The Effects of Climate and Weather on Economic Output: Evidence from Global Subnational Data

Jinchi Dong; Jinnan Wang; Richard S.J. Tol

REVIEW 3 major objections 5 minor 1 cited by

The Effects of Climate and Weather on Economic Output: Evidence from Global Subnational Data

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Using GDP data for more than 1,600 regions in 196 countries, this paper finds that temperature shocks significantly affect economic growth in the short run but that the effect fades over ten-year horizons everywhere except the coldest and…

desk verdict The sample-selection result is the real contribution; the long-run insignificance claim is underpowered and should be softened. read the letter →

arxiv 2505.17946 v1 pith:4KPREDIQ submitted 2025-05-23 econ.GN q-fin.EC

Jinchi Dong , Richard S.J. Tol , Jinnan Wang This is my paper · ORCID

classification econ.GNq-fin.EC

keywords climatedamageimpactspanelregressionlong-differencefunctioneconomicgrowthadaptationsubnationalGDPdata

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Using GDP data for 1,666 subnational regions in 196 countries from 1990 to 2015, this paper asks whether weather and climate affect economic output temporarily or permanently. It finds that temperature shocks have a large and significant effect on economic growth in the short run, but that over ten-year horizons the effect becomes statistically insignificant for most of the world. The exceptions are the coldest and hottest places, where temperature changes continue to affect growth in the long run. The paper argues that earlier studies using data from only 77 countries omitted many hot, poor countries and therefore underestimated the short-run effects of climate. This distinction matters because permanent growth effects would make future climate damages more than twenty times larger than transient level effects.

What carries the argument

The machinery is a pair of regression models derived from a production function in which weather enters both the level of output and the growth rate of total factor productivity. In the annual panel, first-differencing removes fixed regional factors, and the coefficients on the temperature level, its square, the temperature change, and the interaction between the two separate level effects from growth effects. In the long-difference model, variables are averaged over five- or ten-year periods and then differenced across adjacent periods, so the same coefficients capture long-run climate responses. The quadratic temperature specification is the workhorse: it implies an optimal temperature of about 14.6°C in the annual panel and identifies the cold and hot extremes where marginal effects become significant.

What would settle it

Re-estimate the long-difference model with 20- or 30-year averages once longer subnational GDP series become available: if the temperature-growth effect remains significant for non-extreme regions over those horizons, the paper's claim that long-run effects are insignificant would be refuted. A complementary test would be to check whether the insignificance survives when spatial spillovers across regions are explicitly modelled.

Watch

Extended reading notes

Core claim

The paper's central claim is that weather shocks have a transient effect on output: in the global subnational panel, temperature changes significantly affect GDP per capita growth within the year, but at ten-year intervals the effect disappears except in regions with annual mean temperatures below about 5°C or above about 29°C. The annual panel model separates effects on the level of output from effects on the growth rate and finds that temperature operates through growth rather than the level. The long-difference model, which compares five- and ten-year averages across periods, shows that for most regions long-run adaptation offsets the short-run damage, while in extreme climates the damage persists. The paper also claims that vulnerable poor countries are not intrinsically more sensitive: rich and poor regions have nearly identical marginal temperature responses, and poor countries appear more affected simply because they are hotter on average.

Load-bearing premise

The long-run conclusion rests on the assumption that ten-year averages in the long-difference model are long enough to capture true adaptation to climate; the paper itself notes that climate is typically defined over 30 years, and if adaptation takes longer than a decade the long-difference estimates could still include short-run dynamics.

Editorial extensions

If this is right

A dataset limited to the 77 countries used in earlier work makes the temperature-growth effect statistically insignificant, so global coverage is necessary to detect the short-run damage.
For most regions, the temperature effect on growth vanishes at the ten-year horizon, implying substantial long-run adaptation after weather shocks.
In cold and hot extremes, the long-run growth effect persists: about +3.3% annual growth per 1°C at 5°C and -2.9% at 29°C.
Global GDP per capita in 2100 under 2.0°C warming is projected to fall 34.5-35.6% from short-run estimates but the long-difference projection is statistically indistinguishable from zero.
Rich and poor regions show nearly identical marginal temperature effects at the same temperature, so poor countries' greater vulnerability comes from being hotter, not poorer.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Because the long-difference window is ten years rather than the conventional thirty, the paper's long-run estimates may still contain short-run adjustment dynamics; if adaptation continues beyond a decade, true long-run damages for non-extreme regions could be even smaller than reported.
The short-run-growth/long-run-level pattern implies that damage functions are horizon-specific; applying the annual panel coefficients to century-scale projections would overstate climate losses and the social cost of carbon.
A natural extension would be to apply the same level-vs-growth decomposition to sectoral GDP to test whether the transient pattern holds for agriculture, industry, and services alike.
If spatial spillovers (which the paper lists as a caveat) are substantial, the ten-year long-difference estimates could be biased in either direction; modelling spillovers would sharpen the long-run claim.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Read this for the sample-selection result, not for the long-run conclusion. The strongest thing in the paper is the demonstration that global subnational coverage changes the estimated temperature-growth relationship. Using the same estimator, the Kummu data restricted to the 77 Kalkuhl-Wenz countries reproduces the insignificant temperature effect, while the full 196-country sample yields a significant effect (Table 6). That is new, credible, and has direct implications for the level-vs-growth debate and for social cost estimates. The paper also does a lot of robustness work—alternative climate and output data, fixed effects, lags, nonparametrics, bootstrap—and the main short-run result holds up.

The long-run half is much weaker. The 'panel long-difference' with 10-year averages is in practice a single long difference per region: the sample runs 1990-2015, so there are two periods and effectively one differenced observation. That design cannot separate 10-year dynamics from 30-year adaptation. The authors acknowledge in Section VII that climate is usually defined over 30 years and that their 10-year window may underestimate adaptation. The bootstrap adaptation ratios in Figure 5 are mostly indistinguishable from zero, so 'statistically insignificant in the long-run' is better read as 'imprecisely estimated.' I would push the authors to frame it that way.

The 2100 projections are mechanical transformations of the fitted coefficients. They are fine as illustrations but should be labeled as such; they are not independent evidence. A minor point: I did not see replication code or processed data, and for a paper whose main claim rests on a new global subnational dataset, that should be available.

Verdict: the sample-selection finding deserves a serious referee and is publishable. The long-run claims need a major reframing or a longer-horizon test. I would send it to peer review and would cite it for the 196-country comparison.

Referee Report

3 major / 5 minor

Summary. The paper estimates the effect of temperature and precipitation on GDP per capita growth using subnational data for 1,666 regions in 196 countries over 1990–2015. It uses an annual panel model (Eq. 11) and a panel long-difference model (Eq. 12) with 5- and 10-year averages. The annual panel shows a significant short-run effect of temperature on growth, with an optimal temperature around 14.6°C, and the 10-year long-difference model shows significant growth effects only in very cold and very hot regions under region weighting. The paper argues that global coverage matters, reports extensive robustness checks, and projects 2100 GDP losses of about 35% using the panel estimates and 11–20% (statistically insignificant) using the long-difference estimates.

Significance. If the short-run/long-run distinction holds, the paper makes a substantial contribution: it extends subnational evidence to nearly all countries, compares alternative output and climate databases, and provides transparent robustness checks including bootstrap uncertainty. The demonstration that estimates change when hot, poor countries are included is valuable and well supported by Table 6. The main limitation is that the long-run conclusion is identified from a short panel with at most two 10-year differences per region, so the paper's own caveat about 30-year climate definitions is central to interpreting the headline result.

major comments (3)

[Section II.B, Eq. (12); Table 3; Figure 5; Section VII] The central long-run conclusion rests on a low-power design. With 10-year averages over 1990–2015, there are at most two long-difference observations per region; after region fixed effects, identification comes from the change between two 10-year growth windows, i.e., a second difference of log GDP. This design cannot separate 'no long-run effect' from 'an effect that materializes over more than ten years,' which is exactly the adaptation horizon the authors concede in Section VII (climate is conventionally defined over 30 years). Figure 5 shows bootstrap adaptation ratios whose confidence intervals include zero for most regions, so the abstract's 'statistically insignificant in the long-run' should be rephrased as 'imprecisely estimated.' The authors should add a power analysis or a longer-horizon (e.g., 20- or 30-year) long-difference check using country-level data that extend before 1990.
[Table 3, columns (5)–(6); Figure 3] The headline long-run result is specific to region weighting. Under population weighting, the long-difference estimates show no significant growth effect of temperature, but they do show significant negative level effects at warm temperatures (e.g., 18–23% output-level loss at 20–25°C, Figure 3 Panel A). The unqualified statement in the abstract that the long-run impact is statistically insignificant except in the coldest and hottest places therefore holds only for the growth channel under region weighting. The paper should state this qualification prominently and discuss which weighting is policy-relevant.
[Section V, Eq. (13)] The 2100 projections are a mechanical transformation of the estimated damage functions. Equation (13) defines the percentage change as the cumulative sum of g_b(T_ct) − g_b(T_i0), where g_b is the fitted quadratic from the panel or long-difference model. Consequently, the comparison between the 34.5–35.6% panel-based loss and the 11–20% long-difference-based loss is a restatement of the difference between the Table 2 and Table 3 coefficients, not an independent validation. The reported uncertainty covers bootstrap sampling and CMIP6 pattern spread only; it does not cover specification uncertainty, notably the quadratic functional form and the 10-year window. The paper should state this explicitly and present the projections as calibrated scenarios rather than as separate evidence.

minor comments (5)

[Tables 3, 5, 6, 8, 9] The long-difference models include period fixed effects θ_p (Eq. 12), but the table notes label them 'Year FE'; this should be corrected to 'Period FE' for the long-difference columns.
[Appendix I] The serial correlation test is attributed to Born and Breitung (2016), but this reference is missing from the reference list and should be added.
[Table 1] The summary-statistics row for GDP per capita appears garbled: '11315 14141 177 459271.4 43316' is not formatted like the other rows, and the maximum value of 459,271 is inconsistent with the text stating that Qatar has the highest average GDP per capita at $104,617.
[Section III.A and Section II.A] The population grid dataset is referred to as 'LanSan' but should be 'LandScan,' and 'El Ni˜no or La Ni˜na' appears with a formatting/encoding issue that should be fixed.
[Section VI, Bootstrap estimates] The sentence 'which is just 0.9% higher' should read '0.9 percentage points higher' to avoid confusing percentage-point differences with percent differences.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the empirical estimates and projections are self-contained, and the cited self-work is not load-bearing.

full rationale

The paper's central estimates (annual panel, Table 2; long-difference, Table 3) are identified from subnational variation in temperature and precipitation, not constructed from the quantities they are later used to predict. The level-growth decomposition in Equations (4), (9), and (10) is an algebraic consequence of the assumed production function (1)-(2), with the coefficients freely estimated rather than imposed, so the finding of short-run growth effects and statistically insignificant long-run effects is not an artifact of the definitions. The Section V projections apply the estimated damage functions to exogenous SSP1-RCP2.6 climate scenarios; this is standard out-of-sample prediction from fitted regressions, not a fit disguised as a prediction. The paper also reports both panel-based and long-difference-based projections, explicitly acknowledging that the former reflect weather shocks rather than long-run climate effects. Self-citations (e.g., Fankhauser and Tol 2005; Letta and Tol 2019) motivate the TFP-growth channel but are not the evidence for the paper's conclusions, which rest on the regressions and robustness checks; therefore, they are not load-bearing. The 10-year versus 30-year window is an explicitly acknowledged data limitation that affects the interpretation of 'long-run' but does not make the derivation circular.

Assumptions & free parameters 8 free parameters · 7 assumptions · 0 invented entities

The central results are empirical estimates from fitted regression models. The level vs growth interpretation rests on the assumed quadratic production and TFP growth equations, and the long-run conclusions rest on a ten-year window. No new physical entities are introduced.

free parameters (8)

Temperature level effect coefficient alpha0 (on Delta T * T) = 0.000507 (annual panel, Table 2 col 2); 0.0146 (long-difference 10yr, Table 3 col 3)
Fitted coefficient on the interaction of temperature change with temperature level. It identifies the level effect of weather on output and is central to the level vs growth decomposition.
Temperature level effect coefficient beta0 (on Delta T) = -0.00487 (annual panel); -0.351 (long-difference)
Fitted coefficient on temperature change, part of the level effect channel.
Temperature growth effect quadratic coefficient gamma0 (on T^2) = -0.000774 (annual panel); -0.0149 (long-difference)
Fitted coefficient on squared temperature. It drives the nonlinear growth effect and the optimal temperature.
Temperature growth effect linear coefficient delta0 (on T) = 0.0226 (annual panel); 0.534 (long-difference)
Fitted coefficient on temperature level. Together with gamma0 it defines the growth response to temperature.
Precipitation coefficients (Delta P * P, Delta P, P^2, P) = -0.000998, -0.00288, -0.00405, 0.0169 (annual panel, Table 2 col 2)
Fitted coefficients for precipitation level and growth effects. Precipitation results are mostly insignificant and not robust across climate data sources.
Long-difference window m = 5 years and 10 years
Chosen by the authors. The 10-year window underlies the long-run claims but is shorter than the standard 30-year climate normal.
Weighting scheme = Region weight (inverse of number of subnational regions) and population weight
The choice of weighting changes the significance of temperature effects, especially between region-weighted and population-weighted regressions.
Control variable coefficients = Not reported in Table 2
Coefficients on urbanization, schooling, and population are estimated in some specifications but not displayed.

assumptions (7)

domain assumption Output per capita is an exponential function of quadratic temperature and precipitation in levels, times total factor productivity.
Section I, Eq (1). This functional form is adopted from Burke, Hsiang and Miguel (2015b) and Kalkuhl and Wenz (2020). The level vs growth decomposition is only valid if the true relationship is quadratic and additive in logs.
domain assumption TFP growth is a quadratic function of temperature and precipitation.
Section I, Eq (2). This is the growth channel. If weather affects TFP growth through other mechanisms, such as investment, the decomposition mislabels effects.
domain assumption Conditional on fixed effects, region-specific trends, and controls, weather shocks are uncorrelated with other determinants of output growth.
Section II.A. Standard panel identification assumption. If adaptation or expectations correlate with weather, the coefficients are biased.
domain assumption Regional output is independent across regions; there are no cross-region spillovers.
Section VII, second caveat. The authors note that spillovers would make residuals neither independent nor identically distributed.
ad hoc to paper Ten-year averages capture long-run climate adaptation.
Section I and Section VII, first caveat. The authors use m=10 because data end in 2015, but note that climate is normally defined over 30 years and that they may underestimate adaptation.
domain assumption The Kummu et al. (2018) subnational GDP database accurately measures regional output.
Section III and Appendix II. The authors validate against World Bank data, but the database is partly interpolated from national data, especially after 2010.
domain assumption The estimated historical temperature-growth relationship persists to 2100 and applies to all regions.
Section V projections. This is an out-of-sample extrapolation with no independent validation.

how reviews work

0 comments

Cite this review

Pith. "Pith review of The Effects of Climate and Weather on Economic Output: Evidence from Global Subnational Data." pith.science (2026). https://pith.science/paper/4KPREDIQ

@misc{pith2026250517946,
  author       = {Pith},
  title        = {Pith review of: The Effects of Climate and Weather on Economic Output: Evidence from Global Subnational Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4KPREDIQ}},
  note         = {Machine review of arXiv:2505.17946}
}

read the original abstract

Estimating the effects of climate on economic output is crucial for formulating climate policy, but current empirical findings remain ambiguous. Using annual panel model and panel long-difference model with global subnational data from nearly all countries, we find robust evidence that weather shocks have a transient effect on output. The impact on economic growth is large and significant in the short-run but statistically insignificant in the long-run, except in the coldest and hottest places.

Figures

Figures reproduced from arXiv: 2505.17946 by the authors.

**Figure 2.** Marginal effects of temperature, precipitation on output—Panel estimates [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗

**Figure 3.** Marginal effects of temperature, precipitation on output —Long-difference esti [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: Marginal effects of temperature on output growth under different economic lev [PITH_FULL_IMAGE:figures/full_fig_p017_4.png]

**Figure 5.** Figure 5: Marginal effects of temperature on output growth under different economic lev [PITH_FULL_IMAGE:figures/full_fig_p018_5.png]

**Figure 6.** Figure 6: Regional and global aggregated GDP per capita changes [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]

**Figure 7.** Figure 7: Point estimate results for temperature and precipitation [PITH_FULL_IMAGE:figures/full_fig_p026_7.png]

**Figure 8.** Figure 8: Bootstrap estimates results based on Panel and Long-difference models [PITH_FULL_IMAGE:figures/full_fig_p029_8.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

To what extent can long-differencing capture climate adaptation?
econ.EM 2026-07 conditional novelty 6.0 of 10

Long-difference and fixed-effects estimates of climate impacts are each mixtures of short- and long-run responses, so their difference is a biased, often low-power test of adaptation.

Reference graph

Works this paper leans on

32 extracted references · 32 canonical work pages · cited by 1 Pith paper

[1]

Since the panel data cover decades, the coefficients in the panel model jointly capture the weather and climate effects

to control gradual changes in individual regions’ growth rates driven by slowly changing factors, such as the gradual increased adaptation to the climate change. Since the panel data cover decades, the coefficients in the panel model jointly capture the weather and climate effects. However, after including region-specific time trends, the impacts from gra...

work page 2018
[2]

1℃ increase is expected to increase GDP per capita growth by 1.5% in regions with an average temperature of 5℃and decrease GDP per capita growth by 1.6% at 25℃

When weighted by region, only temperature shows a significant marginal effect on output growth when temperatures are below 5℃or above 22℃. 1℃ increase is expected to increase GDP per capita growth by 1.5% in regions with an average temperature of 5℃and decrease GDP per capita growth by 1.6% at 25℃. However, under population weighting, the marginal effects...

work page 2020
[3]

rich” or “poor

Marginal effects of temperature, precipitation on output —Long-difference esti- mates Note:This figure shows the marginal effects of temperature on GDP per capita growth (top), and the marginal effects of precipitation on GDP per capita growth (bottom) based on a long-difference model. The orange line represents the estimates based on region-weighted regr...

work page 2015
[4]

Descriptive Statistics Table 1 summarizes the subnational data used in this study

4https://globaldatalab.org/shdi DONG ET AL: THE EFFECTS OF CLIMATE CONDITIONS 9 B. Descriptive Statistics Table 1 summarizes the subnational data used in this study. Our sample in- cludes 1,666 subnational regions from 196 countries, covering nearly all countries and populations worldwide (excluding control variables). Between 1990 to 2015, the global ave...

work page 1990
[5]

Changes in temperature, precipitation, and GDP per capita from 1990 to

work page 1990
[6]

A and b are the results based on the annual panel model, and c and d are the results based on the long-difference model

Regional and global aggregated GDP per capita changes Note:This figure shows the regional percentage changes in GDP per capita in 2100 (left) and global aggregated GDP per capita changes from 2020 to 2100 (right). A and b are the results based on the annual panel model, and c and d are the results based on the long-difference model. b,d are the GDPpc weig...

work page 2020
[7]

Table 4 summarizes the GDP per capita changes at the subcontinent level

Figure 6 shows the projected regional GDP per capita changes in 2100 based on annual panel and long-difference models. Table 4 summarizes the GDP per capita changes at the subcontinent level. We find that temperature- induced output losses not only differ between countries, but also show pronounced variation within countries. For instance, while prior stu...

work page 2020
[8]

The orange line represents the estimates based on region-weighted regression

Marginal effects of temperature, precipitation on output—Panel estimates Note:This figure shows the marginal effects of temperature on GDP per capita growth (top), and the marginal effects of precipitation on GDP per capita growth (bottom) based on the panel model. The orange line represents the estimates based on region-weighted regression. The green lin...

work page 2025

Show all 32 references

[10]

The orange line represents the poor regions, while the green line represents the rich regions

Marginal effects of temperature on output growth under different economic lev- els Note:This figure shows the marginal effects of temperature on GDP per capita growth based on the panel model (panel A) and the long-difference model (panel B). The orange line represents the poo...

2015
[11]

The whiskers represent the fifth to ninety percentile

Marginal effects of temperature on output growth under different economic lev- els Note:This figure shows the percentage of the short-term effects of temperature on output growth that are mitigated in the longer run for poor (panel A) and rich (panel B) regions. The whiskers r...

2018
[14]

The optimal temperature implied by column (1) is 16.2℃, which is 1.6℃higher than the value derived from subnatioanl-level data. These results confirm the findings of Damania, Desbureaux and Zaveri (2020), which suggests that the data aggregating to large spatial scales masks t...

2020
[15]

These consistences suggest the reliability of the Kummu database used in this study

The mag- nitude and significance of the temperature coefficients are also consistent across columns. These consistences suggest the reliability of the Kummu database used in this study. Second, we compare our results with another subnational database collected by Kalkuhl and W...

2020
[17]

bin counts temperature≥27.P n it is defined similarly for annual total precipitation across 12 bins. Each bin spans 0.2m with the top bin corresponding to precipitation≥2.2m.η i is the region fixed effects.θ t is the year fixed effects.h i(t) is the linear region-specific time...

2021
[18]

Overall, the point estimates support the validity of using the quadratic function to capture the effects of temperature and precipitation on output. However, for precipitation, a piecewise function might provide a better fit for analyzing its effects on output, which is beyond...

2025
[19]

continent

Point estimate results for temperature and precipitation Note:This figure shows the effects of temperature (Panel A) and precipitation (Panel B) on GDP per capita growth based on baseline regression. Both of them are based on region-weighted regression. The shadow areas repres...

2020
[20]

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

Bootstrap estimates results based on Panel and Long-difference models Note:This figure shows the bootstrapped estimates of the marginal effects of temperature on GDP per capita growth based on the panel model (Panel A) and the long-difference model (Panel B). Dots are the medi...

2020
[29]

The data stem from various statistical agencies of central or federal governments

However, it is a highly unbalanced panel database, with over 60% of regions having a duration of less than 25 years. The data stem from various statistical agencies of central or federal governments. 44 out of 77 countries used GDP data, while the others used other data to mea...

2015
[30]

It is a country-level database that contains 266 countries from 1960 to

1960
[32]

Table A6 presents the Pearson and Spearman correlation test results for the three databases

***p<0.01, **p<0.05, *p<0.10. Table A6 presents the Pearson and Spearman correlation test results for the three databases. Correlation coefficients approaching±1 denote stronger corre- lations. The results in Table A6 show significant correlations among all three databases, wi...
[50]

Estimating the impacts of climate change: reconciling disconnects between physical climate and statistical models

Polonik, Pascal, Katharine Ricke, and Jennifer Burney .2025. “Estimating the impacts of climate change: reconciling disconnects between physical climate and statistical models.”Climatic Change, 178(2). Pretis, F elix, Moritz Schwarz, Kevin T ang, Karsten Haustein, and Myles R....

2025
[109]

The impact of weather on economic growth and its production factors

Henseler, Martin, and Ingmar Schumacher.2019. “The impact of weather on economic growth and its production factors.”Climatic Change, 154(3): 417–

2019
[221]

xtqptest

36 WORKING PAPER MAY 2025 Mathematical Appendix: The Effects of Climate Conditions Jinchi Dong, Richard S.J. Tol, Jinnan W ang Appendix I: Pre- and Post-estimation Tests. —We first conduct the unit root test to check the stationary of variables. Since our panel data is a short...

2016
[268]

Long-term macroeco- nomic effects of climate change: A cross-country analysis

Kahn, Matthew E, Kamiar Mohaddes, Ryan NC Ng, M Hashem Pe- saran, Mehdi Raissi, and Jui-Chung Y ang.2021. “Long-term macroeco- nomic effects of climate change: A cross-country analysis.”Energy Economics, 104: 105624. Kalkuhl, Matthias, and Leonie W enz.2020. “The impact of cli...

2021
[374]

Does rainfall matter for economic growth? Evidence from global sub-national data (1990–2014)

Damania, Richard, Sebastien Desbureaux, and Esha Zaveri.2020. “Does rainfall matter for economic growth? Evidence from global sub-national data (1990–2014).”Journal of Environmental Economics and Management, 102: 102335. Dell, Melissa, Benjamin F. Jones, and Benjamin A. Olken....

2020
[433]

Bootstrap methods in econometrics

DONG ET AL: THE EFFECTS OF CLIMATE CONDITIONS 33 Horowitz, Joel L.2019. “Bootstrap methods in econometrics.”Annual Review of Economics, 11(1): 193–224. Hsiang, Solomon.2016. “Climate econometrics.”Annual Review of Resource Economics, 8(1): 43–75. Jongman, Brenden, Stefan Hochr...

2019
[557]

Day-to-day temperature vari- ability reduces economic growth

Kotz, Maximilian, Leonie W enz, Annika Stechemesser, Matthias Kalkuhl, and Anders Levermann.2021. “Day-to-day temperature vari- ability reduces economic growth.”Nature Climate Change, 11(4): 319–325. Kummu, Matti, Maija T aka, and Joseph HA Guillaume.2018. “Gridded global data...

2021
[1990]

These increasing trends indicate an underlying non-stationary process, which may result in spurious results when panel models

On average, the global temperature increased by 0.50℃, precipitation increased by 46 mm, GDP per capita increased by$6,168, and the GDP per capita growth rate increased by 2.7% when comparing the average values from 1990-1994 to those from 2011-2015. These increasing trends in...

2015
[2014]

Table 6 shows the regression results

To ensure consistency, we limit the time series for both the Kummu and Kalkuhl databases to the period from 1990 to 2014 and apply region weighting for each regression. Table 6 shows the regression results. Column (1) is the results based on the Kummu database with all observa...

1990
[2015]

Note:This figure shows the global average temperature, precipitation, GDP per capita, and GDP per capita growth from 1990 to

1990
[2021]

Table A4 shows the descriptive statistics of Kummu, Kalkuhl and the World Bank database

The values of the data are converted to 2017 constant international US dollars (2017 PPP). Table A4 shows the descriptive statistics of Kummu, Kalkuhl and the World Bank database. To ensure comparability among these three databases, we aggre- gated the Kummu and Kalkuhl data f...

2017
[2023]

Population data after 1950 is annual, sourced from the United Nations World Population Prospects and downscaled to a 5 arc-minute ×5 arc-minute grid based on the global 1 km×1 km population grid data (LanSan) published by the Oak Ridge National Laboratory (USA). Additionally, ...

1950
[2025]

Climate shocks, economic activity and cross-country spillovers: Evidence from a new global model

“Climate shocks, economic activity and cross-country spillovers: Evidence from a new global model.”Economic Modelling, 148: 107082. Anthoff, David, and Johannes Emmerling.2019. “Inequality and the social cost of carbon.”Journal of the Association of Environmental and Resource ...

2019

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Since the panel data cover decades, the coefficients in the panel model jointly capture the weather and climate effects

to control gradual changes in individual regions’ growth rates driven by slowly changing factors, such as the gradual increased adaptation to the climate change. Since the panel data cover decades, the coefficients in the panel model jointly capture the weather and climate effects. However, after including region-specific time trends, the impacts from gra...

work page 2018

[2] [2]

1℃ increase is expected to increase GDP per capita growth by 1.5% in regions with an average temperature of 5℃and decrease GDP per capita growth by 1.6% at 25℃

When weighted by region, only temperature shows a significant marginal effect on output growth when temperatures are below 5℃or above 22℃. 1℃ increase is expected to increase GDP per capita growth by 1.5% in regions with an average temperature of 5℃and decrease GDP per capita growth by 1.6% at 25℃. However, under population weighting, the marginal effects...

work page 2020

[3] [3]

rich” or “poor

Marginal effects of temperature, precipitation on output —Long-difference esti- mates Note:This figure shows the marginal effects of temperature on GDP per capita growth (top), and the marginal effects of precipitation on GDP per capita growth (bottom) based on a long-difference model. The orange line represents the estimates based on region-weighted regr...

work page 2015

[4] [4]

Descriptive Statistics Table 1 summarizes the subnational data used in this study

4https://globaldatalab.org/shdi DONG ET AL: THE EFFECTS OF CLIMATE CONDITIONS 9 B. Descriptive Statistics Table 1 summarizes the subnational data used in this study. Our sample in- cludes 1,666 subnational regions from 196 countries, covering nearly all countries and populations worldwide (excluding control variables). Between 1990 to 2015, the global ave...

work page 1990

[5] [5]

Changes in temperature, precipitation, and GDP per capita from 1990 to

work page 1990

[6] [6]

A and b are the results based on the annual panel model, and c and d are the results based on the long-difference model

Regional and global aggregated GDP per capita changes Note:This figure shows the regional percentage changes in GDP per capita in 2100 (left) and global aggregated GDP per capita changes from 2020 to 2100 (right). A and b are the results based on the annual panel model, and c and d are the results based on the long-difference model. b,d are the GDPpc weig...

work page 2020

[7] [7]

Table 4 summarizes the GDP per capita changes at the subcontinent level

Figure 6 shows the projected regional GDP per capita changes in 2100 based on annual panel and long-difference models. Table 4 summarizes the GDP per capita changes at the subcontinent level. We find that temperature- induced output losses not only differ between countries, but also show pronounced variation within countries. For instance, while prior stu...

work page 2020

[8] [8]

The orange line represents the estimates based on region-weighted regression

Marginal effects of temperature, precipitation on output—Panel estimates Note:This figure shows the marginal effects of temperature on GDP per capita growth (top), and the marginal effects of precipitation on GDP per capita growth (bottom) based on the panel model. The orange line represents the estimates based on region-weighted regression. The green lin...

work page 2025