.175)	(1.114)	(1.119)
LIDC (T=25°C)	−5.871^**	−3.747^**	0.824^*	−0.467^**	−0.163	−1.778
LIDC (T=25°C)	(2.074)	(1.516)	(0.426)	(0.195)	(1.306)	(1.365)
C. Impact of a 100 mm per Year Increase in Precipitation on Dependent Variable Level at Horizon 0
AE (P=800 mm per year)	−0.329	−0.558^***	0.008	0.007	−0.009	0.14B
AE (P=800 mm per year)	(0.262)	(0.180)	(0.015)	(0,013)	(0.133)	(0.136)
EM (P=900 mm per year)	−0.323	−0.547^***	0.009	−0.008	−0.016	0.132
EM (P=900 mm per year)	(0.246)	(0.170)	(0.015)	(0.012)	(0.125)	(0.130)
LIDC (P=1,100 mm per year)	−0.311	−0.523^***	0.011	−0.010	−0.030	0.101
LIDC (P=1,100 mm per year)	(0.216)	(0.151)	(0.013)	(0.011)	(0.109)	(0.118)
Impact of a 100 mm per Year Increase in Precipitation on Dependent Variable Level at Horizon 7
AE (P=800 mm per year)	−0.478	−0.984^**	0.071	−0.102^*	−0.295	0.072
AE (P=800 mm per year)	(0.689)	(0.498)	(0.163)	(0.061)	(0.832)	(0.554)
EM (P=900 mm per year)	−0.423	−0.961^**	0.074	−0.097^*	−0.265	0.041
EM (P=900 mm per year)	(0.649)	(0.472)	(0.149)	(0.057)	(0.776)	(0.524)
LIDC (P=1,100 mm per year)	−0.313	−0.914^**	0.080	−0.087^*	−0.206	−0.022
LIDC (P=1,100 mm per year)	(0.573)	(0.4221	(0.123)	(0.050)	(0.666)	(0.467)

^{^*}p < 0.1;

^{^**}p < 0.05;

^{^***}p < 0.01.

Table 3.

Effect of Weather Shocks on Productivity, Capital, and Labor

A. Dependent Variable	Capital Input		Labor Input		Labor Productivity
	Investment	Imports	Infant Mortality	HDI	Non-Heat Exposed	Heat Exposed
	(1)	(2)	(3)	(4)	(5)
Temperature	0.850	0.467	−0.147	0.269^***	0.246	1.902^*
Temperature	(2.042)	(0.943)	(0.117)	(0.078)	(0.681)	(1.002)
Temperature²	−0.045	−0.068^**	0.005^*	−0.008^***	−0.010	−0.087^***
Temperature²	(0.059)	(0.033)	(0.003)	(0.002)	(0.018)	(0.026)
Precipitation	−0.377	−0.654^**	−0.001	0.000	0.047	0.272
Precipitation	(0.398)	(0.271)	(0.024)	(0.018)	(0,201)	(0.195)
Precipitation²	0.003	0.006	0.001	−0.000	−0.003	−0.008^*
Precipitation²	(0.009)	(0.007)	(0.001)	(0.000)	(0.005)	(0.004)
Adjusted R²	0.03	0.08	0.64	0.31	0.03
Number of Countries	169	1/8	182	181	40
Number of Observations	6,093	6,866	8,685	3,864	17,848
B. Impact of a 1°C Increase in Temperature on Dependent Variable Level at Horizon 0
AE(T=11°C)	−0.138	−1.029^**	−0.028	0.094^**	0.030	−0.003
AE(T=11°C)	(0,976)	(0.455)	(0.067)	(0.043)	(0,396)	(0,502)
EM (T=22°C)	−1.126	−2.525^***	0.092^*	−0.082	−0.185	−1.909^***
EM (T=22°C)	(1.064)	(0.753)	(0.055)	(0,056)	(0.412)	(0,363)
LIDC (T=25°C)	−1.395	−2.934^***	0.124^*	−0.129^*	−0.244	−2.428^***
LIDC (T=25°C)	(1.331)	(0.919)	(0.063)	(0.067)	(0.478)	(0.456)
Impact of a 1°C Increase in Temperature on Dependent Variable Level at Horizon 7
AE (T=11°C)	1.812	2.361	−0.364	0.609^**	0.3O5	−1.142
AE (T=11°C)	(2.029)	(1.494)	(0.427)	(0,259)	(1,183)	(0.986)
EM (T=22°C)	−4.225^**	−2.439^*	0.569	−0.237	−0.063	−1.642
EM (T=22°C)	(1.803)	(1.303)	(0.375)	(0.175)	(1.114)	(1.119)
LIDC (T=25°C)	−5.871^**	−3.747^**	0.824^*	−0.467^**	−0.163	−1.778
LIDC (T=25°C)	(2.074)	(1.516)	(0.426)	(0.195)	(1.306)	(1.365)
C. Impact of a 100 mm per Year Increase in Precipitation on Dependent Variable Level at Horizon 0
AE (P=800 mm per year)	−0.329	−0.558^***	0.008	0.007	−0.009	0.14B
AE (P=800 mm per year)	(0.262)	(0.180)	(0.015)	(0,013)	(0.133)	(0.136)
EM (P=900 mm per year)	−0.323	−0.547^***	0.009	−0.008	−0.016	0.132
EM (P=900 mm per year)	(0.246)	(0.170)	(0.015)	(0.012)	(0.125)	(0.130)
LIDC (P=1,100 mm per year)	−0.311	−0.523^***	0.011	−0.010	−0.030	0.101
LIDC (P=1,100 mm per year)	(0.216)	(0.151)	(0.013)	(0.011)	(0.109)	(0.118)
Impact of a 100 mm per Year Increase in Precipitation on Dependent Variable Level at Horizon 7
AE (P=800 mm per year)	−0.478	−0.984^**	0.071	−0.102^*	−0.295	0.072
AE (P=800 mm per year)	(0.689)	(0.498)	(0.163)	(0.061)	(0.832)	(0.554)
EM (P=900 mm per year)	−0.423	−0.961^**	0.074	−0.097^*	−0.265	0.041
EM (P=900 mm per year)	(0.649)	(0.472)	(0.149)	(0.057)	(0.776)	(0.524)
LIDC (P=1,100 mm per year)	−0.313	−0.914^**	0.080	−0.087^*	−0.206	−0.022
LIDC (P=1,100 mm per year)	(0.573)	(0.4221	(0.123)	(0.050)	(0.666)	(0.467)

^{^*}p < 0.1;

^{^**}p < 0.05;

^{^***}p < 0.01.

Our analysis suggests that at higher temperatures, an increase in temperature significantly lowers labor productivity in heat-exposed industries. Temperature increases, however, have no discernible effect on the productivity of workers in non-heat-exposed sectors, even in countries with hot climates. Of course, productivity may decline if weather shocks provoke political instability, incite conflict, or undermine governing institutions in other ways, as discussed in numerous studies that document a strong link between weather shocks and these outcomes.¹⁵

Capital Accumulation

Temperature increases are largely supply-side shocks, but they could lead to persistent output losses and affect growth if they influence the rate of factor accumulation. Investment may fall in response to temperature shocks because there are fewer resources to invest, because the rate of return on capital is lower and/or because the temporary negative shock to income raises the cost of financing investment in an environment of imperfect capital markets (see, for example, Fankhauser and Tol 2005). When access to formal savings, credit, or insurance is limited, households may also sell productive assets to smooth consumption in response to weather shocks.

Using national accounts data, we examine the response of the main components of aggregate demand—gross capital formation, consumption, exports, and imports—to weather shocks using the empirical framework given by equation (1). At the temperature of the median low-income country, all four components respond negatively to a 1°C increase in temperature, although the uncertainty surrounding the estimated contemporaneous effects is large. However, in the medium term, the effect is most pronounced for investment (Table 3, column 1). Seven years after the shock, investment is estimated to be 6 percent lower than it would have been in the absence of the shock as depicted in Figure 9, panel 3. Imports, which are typically closely tied to investment, also exhibit a significant and long-lasting drop as temperature rises (Table 3, column 2 and Figure 9, panel 4).

The negative effect of temperature shocks on aggregate investment is consistent with evidence from household-level studies, which find that weather shocks could slow or even reverse capital accumulation as households try to smooth consumption or perceive investment as too risky (Hallegatte et al. 2016).

Labor Supply

Finally, we examine whether labor supply may be affected by weather fluctuations, for example, through their effect on health. In the absence of comprehensive and comparable data on adult health outcomes, we focus on infant mortality as an imperfect proxy. Estimating equation (1) with infant mortality as dependent variable reveals that, in hot climates, higher temperatures may reduce (future) labor supply because of their influence on mortality rates (Table 3, column 3 and Figure 9, panel 5). A 1°C increase in temperature raises infant mortality by 0.12 percentage point in the year of the shock. The effect grows through the estimation period as weather-related lower income (and potential food insecurity) reinforces the direct physiological impact of higher temperatures in hot climates.

This cross-country panel evidence corroborates findings in numerous studies of links between weather and mortality, prenatal health, and other health outcomes in different countries. Deschênes (2012) and Guo et al. (2014) provide comprehensive reviews of the literature on the link between temperature and mortality and health from a large number of countries.¹⁶ Graff Zivin and Neidell (2014), Deryugina and Hsiang (2014), Park (2017), and Somanathan et al. (2017) find a direct effect of higher temperature on labor supply and productivity. The adverse effects on the health and educational attainment of children could be one of the key reasons behind the long-lasting nature of weather’s consequences. Indeed, we find that higher temperatures also have a negative effect on a broader measure of human well-being, the Human Development Index, a weighted average of per capita income, educational achievement, and life expectancy as documented in Table 3, column 4 and Figure 9, panel 6.

VI. Long-Term Effects of Temperature Increase–A Model-Based Approach

Our empirical work assessed the macroeconomic effects of weather shocks in the short and medium term. We incorporate these estimates into a dynamic general equilibrium model to shed light on the potential long-term effects of temperature increases on GDP, investment, and public debt for a representative small open low-income economy. The model also highlights the role that structural transformation of low-income countries (that is, the transition from agriculture to a more services-based economy) could play in attenuating the impact of climate change.

Our simulations are based on the Debt, Investment and Growth (DIG) model of Buffie et al. (2012), which captures various aspects pertinent to low-income countries—such as low public investment efficiency and high capital adjustment costs—and can be extended easily to incorporate the structural transformation process.¹⁷ These aspects of the DIG model make it preferable for studying the impact of climate change in low-income countries relative to the IAMs more commonly used to assess climate change effects.

In the DIG model, firms combine labor, private capital, and infrastructure to produce output. Consumers supply labor and derive utility from consuming traded and nontraded goods, while the government collects revenue, redistributes income, and invests in infrastructure, which it funds through domestic and external borrowing, grants, and remittances. Based on the empirical results, we model changes in the exogenously given sector-specific total factor productivity levels as quadratic functions of temperature, while we calibrate all other parameters broadly as in Buffie et al. (2012). In particular, $T F P_{t + 1} - T F P_{t} = β_{1}^{1} (T_{t + 1} - T_{t}) + {2 β}_{2}^{1} (T_{t + 1} - T_{t}) T_{t} + Δ T E P_{t}^{*}$ , in which $Δ T E P_{t}^{*}$ is the total factor productivity growth rate that would prevail under no climate change, assumed to be 2.8 percent based on the World Economic Outlook medium-term growth forecast for low-income countries. β_1,1 and β_1,2 are the estimated coefficients on the linear and squared temperature terms in equation (1), as reported in column (5) of Table 1, rescaled to match the modeled decline of GDP when temperature increases by 1°C, and T_t is the average annual temperature for the median low-income country at time t, where the initial temperature is set at 25°C.

We examine the effects of climate change by simulating the macroeconomic response of output, the public-debt-to-GDP ratio, and private investment to the temperature increases projected under two of the scenarios prepared by the IPCC, as discussed in Section II. The simulations suggest that under both scenarios, the representative low-income country will experience sizable economic losses relative to a baseline of no changes in temperature, with significant downside risks (Figure 10).

In the milder scenario, the increase in temperature will lower output by 4 percent by 2100 and depress private investment by 5 percent as firms respond to lower productivity from rising temperatures by cutting back capital spending. The relative decline in output implies an increase in the public-debt-to-GDP ratio of 2 percentage points by 2100. In the unmitigated climate change scenario, the macroeconomic effect would be much larger. Output would fall by close to 9 percent relative to no climate change, private investment would fall by 11 percent, and the public-debt-to-GDP ratio would rise by 5 percentage points by 2100. These results are broadly in line with other model-based estimates of the impact of climate change as discussed in Tol (2009).¹⁸

There is great uncertainty surrounding these central projections because empirical estimates of the effect of temperature shocks are imprecise and temperature projections are uncertain. We combine these sources of uncertainty as follows. The upper-bound scenario is simulated assuming that the temperature increase is equal to the lowest 5th percentile for each RCP.¹⁹ To account for the uncertainty of estimated parameters, the TFP parameters are set to the conditional expected value for the upper 50 percent of the TFP distribution. The worst lower-bound scenario is simulated analogously.

As a result of these uncertainties, wide confidence intervals surround our central projections. There is a 2.5 percent chance of output declining more than 8 percent below the trend in the milder scenario and more than 16 percent in the unmitigated climate change scenario. In line with lower output, public debt would increase significantly relative to output (about 10 percent of GDP in the worst-case scenario), and the private-investment-to-GDP ratio could plummet by as much as 20 percent below the trend.

An alternative way to quantify climate change damage for a representative low-income country is to compute the present value of the shortfall in economic output relative to the baseline of no climate change and to express this present value as a share of current output.²⁰ Using a moderate growth-adjusted discount rate of 1.4 percent, the present value of output losses is large, at 48 percent and 100 percent of current output under the RCP 4.5 and RCP 8.5 scenarios, respectively.

The above simulations assume a static economic structure. However, as discussed in Section V, rising temperatures affect some economic sectors more than others. For example, compared with agriculture, the services sector is relatively sheltered from the adverse effects of higher temperature. Hence, structural economic transformation from a mostly agrarian to a more services-based economy could lower the economic cost of climate change. We extend the baseline DIG model to include an exogenous process of reallocating labor from agriculture and manufacturing to services. The pace of structural transformation is assumed to be moderate and replicates past trends for low-income countries: in the absence of shocks, the employment share of the services sector rises by 2.5 percentage points a decade. Simulations in this extended model indicate that over the long term, for the median low-income country, structural transformation can reduce the cost of climate change by about 25 and 30 percent in the RCP 4.5 and RCP 8.5 scenarios, respectively.

The potential impact of climate change quantified in this section is subject to important caveats. First, extrapolating from the short- to medium-term causal effects of weather shocks estimated from historical data to the long-term impact of potential global warming may overstate the case if persistent changes in climate induce agents to adapt their economic activity to the new environment. Conversely, permanent changes in climate may have consequences that fluctuations in annual weather do not. Moreover, the model does not capture the effects of extreme weather events, which inflict long-lasting macroeconomic damage, and could increase in frequency, potentially amplifying the damage they cause. Certain expected or possible events (such as rising sea levels) have no historic precedents from which to draw inference but may have very significant economic consequences for many low-income countries, which are also not quantified in our simulations. Moreover, the long-term projections do not incorporate several of the channels through which temperature increases, and climate change in general, could affect economic activity, such as declining labor supply from higher mortality and migration.

Even abstracting from these difficulties, considerable uncertainty exists about how to incorporate the empirical estimates of economic losses into the dynamic general equilibrium model. We have taken a very conservative approach and assume that weather shocks have a permanent effect on the level of output. However, several studies have argued that the empirical evidence is not inconsistent with a persistent effect on the growth rate of output (Dell, Jones, and Olken 2012; Burke, Hsiang, and Miguel 2015a). Permanent growth effects could arise if weather shocks scar productivity growth through their effects on institutions, innovation, or human capital accumulation. Since even a small growth effect would ultimately dwarf a level effect, the adverse consequence of temperature increases for the median low-income country would be many times larger if rising temperatures were incorporated into the model as affecting the growth path of output, as done by Burke, Hsiang and Miguel (2015a).

VII. Summary and Policy Implications

Coping with climate change is one of the fundamental challenges of the 21^st century, and this challenge looms particularly large for low-income economies. We document the extraordinarily fast rise in temperature over the past century across advanced, emerging market, and low-income economies. Low-income countries, which tend to be in some of the hottest parts of the planet and are projected to experience sizable increases in temperature depending on our ability to contain future GHG emissions, have contributed very little to the atmospheric concentration of GHGs.

The analysis suggests that rising temperatures have highly uneven macroeconomic effects, with the adverse consequences borne disproportionately by countries with hot climates, such as most low-income countries. We find that a rise in temperature lowers per capita output in countries with high average temperatures, in both the short and medium term, through a wide array of channels. In areas with hot climates, higher temperatures reduce agricultural output, lower productivity of workers exposed to the heat, slow the rate of capital accumulation, and damage health. These findings reflect impacts of weather shocks on average country outcomes. But weather shocks could also have sizable unfavorable distributional consequences within a country. Poor households tend to be more vulnerable to weather fluctuations as a result of their heavy reliance on agricultural income, higher proportion of income devoted to food items, and limited access to savings and credit (Hallegatte et al. 2016; Hallegatte and Rozenberg 2017; IMF 2016b).

The negative effects of projected climate change for low-income countries could be large. Focusing on one particular aspect of climate change—namely, projected rise in temperature—and under the conservative assumption that temperature increases affect the level, rather than growth path, of output, model simulations suggest that absent efforts to reduce global emissions, the output of a representative low-income country could be 9 percent lower than without an increase in temperature. Moreover, the negative welfare consequences of changing climate conditions will likely exceed output losses. Uncomfortably high temperatures may spur investment as households adapt, but the increase in economic activity may not be welfare-enhancing. The significant uncertainty about the magnitude and effects of climate change—not only how much temperatures will rise but also how the environment will react—calls for careful consideration of the sizable downside risks that surround our projections.

While our analysis emphasized the impact of global warming on low-income countries, it is important to note that all countries will increasingly feel direct negative effects from unmitigated climate change, through more frequent (and damaging) natural disasters, rising sea level, loss of biodiversity and many other difficult-to-quantify consequences. Warming will also begin to weigh on growth in many advanced economies, as their temperatures rise above optimal levels. By combining the estimated sensitivity of per capita output to temperature increase, baseline annual temperatures and projected temperature increase under the two RCP scenarios, Figure 11 depicts the potential cumulative impacts on 2100 per capita GDP across the globe. This exercise confirms the highly uneven effects of warming across the glove, but also reveals that the projected increase in temperature, especially under the RCP 8.5 scenario, will push many advanced economies beyond the threshold temperature levels, thus triggering direct economic losses for these countries as well. And even in countries where the effect might be moderate or positive on average, climate change will create winners and losers at both the individual and sectoral levels. Moreover, the international spillovers from the most vulnerable countries, through depressed economic activity and potentially higher conflict and migration flows, could be considerable. Going forward, only a global effort to contain carbon emissions to levels consistent with an acceptable increase in temperature can limit the long-term risks of climate change (Farid et al. 2016; Hallegatte et al. 2016; IMF 2015; Stern 2015; IPCC 2014).

Annex 1. Data Sources and Country Groupings

Annex Table 1.1.

Dala Sources

Source: IMF staff compilation.

Annex Table 1.1.

Dala Sources

Indicator	Source
Temperature, Historical	Intergovernmental Panel on Climate Change (IPCC) Coupled Model Intercomparison Project Phase Five AR5 Atlas subset; Marcott and others (2013); Matsuura and Willmott (2007); National Aeronautics and Space Administration (NASA) Goddard Institute for Space Studies (GISS); Royal Netherlands Meteorological Institute (KNMI) Climate Change Atlas; Shakun and others (2012)
Temperature and Precipitation, Forecast (Grid level)	NASA Earth Exchange Global Daily Downscaled Projections data set (NEX-GDDP)
Temperature and Precipitation, Historical (Grid level)	University of East Anglia, Climate Research Unit (CRU TS v.3.24); University of Delaware (UDEL v.4.01)
Population 2010, 1990, 1950 (Grid level)	Center for International Earth Science Information Network (CIESIN v.3 and v.4); History Database of the Global Environment (HYDE v3.2); Klein and others (2016)
Population 2015 and Projected Population 2100	United Nations World Population Prospects database, 2015 Revision
CO₂ Emissions	Carbon Dioxide Information Analysis Center
Temperature Forcings	Carbon Dioxide Information Analysis Center: NASA GISS; Roston and Migliozzi (2015)
Real GDP per Capita	IMF, World Economic Outlook database; World Bank, World Development Indicators database
Crop Production Index	Food and Agriculture Organization; World Bank, World Development Indicators database
Sectoral Real Value Added (Agriculture, manufacturing, services)	World Bank, World Development Indicators database
Sectoral Labor Productivity	Groningen Growth and Development Centre 10-Sector Database; Timmer, de Vries, and de Vries (2015)
Real Gross Capital Formation	IMF, World Economic Outlook database; World Bank, World Development Indicators database
Real Imports of Goods and Services	IMF, World Economic Outlook database; World Bank, World Development Indicators database
Infant Mortality Rate	World Bank, World Development Indicators database
Human Development Index	United Nations Development Programme, Human Development Report database

Source: IMF staff compilation.

Annex Table 1.1.

Dala Sources

Indicator	Source
Temperature, Historical	Intergovernmental Panel on Climate Change (IPCC) Coupled Model Intercomparison Project Phase Five AR5 Atlas subset; Marcott and others (2013); Matsuura and Willmott (2007); National Aeronautics and Space Administration (NASA) Goddard Institute for Space Studies (GISS); Royal Netherlands Meteorological Institute (KNMI) Climate Change Atlas; Shakun and others (2012)
Temperature and Precipitation, Forecast (Grid level)	NASA Earth Exchange Global Daily Downscaled Projections data set (NEX-GDDP)
Temperature and Precipitation, Historical (Grid level)	University of East Anglia, Climate Research Unit (CRU TS v.3.24); University of Delaware (UDEL v.4.01)
Population 2010, 1990, 1950 (Grid level)	Center for International Earth Science Information Network (CIESIN v.3 and v.4); History Database of the Global Environment (HYDE v3.2); Klein and others (2016)
Population 2015 and Projected Population 2100	United Nations World Population Prospects database, 2015 Revision
CO₂ Emissions	Carbon Dioxide Information Analysis Center
Temperature Forcings	Carbon Dioxide Information Analysis Center: NASA GISS; Roston and Migliozzi (2015)
Real GDP per Capita	IMF, World Economic Outlook database; World Bank, World Development Indicators database
Crop Production Index	Food and Agriculture Organization; World Bank, World Development Indicators database
Sectoral Real Value Added (Agriculture, manufacturing, services)	World Bank, World Development Indicators database
Sectoral Labor Productivity	Groningen Growth and Development Centre 10-Sector Database; Timmer, de Vries, and de Vries (2015)
Real Gross Capital Formation	IMF, World Economic Outlook database; World Bank, World Development Indicators database
Real Imports of Goods and Services	IMF, World Economic Outlook database; World Bank, World Development Indicators database
Infant Mortality Rate	World Bank, World Development Indicators database
Human Development Index	United Nations Development Programme, Human Development Report database

Source: IMF staff compilation.

Annex Table 1.2.

Country and Territory Groups

Source: IMF staff compilation.

^{^*}Not included in the main regression analysis.

Annex 2. Model-Based Analysis

The model used to analyze the long-term impact of climate change is developed and presented in Buffie et al. (2012). It is commonly known as the Debt, Investment and Growth (DIG) model and has served as a workhorse in many IMF studies of low-income countries. The DIG is an optimizing intertemporal model with perfect foresight. It describes a two-sector small open economy model with private and public capital, learning by doing, and endogenous fiscal policies. Public capital is productive and is used in the production function in both sectors. Government spending can raise output directly by augmenting the stock of public capital and can crowd in as well as crowd out private investment.

Firms operate Cobb-Douglas technologies to combine labor, private capital, and public capital (infrastructure) into output in the traded and nontraded sectors. The evolution of total factor productivity (TFP) is exogenous in both sectors. Firms face separate prices for exports, and imports and are assumed to be profit maximizing.

Consumers supply labor and derive utility from consuming the domestic traded good, the foreign traded good, and the domestic nontraded good. These goods are combined into a constant elasticity of substitution basket, and savers maximize the present value of their lifetime utility. The model breaks Ricardian equivalence by including both savers and hand-to-mouth consumers.

The government spends on transfers, debt service, and (partially inefficient) infrastructure investment. It collects revenue from the consumption value-added tax (VAT) and from user fees for infrastructure services. The deficit is financed through domestic borrowing, external concessional borrowing, or external commercial borrowing. Policymakers accept all concessional loans offered by official creditors. The borrowing and amortization schedule for these loans is fixed exogenously. Debt sustainability requires that the VAT and transfers eventually adjust to cover the entire deficit, given the exogenously determined upper limit on taxes and lower limit on transfers. The model incorporates shocks to the government external debt risk premium (or world interest rates).

The majority of the model parameters are set to the same values as in Buffie et al. (2012), with few exceptions, mostly to reflect the decline in global interest rates, the projection of trend GDP growth in low-income countries, and the sample median of public-debt-to-GDP ratios. The parameters that differ from the ones in Buffie et al. (2012) are presented in Annex Table 2.1.

Annex Table 2.1.

Parameterization of the Debt, Investment, and Growth Model

Sources: Bulfie and others (2012); and IMF staff calculations.

Annex Table 2.1.

Parameterization of the Debt, Investment, and Growth Model

Parameter	Value (percent)
Initial Return on Infrastructure Investment	30
Public Domestic Debt-to-GDP Ratio	10
Public Concessional Debt-to-GDP Ratio	30
Public External Commercial Debt-to-GDP Ratio	5
Oil Revenues-1o-GDP Ratio	2
Real Interest Rate on Public Domestic Debt	7
Real Interest Rate on Public External Commercial Debt	4
Trend per Capita Growth Rate	2.8

Sources: Bulfie and others (2012); and IMF staff calculations.

Annex Table 2.1.

Parameterization of the Debt, Investment, and Growth Model

Parameter	Value (percent)
Initial Return on Infrastructure Investment	30
Public Domestic Debt-to-GDP Ratio	10
Public Concessional Debt-to-GDP Ratio	30
Public External Commercial Debt-to-GDP Ratio	5
Oil Revenues-1o-GDP Ratio	2
Real Interest Rate on Public Domestic Debt	7
Real Interest Rate on Public External Commercial Debt	4
Trend per Capita Growth Rate	2.8

Sources: Bulfie and others (2012); and IMF staff calculations.

Simulating the Long-Term Impacts of Climate Change

To trace the long-term impact of climate change, the model incorporates the estimated relationship between temperature and per capita output presented in ^{Table 1}, column (5). The effect is assumed to occur through temperature’s effect on TFP; therefore, the estimated parameters are rescaled so that the model matches the empirically estimated decline of GDP if temperature increases by 1°C. Estimates of the damage to GDP cannot be used directly since GDP is endogenous.

The temperature during 2017–2100 is assumed to follow one of two alternative scenarios: Representative Concentration Pathway (RCP) 4.5 or RCP 8.5. The temperature increases during 2017–2100 are calculated for the median low-income country in the sample and are equal to 2.0°C and 3.9°C for RCP 4.5 and RCP 8.5, respectively.

Modeling Structural Transformation

Structural transformation is generated in the DIG model by introducing diverging trends in sectoral TFP growth, along the lines of Ngai and Pissarides (2007). In their model, faster productivity growth in the traded good sector goes along with a decline in the relative price of traded versus nontraded goods. Given complementarity in final demand, production in the former sector relative to the latter does not increase in the same proportion. The value share of the traded good sector eventually shrinks, even in the presence of international trade. While this approach relies on only one potential driver of structural transformation, it generates the desired increase in employment and nominal-value-added shares of the nontraded goods sector, which is mostly composed of services. The gap in sectoral TFP growth rates is set to replicate the average increase in the service share of value added in LIDCs in the period 1990–2015, which has risen at the rate of 2.5 percentage points a decade. Given this calibration, in the scenario without rising temperatures the employment share of nontraded goods increases from the baseline value of 42.27 to 65 percent over 90 years.

References

Anthoff, David, and Richard Tol. 2010. “FUND–Climate Framework for Uncertainty, Negotiation and Distribution.”
- Search Google Scholar
- Export Citation
Auffhammer, Maximilian, Solomon M. Hsiang, Wolfram Schlenker and Adam Sobel. 2011. “Global Climate Models and Climate Data: A User Guide for Economists.” Unpublished.
- Search Google Scholar
- Export Citation
Barreca, Alan I. 2012. “Climate Change, Humidity, and Mortality in the United States.” Journal of Environmental Economics and Management 63 (1): 19–34.
- Search Google Scholar
- Export Citation
Barreca, Alan I., Karen Clay, Olivier Deschênes, Michael Greenstone, and Joseph Shapiro. 2016. “Adapting to Climate Change: The Remarkable Decline in the US Temperature-Mortality Relationship over the 20th Century.” Journal of Political Economy 124 (1): 105–59.
- Search Google Scholar
- Export Citation
Barrios, Salvador, Luisito Bertinelli, and Eric Strobl. 2006. “Climatic Change and Rural-Urban Migration: The Case of Sub-Saharan Africa.” Journal of Urban Economics 60 (3): 357–71.
- Search Google Scholar
- Export Citation
Barrios, Salvador, Luisito Bertinelli, and Eric Strobl. 2010. “Trends in Rainfall and Economic Growth in Africa: A Neglected Cause of the African Growth Tragedy.” Review of Economics and Statistics 32 (2): 350–66.
- Search Google Scholar
- Export Citation
Barrios, Salvador, Ouattara Bazoumana, and Eric Strobl. 2008. “The Impact of Climatic Change on Agricultural Production: Is It Different for Africa?” Food Policy 33: 287–98.
- Search Google Scholar
- Export Citation
Buckley, B. M., K. J. Anchukaitis, D. Penny, R. Fletcher, E. R. Cook, M. Sano, C. Nam le, A. Wichienkeeo, T. T. Minh, and T. M. Hong. 2010. “Climate as a Contributing Factor in the Demise of Angkor, Cambodia.” Proceedings of the National Academy of Sciences 107 (15): 6748–52.
- Search Google Scholar
- Export Citation
Buffie, Edward, Andrew Berg, Catherine Pattillo, Rafael Portillo, and Luis-Felipe Zanna. 2012. “Public Investment, Growth, and Debt Sustainability: Putting Together the Pieces.” IMF Working Paper 12/144, International Monetary Fund, Washington, DC.
- Search Google Scholar
- Export Citation
Büntgen, U., W. Tegel, K. Nicolussi, M. McCormick, D. Frank, V. Trouet, J. O. Kaplan, Franz Herzig, Karl-Uwe Heussner, Heinz Wanner, Jürg Luterbacher, and Jan Esper. 2011. “2500 Years of European Climate Variability and Human Susceptibility.” Science 331 (6017): 578–82.
- Search Google Scholar
- Export Citation
Burgess, Robin, Olivier Deschênes, Dave Donaldson, and Michael Greenstone. 2014. “The Unequal Effects of Weather and Climate Change: Evidence from Mortality in India.” Unpublished.
- Search Google Scholar
- Export Citation
Burke, Marshall, Solomon M. Hsiang, and Edward Miguel. 2015a. “Global Non-Linear Effect of Temperature on Economic Production.” Nature 527:235–39.
- Search Google Scholar
- Export Citation
Burke, Marshall, Solomon M. Hsiang, and Edward Miguel. 2015b. “Climate and Conflict.” Annual Review of Economics 7: 577–617.
- Search Google Scholar
- Export Citation
Burke, Marshall, John Dykema, David Lobell, Edward Miguel, and Shanker Satyanath. 2015. “Incorporating Climate Uncertainty into Estimates of Climate Change Impacts, with Applications to U.S and African Agriculture.” Review of Economics and Statistics 97(2): 461–471.
- Search Google Scholar
- Export Citation
Burke, Marshall, and Kyle Emerick. 2016. “Adaptation to Climate Change: Evidence from US Agriculture.” American Economic Journal: Economic Policy 8(3): 106–40.
- Search Google Scholar
- Export Citation
Burke, Paul J., and Andrew Leigh. 2010. “Do Output Contractions Trigger Democratic Change?” American Economic Journal: Macroeconomics 2 (4): 124–57.
- Search Google Scholar
- Export Citation
Carleton, Tamma A., and Solomon M. Hsiang. 2016. “Social and Economic Impacts of Climate.” Science 353 (6304).
- Search Google Scholar
- Export Citation
Cashin, Paul, Kamiar Mohaddes, and Mehdi Raissi. 2017. “Fair Weather or Foul? The Macroeconomic Effects of El Niño.” Journal of International Economics 106: 37–54.
- Search Google Scholar
- Export Citation
Center for International Earth Science Information Network (CIESIN) Columbia University. 2016. Gridded Population of the World, Version 4 (GPWv4): Population Count. NASA Socioeconomic Data and Applications Center (SEDAC), Palisades, NY. http://dx.doi.org/10.7927/H4X63JVC. Accessed March 15, 2017.
- Search Google Scholar
- Export Citation
Center for International Earth Science Information Network (CIESIN) Columbia University, United Nations Food and Agriculture Programme (FAO), and Centro Internacional de Agricultura Tropical (CIAT). 2005. Gridded Population of the World, Version 3 (GPWv3): Population Count Grid. NASA Socioeconomic Data and Applications Center (SEDAC), Palisades, NY. http://dx.doi.org/10.7927/H4639MPP. Accessed March 15, 2017.
- Search Google Scholar
- Export Citation
Chaney, Eric. 2013. “Revolt on the Nile: Economic Shocks, Religion, and Political Power.” Econometrica 81: 2033–53.
- Search Google Scholar
- Export Citation
Colmer, Jonathan. 2016. “Weather Labor Reallocation and Industrial Production: Evidence from India.” Unpublished, MIT, Cambridge, MA.
- Search Google Scholar
- Export Citation
Costinot, Arnaud, Dave Donaldson, and Cory Smith. 2016. “Evolving Comparative Advantage and the Impact of Climate Change in Agricultural Markets: Evidence from 1.7 Million Fields around the World.” Journal of Political Economy 124 (1): 205–48.
- Search Google Scholar
- Export Citation
Cullen, H., P. B. deMenocal, S. Hemming, G. Hemming, F. H. Brown, T. Guilderson, and F. Sirocko. 2000. “Climate Change and the Collapse of the Akkadian Empire: Evidence from the Deep Sea.” Geology 28 (4): 379–82.
- Search Google Scholar
- Export Citation
Dell, Melissa, Benjamin F. Jones, and Benjamin A. Olken. 2012. “Temperature Shocks and Economic Growth: Evidence from the Last Half Century.” American Economic Journal: Macroeconomics 4 (3): 66–95.
- Search Google Scholar
- Export Citation
Dell, Melissa, Benjamin F. Jones, and Benjamin A. Olken. 2014. “What Do We Learn from the Weather? The New Climate–Economy Literature.” Journal of Economic Literature 52 (3): 740–98.
- Search Google Scholar
- Export Citation
Deryugina, Tatyana, and Solomon M. Hsiang. 2014. “Does the Environment Still Matter? Daily Temperature and Income in the United States.” NBER Working Paper 20750, National Bureau of Economic Research, Cambridge, MA.
- Search Google Scholar
- Export Citation
Deschênes, Olivier. 2012. “Temperature, Human Health and Adaptation: A Review of the Empirical Literature.” NBER Working Paper 18345, National Bureau of Economic Research, Cambridge, MA.
- Search Google Scholar
- Export Citation
Deschênes, Olivier, and Michael Greenstone. 2011. “Climate Change, Mortality, and Adaptation: Evidence from Annual Fluctuations in Weather in the US.” American Economic Journal: Applied Economics 3 (4): 152–85.
- Search Google Scholar
- Export Citation
Fankhauser, Samuel, and Richard Tol. 2005. “On Climate Change and Economic Growth.” Resource and Energy Economics 27: 1–17.
- Search Google Scholar
- Export Citation
Farid, Mai, Michael Keen, Michael Papaioannou, Ian Parry, Catherine Pattillo, and Anna Ter-Martirosyan. 2016. “After Paris: Fiscal, Macroeconomic, and Financial Implications of Climate Change.” IMF Staff Discussion Note 16/01, International Monetary Fund, Washington, DC.
- Search Google Scholar
- Export Citation
Feng, Shuaizhang, Alan B. Krueger, and Michael Oppenheimer. 2010. “Linkages among Climate Change, Crop Yields, and Mexico–US Cross-Border Migration.” Proceedings of the National Academy of Sciences 107 (32): 14257–62.
- Search Google Scholar
- Export Citation
Feng, Shuaizhang, Michael Oppenheimer, and Wolfram Schlenker. 2012. “Climate Change, Crop Yields, and Internal Migration in the United States.” NBER Working Paper 17734, National Bureau of Economic Research, Cambridge, MA.
- Search Google Scholar
- Export Citation
Füssel, Hans-Martin. 2010. “Modeling impacts and adaptation in global IAMs.” Wiley Interdisciplinary Reviews: Climate Change 1 (2): 288–303.
- Search Google Scholar
- Export Citation
Garg, Teevrat, Maulik Jagnani, and Viz Taraz. 2017. “Human Capital Costs of Climate Change: Evidence from Test Scores in India.” Unpublished.
- Search Google Scholar
- Export Citation
Graff Zivin, Joshua S., Solomon M. Hsiang, and Matthew J. Neidell. 2015. “Temperature and Human Capital in the Short- and Long-Run.” NBER Working Paper 21157, National Bureau of Economic Research, Cambridge, MA.
- Search Google Scholar
- Export Citation
Graff Zivin, Joshua S., and Matthew J. Neidell. 2014. “Temperature and the Allocation of Time: Implications for Climate Change.” Journal of Labor Economics 32 (1): 1–26.
- Search Google Scholar
- Export Citation
Guo, Y., A. Gasparrini, B. Armstrong, S. Li, B. Tawatsupa, A. Tobias, E. Lavigne, M. de Sousa Zanotti Stagliorio Coelho, M. Leone, X. Pan, S. Tong, L. Tian, H. Kim, M. Hashizume, Y. Honda, YL Guo, CF Wu, K. Punnasiri, SM Yi, P. Michelozzi, PH Saldiva, and G. Williams. 2014. “Global Variation in the Effects of Ambient Temperature on Mortality: A Systematic Evaluation.” Epidemiology 25 (6): 781–89.
- Search Google Scholar
- Export Citation
Hallegatte, Stéphane. 2009. “Strategies to Adapt to an Uncertain Climate Change.” Global Environmental Change 19 (2): 240–47.
- Search Google Scholar
- Export Citation
Hallegatte, Stéphane, M. Bangalore, L. Bonzanigo, M. Fay, T. Kane, U. Narloch, J. Rozenberg, D. Treguer, and A. Vogt-Schilb. 2016. Shock Waves: Managing the Impacts of Climate Change on Poverty. Climate Change and Development Series. Washington, DC: World Bank.
- Search Google Scholar
- Export Citation
Hallegatte, Stéphane, Jean-Charles Hourcade, Patrice Dumas. 2007. “Why Economic Dynamics Matter in Assessing Climate Change Damages: Illustration on Extreme Events.” Ecological Economics 62: 330–40.
- Search Google Scholar
- Export Citation
Hallegatte, Stéphane, and Julie Rozenberg. 2017. “Climate Change through a Poverty Lens.” Nature Climate Change 4: 250–56.
- Search Google Scholar
- Export Citation
Haug, Gerald H., Detlef Günther, Larry C. Peterson, Daniel M. Sigman, Konrad A. Hughen, and Beat Aeschlimann. 2003. “Climate and the Collapse of Maya Civilization.” Science 299 (5613): 1731–35.
- Search Google Scholar
- Export Citation
Heal, Geoffrey, and Jisung Park. 2016. “Temperature Stress and the Direct Impact of Climate Change: A Review of an Emerging Literature.” Review of Environmental Economics and Policy 10 (2): 347–62.
- Search Google Scholar
- Export Citation
Hope, Chris. 2011. “The PAGE09 Integrated Assessment Model: A Technical Description.” Cambridge Judge Business School Working Papers 4/11.
- Search Google Scholar
- Export Citation
Hsiang, Solomon M. 2016. “Climate Econometrics.” Annual Review of Resource Economics 8: 43–75.
- Search Google Scholar
- Export Citation
Intergovernmental Panel on Climate Change (IPCC). 2013. Climate Change 2013: The Physical Science Basis. Summary for Policymakers. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, U. K. and New York, NY: Cambridge University Press.
- Search Google Scholar
- Export Citation
Intergovernmental Panel on Climate Change (IPCC). 2014. Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, U.K., and New York, NY: Cambridge University Press.
- Search Google Scholar
- Export Citation
International Monetary Fund (IMF). 2015. “Macroeconomic Developments and Prospects in Low-Income Developing Countries.” Policy Paper, Washington, DC.
- Search Google Scholar
- Export Citation
International Monetary Fund (IMF). 2016a. “Enhancing Resilience to Natural Disasters in Sub-Saharan Africa.” In Regional Economic Outlook: Sub-Saharan Africa. Washington DC, Fall.
- Search Google Scholar
- Export Citation
International Monetary Fund (IMF) 2016b. “Small States’ Resilience to Natural Disasters and Climate Change: Role for the IMF.” Board Paper, Washington, DC.
- Search Google Scholar
- Export Citation
Jordà, Òscar. 2005. “Estimation and Inference of Impulse Responses by Local Projections.” American Economic Review 95 (1): 161–82.
- Search Google Scholar
- Export Citation
Kelley, Colin P., Shahrzad Mohtadi, Mark A. Cane, Richard Seager, and Yochanan Kushnir. 2015. “Climate Change in the Fertile Crescent and Implications of the Recent Syrian Drought.” Proceedings of the National Academy of Sciences 112 (11): 3241–46.
- Search Google Scholar
- Export Citation
Kim, Nam Kyu. 2014. “Revisiting Economic Shocks and Coups.” Journal of Conflict Resolution 60: 3–31.
- Search Google Scholar
- Export Citation
Klein Goldewijk, K., A. Beusen, J. Doelman and E. Stehfest. 2016. “New Anthropogenic Land Use Estimates for the Holocene; HYDE 3.2.”
- Search Google Scholar
- Export Citation
Kudamatsu, Masayuki, Torsten Persson, and David

Advanced Economies	Australia, Austria, Belgium, Canada, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hong Kong SAR,^* Iceland, Ireland, Israel, Italy, Japan, Korea, Latvia, Lithuania, Luxembourg, Macao SAR,^ Malta, Netherlands, New Zealand, Norway, Portugal, Puerto Rico, San Marino,^ Singapore, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Taiwan Province of China,^** United Kingdom, United States
Emerging Market Economies	Albania, Algeria, Angola, Antigua and Barbuda, Argentina, Armenia, Azerbaijan, The Bahamas,^ Bahrain, Barbados, Belarus, Belize, Bosnia and Herzegovina, Botswana, Brazil, Brunei Darussalam, Bulgaria, Cabo Verde, Chile, China, Colombia, Costa Rica, Croatia, Dominica, Dominican Republic, Ecuador, Egypt, El Salvador, Equatorial Guinea, Fiji, Gabon, Georgia, Grenada, Guatemala, Guyana, Hungary, India, ndonesia, Iran, Iraq, Jamaica, Jordan, Kazakhstan, Kosovo,^ Kuwait, Lebanon, Libya, Macedonia FYR, Malaysia, Maldives,^ Marshall slands.^ Mauritius, Mexico, Micronesia,^ Montenegro, Morocco, Namibia, Nauru,^ Oman, Pakistan, Palau,^ Panama, Paraguay, Peru, Philippines, Poland, Qatar, Romania, Russia, Samoa, Saudi Arabia, Serbia, Seychelles.^ South Africa, Sri Lanka, St. Kitts and Nevis, St. Lucia, St. Vincent and the Grenadines, Suriname, Swaziland, Syria, Thailand, Timor-Leste, Tonga, Trinidad and Tobago, Tunisia, Turkey, Turkmenistan, Tuvalu,^** Ukraine. United Arab Emirates, Uruguay, Vanuatu, Venezuela
Low-Income Developing Countries	Afghanistan, Bangladesh, Benin, Bhutan, Bolivia, Burkina Faso, Burundi, Cambodia, Cameroon, Central African Republic, Chad, Comoros, Democratic Republic of the Congo. Republic of Congo, Côte d’Ivoire, Djibouti, Eritrea, Ethiopia, The Gambia, Ghana, Guinea, Guinea-Bissau, Haiti, Honduras, Kenya, Kiribati,^ Kyrgyz Republic, Lao P.D.R., Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Moldova, Mongolia Mozambique, Myanmar, Nepal, Nicaragua, Niger, Nigeria, Papua New Guinea, Rwanda, Senegal, Sierra Leone, Solomon Islands, Somalia,^ South Sudan, Sudan, São Tomé and Príncipe, Tajikistan, Tanzania, Togo, Uganda, Uzbekistan, Vietnam, Yemen, Zambia, Zimbabwe
Countries and Territories with Average Annua Temperature above 15°C	Algeria, American Samoa, Angola, Anguilla, Antigua and Barbuda, Argentina, Australia, Bahrain, Bangladesh. Barbados, Belize, Benin, Bhutan Botswana, Brazil, Brunei Darussalam, Burkina Faso, Burundi, Cabo Verde, Cambodia, Cameroon, Central African Republic, Chad, Colombia, Comoros, Democratic Republic of the Congo, Republic of Congo, Costa Rica, Cuba, Curaçao,^ Cyprus, Côte d’Ivoire, Djibouti, Dominica, Dominican Republic, Ecuador, Egypt, El Salvador, Equatorial Guinea, Eritrea, Ethiopia, Fiji, Gabon, The Gambia, Ghana, Grenada, Guadeloupe,^ Guatemala, French Guiana,^ Guinea, Guinea-Bissau, Guyana, Haiti, Honduras, India, Indonesia, Iraq, Israel, Jamaica, Jordan, Kenya, Kuwait, Lao P.D.R., Lebanon, Liberia, Libya, Madagascar, Malawi, Malaysia, Mali, Malta, Martinique,^ Mauritania, Mauritius, Mexico, Montserrat, Morocco, Mozambique, Myanmar, Namibia, Nepal, New Caledonia, Nicaragua, Niger, Nigeria, Oman, Pakistan, Panama, Papua New Guinea, Paraguay, Philippines, Puerto Rico, Qatar, Reunion,^ Rwanda, Samoa, Saudi Arabia, Senegal, Sierra Leone, Singapore, Solomon Islands, Somalia, South Africa, South Sudan, Sri Lanka, St. Kitts and Nevis, St. Lucia, St. Vincent and the Grenadines, Sudan, Suriname, Swaziland Syria, São Tomé and Príncipe, Tanzania, Thailand, Timor-Leste, Togo, Tonga, Trinidad and Tobago, Tunisia, Turkmenistan, Turks and Caicos,^ Uganda, United Arab Emirates, Uruguay, Vanuatu, Venezuela, Vietnam, Virgin Islands (US), West Bank and Gaza, Yemen, Zambia, Zimbabwe
Countries with Sectoral-Level Data	Argentina, Bolivia, Botswana, Brazil, Chile, China, Colombia, Costa Rica, Denmark, Egypt, Ethiopia, France, Germany, Ghana, Hong Kong SAR,^ India, Indonesia, Italy, Japan, Kenya, Korea, Malawi, Malaysia, Mauritius, Mexicc, Morocco, Netherlands, Nigeria, Peru, Philippines, Senegal, Singapore, South Africa, Spain, Sweden, Taiwan Province of China,^ Tanzania, Thailand, United Kingdom, United States, Venezuela, Zambia

Advanced Economies	Australia, Austria, Belgium, Canada, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hong Kong SAR,^* Iceland, Ireland, Israel, Italy, Japan, Korea, Latvia, Lithuania, Luxembourg, Macao SAR,^ Malta, Netherlands, New Zealand, Norway, Portugal, Puerto Rico, San Marino,^ Singapore, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Taiwan Province of China,^** United Kingdom, United States
Emerging Market Economies	Albania, Algeria, Angola, Antigua and Barbuda, Argentina, Armenia, Azerbaijan, The Bahamas,^ Bahrain, Barbados, Belarus, Belize, Bosnia and Herzegovina, Botswana, Brazil, Brunei Darussalam, Bulgaria, Cabo Verde, Chile, China, Colombia, Costa Rica, Croatia, Dominica, Dominican Republic, Ecuador, Egypt, El Salvador, Equatorial Guinea, Fiji, Gabon, Georgia, Grenada, Guatemala, Guyana, Hungary, India, ndonesia, Iran, Iraq, Jamaica, Jordan, Kazakhstan, Kosovo,^ Kuwait, Lebanon, Libya, Macedonia FYR, Malaysia, Maldives,^ Marshall slands.^ Mauritius, Mexico, Micronesia,^ Montenegro, Morocco, Namibia, Nauru,^ Oman, Pakistan, Palau,^ Panama, Paraguay, Peru, Philippines, Poland, Qatar, Romania, Russia, Samoa, Saudi Arabia, Serbia, Seychelles.^ South Africa, Sri Lanka, St. Kitts and Nevis, St. Lucia, St. Vincent and the Grenadines, Suriname, Swaziland, Syria, Thailand, Timor-Leste, Tonga, Trinidad and Tobago, Tunisia, Turkey, Turkmenistan, Tuvalu,^** Ukraine. United Arab Emirates, Uruguay, Vanuatu, Venezuela
Low-Income Developing Countries	Afghanistan, Bangladesh, Benin, Bhutan, Bolivia, Burkina Faso, Burundi, Cambodia, Cameroon, Central African Republic, Chad, Comoros, Democratic Republic of the Congo. Republic of Congo, Côte d’Ivoire, Djibouti, Eritrea, Ethiopia, The Gambia, Ghana, Guinea, Guinea-Bissau, Haiti, Honduras, Kenya, Kiribati,^ Kyrgyz Republic, Lao P.D.R., Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Moldova, Mongolia Mozambique, Myanmar, Nepal, Nicaragua, Niger, Nigeria, Papua New Guinea, Rwanda, Senegal, Sierra Leone, Solomon Islands, Somalia,^ South Sudan, Sudan, São Tomé and Príncipe, Tajikistan, Tanzania, Togo, Uganda, Uzbekistan, Vietnam, Yemen, Zambia, Zimbabwe
Countries and Territories with Average Annua Temperature above 15°C	Algeria, American Samoa, Angola, Anguilla, Antigua and Barbuda, Argentina, Australia, Bahrain, Bangladesh. Barbados, Belize, Benin, Bhutan Botswana, Brazil, Brunei Darussalam, Burkina Faso, Burundi, Cabo Verde, Cambodia, Cameroon, Central African Republic, Chad, Colombia, Comoros, Democratic Republic of the Congo, Republic of Congo, Costa Rica, Cuba, Curaçao,^ Cyprus, Côte d’Ivoire, Djibouti, Dominica, Dominican Republic, Ecuador, Egypt, El Salvador, Equatorial Guinea, Eritrea, Ethiopia, Fiji, Gabon, The Gambia, Ghana, Grenada, Guadeloupe,^ Guatemala, French Guiana,^ Guinea, Guinea-Bissau, Guyana, Haiti, Honduras, India, Indonesia, Iraq, Israel, Jamaica, Jordan, Kenya, Kuwait, Lao P.D.R., Lebanon, Liberia, Libya, Madagascar, Malawi, Malaysia, Mali, Malta, Martinique,^ Mauritania, Mauritius, Mexico, Montserrat, Morocco, Mozambique, Myanmar, Namibia, Nepal, New Caledonia, Nicaragua, Niger, Nigeria, Oman, Pakistan, Panama, Papua New Guinea, Paraguay, Philippines, Puerto Rico, Qatar, Reunion,^ Rwanda, Samoa, Saudi Arabia, Senegal, Sierra Leone, Singapore, Solomon Islands, Somalia, South Africa, South Sudan, Sri Lanka, St. Kitts and Nevis, St. Lucia, St. Vincent and the Grenadines, Sudan, Suriname, Swaziland Syria, São Tomé and Príncipe, Tanzania, Thailand, Timor-Leste, Togo, Tonga, Trinidad and Tobago, Tunisia, Turkmenistan, Turks and Caicos,^ Uganda, United Arab Emirates, Uruguay, Vanuatu, Venezuela, Vietnam, Virgin Islands (US), West Bank and Gaza, Yemen, Zambia, Zimbabwe
Countries with Sectoral-Level Data	Argentina, Bolivia, Botswana, Brazil, Chile, China, Colombia, Costa Rica, Denmark, Egypt, Ethiopia, France, Germany, Ghana, Hong Kong SAR,^ India, Indonesia, Italy, Japan, Kenya, Korea, Malawi, Malaysia, Mauritius, Mexicc, Morocco, Netherlands, Nigeria, Peru, Philippines, Senegal, Singapore, South Africa, Spain, Sweden, Taiwan Province of China,^ Tanzania, Thailand, United Kingdom, United States, Venezuela, Zambia

Topics

Countries

Series

About

Resources

Contributor Notes

Abstract

I. Introduction

II. Temperature and Precipitation: Historical Patterns and Projections

III. Empirical Strategy and Data

IV. RESULTS

A. Short-term Effect on Per Capita Output

B. Medium-term Effect on Per Capita Output

C. Robustness

V. Channels of Impact

Sectoral Output: Agriculture, Manufacturing, Services

Productivity

Capital Accumulation

Labor Supply

VI. Long-Term Effects of Temperature Increase–A Model-Based Approach

VII. Summary and Policy Implications

Annex 1. Data Sources and Country Groupings

Annex 2. Model-Based Analysis

Simulating the Long-Term Impacts of Climate Change

Modeling Structural Transformation

References