## wpiea2019215-print-pdf

## Source details

**Canonical URL:** [wpiea2019215-print-pdf](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019215-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019215-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019215-print-pdf.pdf.json)

---

### 1. Introduction: purpose, scope, and approach
- Focus: long-term economic impacts of climate change using a multi-country stochastic growth framework; includes empirical and counterfactual analyses and a United States case study.
- Research questions: how deviations in climate variables affect per capita economic growth and welfare over the long run.
- Methods: country-level historical climate and GDP data, stochastic growth modeling, panel ARDL estimation, HPJ-FE bias correction, and scenario-based counterfactual projections.
- Organization: empirical estimation of climate–growth relationships, robustness checks, counterfactual projections under alternative climate scenarios, sectoral analysis for the United States, appendices with additional results and data documentation.

### 3.1 Historical temperature patterns (empirical facts)
- Country-specific trends:
  - Regression: T_it = a_Ti + b_Ti t + v_Ti,t for i = 1,...,N = 174; T_it in degrees Celsius (  C).
  - Individual b_Ti over 1960ñ2014 range: -0:0044 (Samoa) to 0:0390 (Afghanistan).
  - 169 countries (97.1% of cases) have positive b_Ti estimates.
  - 161 countries (95.3% of cases) have b_Ti estimates statistically significant at the 5% level.
  - Five non-positive cases (1960ñ2014): Bangladesh, Bolivia, Cuba, Ecuador, and Samoa (none statistically significant at 5%).
- Longer horizon (1900ñ2014):
  - b_Ti range: -0:0008 (Greece) to 0:0190 (Haiti).
  - 172 countries (98.9% of cases) positive; 156 countries (90.7% of cases) significant at 5%.
  - Non-positive: Greece and Macedonia (not significant).
- Global pooled estimate (1960ñ2014):
  - ^b_T = 0:0181(0:0007)degrees Celsius per annum (statistically highly significant).
- Comparisons with observational series:
  - NASA: rise 0:89  C between 1960 and 2014 ≡ 0:0165  C per year.
  - NCEI (NOAA): rise 1:07  C between 1960 and 2014 ≡ 0:0198  C per year.
  - Pooled estimate 0:0181  C per year lies between NASA and NCEI estimates.

### 3.2 Long-term impact on economic growth — methodology and main estimates
- Model and estimation:
  - Panel ARDL with p = 4 lags; dependent variable y_it = log real GDP per capita; climate regressors C_it = (T_it, P_it)′ with historical norms as m-year moving averages (baseline m = 30; robustness m = 20 and m = 40).
  - Positive and negative deviations separated: z+ = z I(z ≥ 0), z− = −z I(z < 0).
  - Estimators: FE and half-panel Jackknife FE (HPJ-FE) bias correction; long-run effects ρ computed via standard ARDL formula.
  - Sample: 1960ñ2014; N = 174 (some specs N = 165); max T = 50; avg T ≈ 38.36–38.98.
- Baseline long-run temperature deviation effects (HPJ-FE, m = 30, specification highlights):
  - b̂_ρ[(T_it − T′_{i,t-1})+] = −0.0577 (per 0.01 °C annual increase) [HPJ-FE reported in text].
  - b̂_ρ[(T_it − T′_{i,t-1})−] = −0.0505 (per 0.01 °C annual decrease) [HPJ-FE reported in text].
  - Specification 2 (temperature only): persistent 0.01 °C increase above historical norm reduces real GDP per capita growth by 0.0586 percentage points per annum (1% level); persistent 0.01 °C decrease reduces growth by 0.0520 percentage points per annum (5% level).
- Ranges across specifications/estimators:
  - b̂_ρ[(T_it − T′_{i,t-1})+] ranges from −0.0352 to −0.0692 across FE and HPJ-FE.
  - HPJ-FE estimates for (T_it − T′_{i,t-1})+ range from −0.0545 to −0.0692.
  - b̂_ρ[(T_it − T′_{i,t-1})−] ranges from −0.0432 to −0.0677; HPJ-FE range −0.0480 to −0.0677.
- Absolute-deviation specification (HPJ-FE, Table 3):
  - b̂_ρ[|T_it − T′_{i,t-1}|] (m = 20) = −0.0498 (std. err. 0.0191).
  - b̂_ρ[|T_it − T′_{i,t-1}|] (m = 30) = −0.0539 (std. err. 0.0183).
  - b̂_ρ[|T_it − T′_{i,t-1}|] (m = 40) = −0.0479 (std. err. 0.0176).
- Error-correction (speed of adjustment) (selected HPJ-FE estimates):
  - b̂_ε (Spec 1 HPJ-FE) = −0.6026*** (std. err. 0.0449).
  - Table 3 HPJ-FE b̂_ε ≈ −0.6038 to −0.6047 (std. err. ≈ 0.0448–0.0449).
- Precipitation:
  - Coefficients on precipitation deviations are generally not statistically significant in baseline specifications; absolute-deviation precipitation estimates not significant.
- Heterogeneity and symmetry:
  - Interaction terms with poor/hot country dummies not significant: cannot reject no-differential long-run effects for poor vs rich or hot vs cold.
  - Little evidence of asymmetry between positive and negative temperature deviations in cross-country long-run effects.

### 3.3 Robustness to historical norm choice and counterfactual implementation
- Historical norm windows (HPJ-FE absolute-deviation estimates, Specification 2):
  - b̂|Tit − T̃i,t−1| = −0.0504 (0.0191) for m = 20.
  - b̂|Tit − T̃i,t−1| = −0.0543 (0.0183) for m = 30.
  - b̂|Tit − T̃i,t−1| = −0.0486 (0.0176) for m = 40.
- Counterfactual framework:
  - Temperature path specification: Ti,T+j modeled with trend changes bTi;j; baseline historical norm m = 30.
  - Counterfactual horizon H set by Ti,T+H = Ti,2099 and Ti,T+1 = Ti,2015 ⇒ H = 85.
  - Scenarios: RCP 2.6 (Paris Agreement analog) average global temperature increase 0:01  C per annum; RCP 8.5 (no mitigation analog) average increase 0:04  C per annum.
  - Conservative baseline: assume volatility unchanged (σ1Ti = σ0Ti); alternative allows volatility to rise with mean (σ1Ti = (μ1Ti;j / μ0Ti) σ0Ti).
- Lag structure and multipliers (HPJ-FE ARDL, Table 6):
  - Estimated mean lag = 3:1943 years.
  - b̂ψ0 = −0.0038 (0.0021)
  - b̂ψ1 = −0.0056 (0.0029)
  - b̂ψ2 = −0.0084*** (0.0031)
  - b̂ψ3 = −0.0090*** (0.0026)
  - b̂ψ4 = −0.0060*** (0.0021)
  - Associated multipliers: b'1 = 0.2643*** (0.0500); b'2 = 0.0785*** (0.0266); b'3 = 0.0547** (0.0216); b'4 = −0.0016 (0.0327).
  - Sample: N = 174; No. Obs. = 6,674; maxT = 50; avgT = 38.36; minT = 2.

### Counterfactual results (percent losses in GDP per capita, Δih(di); PPP-weighted averages)
- World (Table 7; RCP 2.6 and RCP 8.5; m = 20, 30, 40):
  - RCP 2.6:
    - Year 2030 (h=16): −0.01 (m=20), −0.01 (m=30), −0.02 (m=40)
    - Year 2050 (h=36): 0.06 (m=20), 0.11 (m=30), 0.16 (m=40)
    - Year 2100 (h=86): 0.58 (m=20), 1.07 (m=30), 1.57 (m=40)
  - RCP 8.5:
    - Year 2030: 0.40 (m=20), 0.80 (m=30), 1.25 (m=40)
    - Year 2050: 1.39 (m=20), 2.51 (m=30), 3.67 (m=40)
    - Year 2100: 4.44 (m=20), 7.22 (m=30), 9.96 (m=40)
- Selected countries (RCP 8.5 vs RCP 2.6; m = 30 unless stated):
  - China: RCP 8.5: 2030 = 0.58; 2050 = 1.62; 2100 = 4.35. RCP 2.6: 2030 = −0.45; 2050 = −0.80; 2100 = 0.45.
  - European Union: RCP 8.5: 2030 = 0.50; 2050 = 1.53; 2100 = 4.66. RCP 2.6: 2030 = −0.08; 2050 = −0.13; 2100 = 0.09.
  - India: RCP 8.5: 2030 = 1.16; 2050 = 3.62; 2100 = 8.90. RCP 2.6: 2030 = 0.26; 2050 = 0.81; 2100 = 2.57.
  - Russia: RCP 8.5: 2030 = 1.03; 2050 = 3.08; 2100 = 8.93. RCP 2.6: 2030 = −0.14; 2050 = −0.34; 2100 = −0.71.
  - United States: RCP 8.5: 2030 = 1.20; 2050 = 3.77; 2100 = 10.52. RCP 2.6: 2030 = 0.20; 2050 = 0.60; 2100 = 1.88.
- Country-group summaries (m = 30):
  - Rich countries: RCP 8.5: 2030 = 0.84; 2050 = 2.67; 2100 = 7.76. RCP 2.6: 2030 = 0.05; 2050 = 0.23; 2100 = 1.09.
  - Poor countries: RCP 8.5: 2030 = 0.72; 2050 = 2.18; 2100 = 6.05. RCP 2.6: 2030 = −0.16; 2050 = −0.18; 2100 = 0.99.
  - Hot countries: RCP 8.5: 2030 = 0.76; 2050 = 2.37; 2100 = 6.65. RCP 2.6: 2030 = 0.00; 2050 = 0.15; 2100 = 1.11.
  - Cold countries: RCP 8.5: 2030 = 0.81; 2050 = 2.56; 2100 = 7.40. RCP 2.6: 2030 = −0.02; 2050 = 0.05; 2100 = 1.05.
- Cross-country heterogeneity (textual ranges for focal countries under RCP 8.5, m = 30):
  - 2030 losses vary between 0:50 and 1:20 percent; 2050 between 1:53 and 3:77 percent; 2100 between 4:35 and 10:52 percent (reported for focal countries).

### Role of adaptation (m) and climate variability (Table 8)
- Adaptation via m:
  - Shorter m (m = 20) → faster updating of norms → lower per-capita output losses.
  - Longer m (m = 40) → slower adaptation → higher losses.
- Allowing volatility to rise with mean (RCP 8.5; m = 30; Table 8 PPP-weighted averages):
  - World: 2030 = 2.02; 2050 = 5.18; 2100 = 13.11.
  - United States: 2030 = 2.66; 2050 = 6.81; 2100 = 17.19.
  - Implication: allowing volatility to rise roughly doubles global GDP per capita losses by 2100 (from 7.22 percent in Table 7, RCP 8.5, m = 30 to 13.11 percent in Table 8); U.S. losses ~70 percent higher when climate variability increases.

### Interpretation and policy implications (aggregate)
- Main empirical implications:
  - Persistent temperature deviations from historical norms have robust, statistically significant, negative long-run effects on per-capita GDP growth.
  - Precipitation deviations generally not significant cross-country but matter in within-country U.S. and sectoral analyses.
  - Results contrast with literature focused on short- to medium-term level effects concentrated in hot/poor countries: long-run growth effects appear across poor and rich, hot and cold alike.
- Policy takeaways:
  - Strong support for keeping Paris Agreement pledges or increasing their ambition to avoid substantial long-run output losses.
  - Adaptation reduces but is unlikely to fully offset long-run growth damages; mitigation plus adaptation required.
  - Methodological recommendation: use bias-corrected panel estimators (e.g., HPJ-FE) when N is large relative to T to avoid underestimating cumulative climate-change effects.

### 5.1 United States: state-level trends and macro effects (1963–2016)
- Temperature trends (48 states, 1963ñ2016):
  - Estimated average annual temperature increase across 48 U.S. states: 0.026 degrees Celsius (C) per year; ^b_T = 0.0260 (0.0007) (std. err. in brackets); significant at 1% level.
  - U.S. average per annum increase 0.026 > world average 0:018 per annum.
  - All 48 states experienced statistically significant increases; selected state annual trend estimates include Alabama 0.0212, Arizona 0.0318, California 0.0270, New Jersey 0.0343, Wisconsin 0.0307, etc.
- State-level long-run effects (panel ARDL, HPJ-FE; 1963ñ2016; N = 48; T = 48):
  - Long-run coefficients (selected; standard errors in parentheses) for (T_it − T̄_i,t−1)+ on Real GSP:
    - 20 Year MA: -0.0245*** (0.0081)
    - 30 Year MA: -0.0152** (0.0077)
    - 40 Year MA: -0.0074 (0.0073)
  - For (T_it − T̄_i,t−1)+ on Real GSP per Capita:
    - 20 Year MA: -0.0143** (0.0064)
    - 30 Year MA: -0.0073 (0.0061)
    - 40 Year MA: -0.0014 (0.0058)
  - For (T_it − T̄_i,t−1)− on Real GSP and Real GSP per Capita: large negative and often highly significant coefficients across m = 20,30,40 (examples: Real GSP 30-year MA = -0.0697*** (0.0166); Real GSP per Capita 30-year MA = -0.0454*** (0.0124)).
  - Precipitation deviations (above/below) on Real GSP and Real GSP per Capita: small coefficients but several statistically significant (e.g., b
(P_it − P̄_i,t−1)+ on Real GSP: -0.0001*** (0.0000)).
- Expanded outcomes (1976ñ2016): labour productivity and employment
  - b
(T_it − T̄_i,t−1)+ (30-year MA; selected):
    - Real GSP: -0.0379*** (0.0107)
    - Real GSP per Capita: -0.0273*** (0.0083)
    - Real GSP per Employed: -0.0190*** (0.0067)
    - Employment: -0.0135** (0.0062)
  - b
(T_it − T̄_i,t−1)− large negative effects across outcomes (e.g., 30-year MA Real GSP: -0.1951*** (0.0358); Real GSP per Capita: -0.1505*** (0.0259)).
  - Precipitation below-norm coefficients negative and often significant for multiple outcomes.
- Temporal evolution / adaptation evidence:
  - Rolling-window estimates show long-run coefficients become larger in absolute value over time (Figure 9): relationship between climate deviations and per capita GSP growth has strengthened rather than weakened.
  - Interpretation: some adaptation exists (U.S. estimates smaller than cross-country), but adaptation has been limited and may not keep pace with warming.

### 5.2 U.S. sectoral evidence (1963–2016)
- Purpose: identify channels and heterogeneity across 10 sectors using state-level sectoral output series.
- Sectoral long-run HPJ-FE results (Table 12; coefficients are long-run effects; standard errors in parentheses):
  - Effects of (T_it − T¯_i,t−1)+ (positive temperature deviations):
    - Agriculture: -0.0323* (0.0178)
    - Forestry: -0.0666*** (0.0163)
    - Finance: -0.0587*** (0.0204)
    - Construction: -0.0628*** (0.0209)
    - Manufacturing: -0.0531*** (0.0126)
    - Communication: 0.0730*** (0.0161)
    - Wholesale Trade: -0.0336*** (0.0110)
    - Retail Trade: 0.0182* (0.0107)
    - Mining, Transport: negative or mixed significance across specifications.
  - Effects of (T_it − T¯_i,t−1)− (negative temperature deviations):
    - Broadly large negative effects across Transport, Finance, Forestry, Mining, Construction, Manufacturing, Wholesale Trade, Retail Trade (examples: Construction -0.2365*** (0.0389); Manufacturing -0.1674*** (0.0237)).
  - Precipitation deviations:
    - Positive precipitation deviations: small negative coefficients significant in Agriculture, Transport, Forestry, Wholesale Trade.
    - Negative precipitation deviations: significant negative effects in Agriculture, Forestry, Construction, Manufacturing, Wholesale Trade.
  - Long-run intercepts (persistence b ) are negative and significant across sectors (e.g., Agriculture -1.8892*** (0.0843); Manufacturing -0.5787*** (0.0312)).
- Interpretation:
  - Broad-based impacts: each of the 10 sectors is affected by at least one climate variable deviation.
  - Agriculture: negatively affected by temperature above norm and by precipitation above/below norms; adaptation costly.
  - Manufacturing: adverse effects from deviations of all four climate variables.
  - Some sectors (Communication, Retail Trade, Mining, Government, Finance/Insurance/Real Estate) show positive or mixed coefficients for positive temperature deviations in specific estimates, reflecting sector-specific demand shifts or measurement features.
  - Channels: reduced labour productivity, health impacts, infrastructure damage, supply-chain disruptions, demand shifts.
- Variability and extremes:
  - Increased variability amplifies sectoral damages; persistent positive deviations and larger swings raise frequency of extreme events.

### Data construction and econometric considerations
- Cross-country climate and GDP:
  - Population-weighted climate data from Matsuura and Willmott (2015) 0.5 degree gridded monthly series and CIESIN (2016) gridded population with 2010 density.
  - Temperature in degrees Celsius (  C); precipitation in meters.
  - Real GDP per capita from World Bank WDI for 1960ñ2014; panel unbalanced: max T = 55, avg T ≈ 39, N = 174 countries.
- U.S. state-level data (1963ñ2016):
  - Real GSP series: deflate nominal GSP using state CPI; splice 1963ñ1977 deflated nominal series with real GSP from 1977 using annual growth rates.
  - Sectoral real output (1963ñ2016): reconcile SIC (1963–1997) and NAICS (1997 onward) via SIC–NAICS mapping; deflate nominal sectoral outputs using PPI (SIC mapped to NAICS PPIs) for 1963–1976; splice backwards the 1997–2016 NAICS real series using growth rates from deflated nominal SIC series (1963–1997).
  - Climate data: NOAA NCEI monthly state-level area-weighted series aggregated to annual averages; original units Fahrenheit and inches converted to degrees Celsius and millimeters; snow included as melted precipitation.
  - Employment: BLS annual state-level number of employed persons (16 years and over, civilian noninstitutional population).
- Econometric issues and solutions:
  - Temperature is trending in almost all countries; using raw temperature in growth regressions induces spurious trends.
  - Distinguish short-run weather from persistent climate deviations by using deviations from historical norms (m-year moving average).
  - FE estimators biased when regressors weakly exogenous and N≫T; HPJ-FE used to correct small-T (Nickell-type) bias and allow feedback from growth to climate variables.
  - Lag order p = 4 chosen to capture potentially long-lasting climate impacts and avoid data-mining.

### Aggregate conclusions and policy recommendations
- Core empirical conclusion:
  - Persistent deviations of temperature from historical norms (both increases and decreases) are associated with negative long-run per capita GDP growth effects; effects are economically meaningful and robust to estimator and norm-window choices.
- Counterfactuals and policy relevance:
  - RCP 8.5 analog (0:04  C per year): world real GDP per capita reduced by 7:22percent by 2100 (m = 30 baseline); United States loss 10.52 percent in year 2100.
  - Paris Agreement analog (0:01  C per year): world loss 1:07 percent by 2100; United States loss 1.88 percent in year 2100.
  - Allowing climate variability to rise with mean substantially increases projected losses (world 13.11 percent by 2100 under RCP 8.5, m = 30; U.S. 17.19 percent).
- Recommendations:
  - Strengthen mitigation ambition (keep or exceed Paris Agreement pledges) to avoid substantial long-run output losses.
  - Invest in adaptation to reduce negative growth effects, recognizing adaptation alone is unlikely to fully offset long-run damages.
  - Use bias-corrected panel estimators (e.g., HPJ-FE) for large-N panels with relatively small T when estimating long-run impacts and constructing counterfactuals.

*Italic: Source — wpiea2019215-print-pdf (IMF Working Paper, selected sections and tables reproduced verbatim).*

### 1. Introduction ........................................................................................................

### 1. Introduction

### Purpose and scope
- Introduces the study's focus on the long-term economic impacts of climate change using a multi-country stochastic growth framework.
- Establishes the empirical and counterfactual analyses that follow, including an application to an advanced economy (the United States).

### Key questions and analytical approach
- Frames the principal research questions on how deviations in climate variables affect per capita economic growth and welfare over the long run.
- Indicates the use of country-level historical climate and GDP data, stochastic growth modeling, and scenario analysis to quantify impacts.

### Organization of the paper
- Previews empirical estimation of climate–growth relationships, robustness checks, and counterfactual projections under alternative climate scenarios.
- Notes a dedicated empirical case study for the United States with sector-level analysis and appendices providing additional results and data documentation.

*Source: wpiea2019215-print-pdf - 1. Introduction (IMF Working Paper contents list provided)*

### 12. Long-Run Effects of Climate Change on the Output Growth of Various Sectors in

### 12. Long-Run Effects of Climate Change on the Output Growth of Various Sectors in the United States, 1963–2016

### Introduction and scope
- Study period and coverage:
  - Cross-country panel: 174 countries over the period 1960 to 2014.
  - United States case study: 48 states over the period 1963 to 2016.
- Main focus:
  - Long-term macroeconomic effects of persistent deviations of climate variables (temperature and precipitation) from their historical norms on labour productivity and real output per capita.
  - Distinguishes level effects (e.g., agricultural yields) from long-run growth effects via reduced investment and labour productivity.
- Methodological highlights:
  - Theoretical multi-country stochastic growth model linking climate deviations to labour productivity.
  - Empirical strategy uses deviations from historical norms and allows for nonlinearity and asymmetric effects.
  - Estimator: half-panel Jackknife FE (HPJ-FE) to address weak exogeneity and small-T bias of standard FE estimators.

### Key empirical findings (cross-country)
- Global temperature trends and significance:
  - Global average temperature has risen by 0:0181de-grees Celsius per year over 1960ñ2014.
  - Positive country-specific trend estimates in 169 out of 174 countries (97.1% of cases).
  - Statistically significant at the 5% level in 161 out of 169 countries with positive trends (95.3% of cases).
- Long-run growth effect of persistent temperature deviation:
  - If temperature rises (falls) above (below) its historical norm by 0:01  C annually, income growth will be lower by 0:0543 percentage points per year.
  - No significant evidence of asymmetry between positive and negative deviations in the cross-country estimates.
  - Climate variability (deviations from norms) matters in addition to absolute temperature levels.
- Heterogeneity and contrasts with prior literature:
  - Results show long-run negative growth effects across poor and rich, hot and cold countries alike—contrary to literature that finds adverse short- to medium-term effects concentrated in already hot/poor countries.
  - Cross-country counterfactuals imply substantially larger long-run losses under high-emission scenarios.

### Counterfactual projections and policy implications
- Counterfactuals compare cumulative income effects for 2015ñ2100 relative to baseline where temperature increases follow historical trend (1960ñ2014).
- Representative scenarios and outcomes:
  - RCP 8.5 analog (increase in average global temperature of 0:04  C per year): reduces world real GDP per capita by 7:22percent by 2100.
  - Paris Agreement analog (increase limited to 0:01  C per year): reduces world output loss to 1:07 percent by 2100.
  - United States:
    - Loss under RCP 8.5 in year 2100: 10.52 percent.
    - Loss under Paris Agreement in year 2100: 1.88 percent.
- Sensitivities and caveats:
  - Income losses would be significantly higher if country-specific climate variability rises commensurate with temperature increases.
  - The speed at which historical norms change (20-, 30-, or 40-year moving averages; adaptation speed) affects the magnitude of income losses.
  - Upper and lower bounds in projections illustrated using m= 20 and m= 40 (Figure notes).
- Policy takeaway:
  - Strong support for keeping Paris Agreement pledges or increasing their ambition to avoid substantial long-run output losses.
  - Adaptation can reduce negative long-run growth effects but is unlikely to completely offset them.

### Theoretical model (overview)
- Framework:
  - Multi-country stochastic growth model exteding Merton (1975), Brock and Mirman (1972), Binder and Pesaran (1999).
  - Labour productivity in country i: function of common technology factor t and country-specific climate deviations (Tit and Pit) from historical norms T̃i,t−1 and P̃i,t−1.
  - Climate affects productivity only when variables deviate from historical norms; positive and negative deviations allowed and modeled with threshold/asymmetric terms.
- Key equations and implications:
  - Productivity specification (log-linear with exponential climate penalty): allows decomposition so that expected per capita output growth E(Δyit) = ηi μΘ − 0′i E(Δxit) (Equation (14) in model summary).
  - Use of deviations from historical norms avoids importing the strong upward trend of Tit into growth regressions, preventing spurious trend-induced bias.
  - Model generates long-run negative effect on mean per capita growth when climate deviation parameters i are positive.

### Within-country (United States) results and sectoral channels
- Aggregate state-level effects:
  - If temperature increases by 0:01  C annually above its historical norm across U.S. states, average per-capita real GSP growth will be lower by 0:0273 percentage points per year.
  - This state-level per-capita impact is smaller than the cross-country per-capita estimates.
- Sectoral impacts:
  - Impact of climate change on sectoral output growth is broad-based: each of the 10 sectors considered is affected by at least one of the four climate variables.
  - Within-U.S. estimates tend to show asymmetry in positive vs. negative deviations of climate variables from historical norms (unlike cross-country symmetry).
- Interpretation:
  - Geographic heterogeneity within the U.S. (hot/wet vs. cold/dry states) enables testing whether responses differ regionally; evidence shows negative macro effects are present but adaptation/resilience in advanced economies has not fully offset macro-level impacts.

### Data, estimation and econometric considerations
- Data sources and construction:
  - Population-weighted climate data constructed from Matsuura and Willmott (2015) 0.5 degree gridded monthly series and CIESIN (2016) gridded population with 2010 population density.
  - Temperature measured in degrees Celsius (  C); precipitation in meters.
  - Real GDP per capita from World Bank World Development Indicators for 1960ñ2014.
  - Resulting panel: unbalanced, maximum T = 55, average T ≈ 39, N = 174 countries.
- Econometric challenges addressed:
  - Temperature is trending in almost all countries; including raw temperature as regressor induces undesirable trends in per capita growth regressions.
  - FE estimators can be biased when regressors are weakly exogenous and when N≫T; HPJ-FE estimator used to correct Nickell-type bias and allow for feedback from growth to climate variables.
  - Distinction between short-run weather effects and persistent long-run climate deviations emphasized; short-run correlations may render climate variables weakly exogenous.

*Source: IMF Working Paper (chapter/section text provided).*

### 3.1  Climate Change: Historical Patterns

### 3.1  Climate Change: Historical Patterns

### Regression specification and data
- Country-specific regressions estimated: T_it = a_Ti + b_Ti t + v_Ti,t; for i = 1,2,...,N = 174.
- T_it denotes the population-weighted average temperature of country i at year t ( C).
- b_Ti is the per annum average increase in land temperature for country i.
- Global measure defined by b_T = N^{-1} ∑_{i=1}^N b_Ti.
- Appendix A reports estimates of b_Ti over 1900–2014.

### Key empirical findings (1960–2014)
- Individual country estimates of b_Ti range from -0:0044 (Samoa) to 0:0390 (Afghanistan).
- 169 countries (97.1% of cases) have positive b_Ti estimates.
- 161 countries (95.3% of cases) have b_Ti estimates statistically significant at the 5% level.
- Five countries with non-positive estimates: Bangladesh, Bolivia, Cuba, Ecuador, and Samoa; none of these five estimates are statistically significant at the 5% level.
- Table 1 provides country-level b_Ti estimates and standard errors (selected examples from Table 1 preserved in the source).

### Longer horizon results (1900–2014)
- Country-specific b_Ti estimates over 1900–2014 range from -0:0008 (Greece) to 0:0190 (Haiti).
- 172 countries (98.9% of cases) have positive b_Ti estimates over 1900–2014.
- 156 countries (90.7% of cases) have estimates statistically significant at the 5% level.
- Only two countries with non-positive estimates over 1900–2014: Greece and Macedonia; these are not statistically significant.
- Conclusion: Temperature has been rising (T_it is trended) for almost all countries in the sample across both sample periods.

### Global pooled estimate and comparisons
- Using individual country estimates in Table 1, the average rise in global temperature over 1960–2014 is
  - ^b_T = 0:0181(0:0007)degrees Celsius per annum (statistically highly significant).
- NASA observations:
  - Global land temperature rose by 0:89  C between 1960 and 2014, equivalent to around 0:0165  C per year.
- NCEI (NOAA) observations:
  - Global land-surface air temperature rose by 1:07  C between 1960 and 2014, equivalent to around 0:0198  C per year.
- The pooled estimate of 0:0181  C per year lies between the NASA and NCEI estimates and benefits from a small standard error as a pooled estimate across many countries.

### External dataset comparisons and documented trends
- The country-specific estimates align with average increases published by:
  - Goddard Institute for Space Studies (GISS) at National Aeronautics and Space Administration (NASA).
  - National Centers for Environmental Information (NCEI) at the National Oceanic and Atmospheric Administration (NOAA).
- IPCC (2013) trend estimates of global land-surface air temperature ( C per decade) over 1951–2012:
  - CRUTEM4.1.1.0: 0.175 (0:037)
  - GHCNv3: 0.197 (0:031)
  - Berkeley Earth: 0.175 (0:029)
  - 90% confidence intervals shown in brackets in the source.

### Temperature series (NOAA and NASA) and land vs. ocean behavior
- NOAA and NASA series for 1960–2014 show upward trends in:
  - Global land-surface air and sea-surface water temperatures (left panel of Figure 4).
  - Global land temperatures (right panel of Figure 4).
- NOAA reported rise between 1960 and 2014:
  - Global land-surface air and sea-surface water temperatures: rise of 0:72  C (equivalent to 0:0134  C per year).
- NASA reported rise between 1960 and 2014:
  - Global land-surface air and sea-surface water temperatures: rise of 0:77  C (equivalent to 0:0143  C per year).
- Land-surface air temperature has risen more than sea-surface water temperature over this period because oceans have a larger effective heat capacity and lose more heat through evaporation.

*Source: 3.1 Climate Change: Historical Patterns (wpiea2019215-print-pdf)*

### 3.2  Long-Term Impact of Climate Change on Economic Growth

### 3.2  Long-Term Impact of Climate Change on Economic Growth

### Methodology
- Model:
  - Panel ARDL model estimated as equation (17) with p = 4 lags for all variables/countries.
  - Dependent variable y_it: log of real GDP per capita.
  - Climate variables C_it = (T_it, P_it)′; historical norms C′_{i,t-1} are m-year moving averages with baseline m = 30 (robustness with m = 20 and m = 40).
  - Positive and negative deviations separated: z+ = z I(z ≥ 0) and z− = −z I(z < 0).
- Estimators and identification:
  - Fixed effects (FE) estimator reported and corrected for small-T bias using the half-panel Jackknife FE (HPJ-FE) of Chudik et al. (2018).
  - Long-run effects ρ calculated as ρ = ψ−1 Σ_{`=0}^p φ` with ψ = 1 − Σ_{`=1}^p φ`.
  - HPJ-FE preferred due to N = 174 and average T ≈ 38 (small-T bias concerns).
- Sample and timing:
  - Sample period 1960–2014.
  - Panel dimensions: N = 174 (some specifications N = 165), max T = 50, avg T ≈ 38.36–38.98, min T = 2 (some specs min T = 8).
- Additional specifications:
  - Specification 2 drops precipitation variables.
  - Specifications 3 and 4 interact temperature deviations with dummies for poor countries and hot countries respectively (poor: PPP GDP per capita below global median in 1980; hot/cold: historical average temperature above/below global median).

### Main empirical findings
- Temperature deviations from historical norms have persistent negative long-run effects on per capita GDP growth; precipitation deviations generally not significant.
- Baseline (Specification 1) HPJ-FE estimates (Historical norms = 30-year moving average):
  - b̂_ρ[(T_it − T′_{i,t-1})+] = −0.0577 (per 0.01 °C annual increase) [HPJ-FE reported in text].
  - b̂_ρ[(T_it − T′_{i,t-1})−] = −0.0505 (per 0.01 °C annual decrease) [HPJ-FE reported in text].
- Specification 2 (temperature only, 30-year norm) HPJ-FE:
  - Persistent 0.01 °C increase above historical norm reduces real GDP per capita growth by 0.0586 percentage points per annum (statistically significant at the 1% level).
  - Persistent 0.01 °C decrease below historical norm reduces real GDP per capita growth by 0.0520 percentage points per annum (statistically significant at the 5% level).
- Range of estimates across specifications and estimators:
  - b̂_ρ[(T_it − T′_{i,t-1})+] ranges from −0.0352 to −0.0692 across FE and HPJ-FE estimates (text summary).
  - HPJ-FE estimates for (T_it − T′_{i,t-1})+ range from −0.0545 to −0.0692 (text summary).
  - b̂_ρ[(T_it − T′_{i,t-1})−] ranges from −0.0432 to −0.0677 across FE and HPJ-FE estimates (text summary).
  - HPJ-FE estimates for (T_it − T′_{i,t-1})− range from −0.0480 to −0.0677 (text summary).
- Alternative absolute-deviation specification (Table 3; HPJ-FE):
  - b̂_ρ[|T_it − T′_{i,t-1}|] (Historical norm = 20 years): −0.0498*** (std. err. 0.0191).
  - b̂_ρ[|T_it − T′_{i,t-1}|] (Historical norm = 30 years): −0.0539*** (std. err. 0.0183).
  - b̂_ρ[|T_it − T′_{i,t-1}|] (Historical norm = 40 years): −0.0479*** (std. err. 0.0176).
- Error-correction (speed of adjustment) estimates (selected):
  - b̂_ε (Spec 1 FE) = −0.6706*** (std. err. 0.0489).
  - b̂_ε (Spec 1 HPJ-FE) = −0.6026*** (std. err. 0.0449).
  - Table 3 HPJ-FE b̂_ε values consistently ≈ −0.6038 to −0.6047 (std. err. ≈ 0.0448–0.0449).
- Precipitation:
  - Coefficients on (P_it − P′_{i,t-1})+ and (P_it − P′_{i,t-1})− are not statistically significant in baseline; absolute-deviation precipitation estimates also not significant in Table 3.
- Heterogeneity tests:
  - Interactive terms with I(country is poor) and I(country is hot) are not statistically significant in Table 2 and Table 3: cannot reject the hypothesis of no differential long-run effects for poor versus rich or hot versus cold countries.
  - Examples of interactive-term estimates (Table 2, HPJ-FE): b̂_ρ[(T_it − T′_{i,t-1})+ × I(country is poor)] = −0.0150 (std. err. 0.0382) — not significant.
  - Examples of hot-country interactions (Table 2, HPJ-FE): b̂_ρ[(T_it − T′_{i,t-1})+ × I(country is hot)] = 0.0354 (std. err. 0.0434) — not significant.
- Symmetry:
  - Positive and negative temperature deviations have similar magnitudes and signs — little evidence of asymmetry in long-run effects.

### Robustness and sensitivity
- Lag order and estimator:
  - p = 4 chosen to capture possibly long-lasting climate impacts and to avoid data-mining via variable-specific lag selection.
  - HPJ-FE bias correction materially increases absolute magnitude of temperature coefficients relative to FE.
- Historical norm window:
  - Results robust using historical norms computed with m = 20, 30, and 40 years (Tables 3–5).
  - Table 3 (absolute deviations) shows consistently negative and statistically significant b̂_ρ[|T_it − T′_{i,t-1}|] across 20/30/40 year norms.
- Precipitation exclusion:
  - Dropping precipitation (Specification 2) does not alter main finding that persistent temperature deviations reduce long-run growth.

### Implications and policy recommendations
- Empirical implication:
  - Persistent deviations of temperature from historical norms (both increases and decreases) are associated with negative long-run per capita GDP growth effects.
  - The magnitude of long-run effects is economically meaningful and is larger when small-T bias is corrected (HPJ-FE).
- Comparison with literature:
  - Integrated assessment models (IAMs) often assume level (short-term) effects only; these results indicate long-run growth effects exist and may imply IAMs understate welfare costs of climate change.
  - Panel-data studies focusing on short-to-medium-term effects (level effects) may underestimate long-run costs.
- Policy recommendation:
  - Findings "call for a more forceful policy response to climate change" because welfare effects are likely underestimated if long-run growth impacts are ignored.
- Methodological recommendation:
  - Use bias-corrected panel estimators (e.g., HPJ-FE) when N is large relative to T to avoid underestimating cumulative climate-change effects in counterfactual analyses.

*Source: 3.2 Long-Term Impact of Climate Change on Economic Growth (wpiea2019215-print-pdf).*

### 3.3  Robustness to the Choice of Historical Norms

### 3.3  Robustness to the Choice of Historical Norms

### Robustness checks on historical norms
- Alternative constructions of historical climate norms (T̃i,t−1 and P̃i,t−1) were considered using moving averages of the past 20 (m = 20) and 40 (m = 40) years, in addition to m = 30.
- Precipitation deviations:
  - Estimated coefficients of precipitation variables b̂( (Pit − P̃i,t−1)+ ), b̂( (Pit − P̃i,t−1)− ), and b̂ |Pit − P̃i,t−1| are not statistically significant (Specification 1).
  - No statistically significant differences between "rich" and "poor" or "hot" and "cold" countries in the precipitation interactions (Specifications 3 and 4).
- Temperature deviations:
  - Estimated coefficients of deviations of temperature from its historical norm are statistically significant in all four specifications.
  - Focusing on Specification 2 with xit = |Tit − T̃i,t−1| and using HPJ-FE estimates (preferred model), the absolute temperature deviation coefficient is robust to alternative historical-norm constructions:
    - b̂|Tit − T̃i,t−1| = −0.0504 (0.0191) for m = 20
    - b̂|Tit − T̃i,t−1| = −0.0543 (0.0183) for m = 30
    - b̂|Tit − T̃i,t−1| = −0.0486 (0.0176) for m = 40
  - Standard errors are reported in brackets next to these estimates.

### Counterfactual analysis: methodology
- Empirical approach:
  - Use HPJ-FE estimates based on the ARDL specification (equations (17), (20), and (21)).
  - Climate regressor simplified to xit = |Tit − T̃i,t−1| with (L) = ∑1 j=0 ψj L^j.
- Counterfactual output change for country i over horizon h (equation (22)):
  - Δi,T+h = E[ yi,T+h | zi,T ; x1i,T+1:T+h ] − E[ yi,T+h | zi,T ; x0i,T+1:T+h ]
  - Cumulated impact (equation (22)): Δi,T+h = ∑h j=1 ∑h−j ℓ=0 ψℓ [ x1i,T+j − x0i,T+j ].
- Temperature path modeling (equation (23)):
  - Ti,T+j = aTi + bTi;j (T+j) + vTi,T+j for j = 1,2,...
  - Allow trend change bTi;j to vary over time; historical norm constructed over past m years.
- First-order approximation and distributional assumptions:
  - Temperature shocks vTi,T+j assumed serially uncorrelated Gaussian with zero mean and variance σ^2_Ti.
  - Derive expected positive and negative deviations and expected absolute deviations using truncated Normal expressions (equations (25)–(27)), defining functions g+Ti, g−Ti, and gTi that depend on m, bTi;j, and σTi.

### Counterfactual scenarios and implementation
- Baseline setup:
  - Set m = 30 for baseline.
  - Counterfactual trend change specification (equation (28)): b1Ti;j = Ti,T+j − Ti,T+j−1 = b0Ti + j di, for j = 1,...,H.
  - Average counterfactual trend b1Ti = H−1 ∑H j=1 b1Ti;j = (Ti,T+H − Ti,T)/H (equation (29)).
  - di chosen so average rise over counterfactual period equals hypothesised b1Ti; di = 2 (b1Ti − b0Ti)/(H + 1) (equation (30)).
  - Counterfactual horizon: set Ti,T+H = Ti,2099 and Ti,T+1 = Ti,2015 ⇒ H = 85.
- Scenarios:
  - RCP 2.6 (Paris Agreement): average global temperature increase 0.01 °C per annum.
  - RCP 8.5 (no mitigation): average global temperature increase 0.04 °C per annum.
- Assumptions on volatility:
  - Conservative baseline: assume postulated trend rise does not affect volatility, set σ1Ti = σ0Ti.
  - Alternative specification for climate variability effects: keep coefficient of variation unchanged and set σ1Ti = (μ1Ti;j / μ0Ti) σ0Ti (discussed in Table 8).

### Key empirical inputs and lag structure
- Use HPJ-FE estimates from ARDL with climate variable |Tit − T̃i,t−1|; estimates and standard errors used to construct ψ̂j (reported in Table 6).
- Estimated mean lag from ψ̂j for j = 0,1,2,...,20:
  - Mean lag = ∑1 j=1 j ψ̂j / ∑1 j=0 ψ̂j = 3:1943 years.

### Estimated coefficients from Table 6 (HPJ-FE ARDL results, 1960–2014)
- b̂ψ0 = −0.0038 (0.0021)
- b̂ψ1 = −0.0056 (0.0029)
- b̂ψ2 = −0.0084*** (0.0031)
- b̂ψ3 = −0.0090*** (0.0026)
- b̂ψ4 = −0.0060*** (0.0021)
- Associated multipliers (b'1..b'4):
  - b'1 = 0.2643*** (0.0500)
  - b'2 = 0.0785*** (0.0266)
  - b'3 = 0.0547** (0.0216)
  - b'4 = −0.0016 (0.0327)
- Sample:
  - No. of Countries (N) = 174
  - No. of Obs. (N×T) = 6,674
  - maxT = 50, avgT = 38.36, minT = 2
- Notes: yit is log real GDP per capita; xit = |Tit − T̃i,t−1|; T̃i,t−1 based on moving averages of past 30 years; coefficients estimated by half-panel jackknife FE (HPJ-FE). Asterisks indicate statistical significance at 1% (***), 5% (**), and 10% (*).

### Counterfactual results — aggregate and country-level projections
- Presentation:
  - Results reported as percent losses in GDP per capita Δih(di) for years 2030 (h=16), 2050 (h=36), and 2100 (h=86); PPP-GDP weighted averages across countries and groupings; m = 20, 30, 40 variations shown.
- World (PPP-GDP weighted averages, Table 7):
  - RCP 2.6:
    - Year 2030 (h=16): −0.01 (m=20), −0.01 (m=30), −0.02 (m=40)
    - Year 2050 (h=36): 0.06 (m=20), 0.11 (m=30), 0.16 (m=40)
    - Year 2100 (h=86): 0.58 (m=20), 1.07 (m=30), 1.57 (m=40)
  - RCP 8.5:
    - Year 2030: 0.40 (m=20), 0.80 (m=30), 1.25 (m=40)
    - Year 2050: 1.39 (m=20), 2.51 (m=30), 3.67 (m=40)
    - Year 2100: 4.44 (m=20), 7.22 (m=30), 9.96 (m=40)
- Selected countries (Table 7 highlights; RCP 8.5, m = 30 unless stated):
  - China:
    - RCP 8.5: 2030 = 0.58, 2050 = 1.62, 2100 = 4.35
    - RCP 2.6: 2030 = −0.45, 2050 = −0.80, 2100 = 0.45
  - European Union:
    - RCP 8.5: 2030 = 0.50, 2050 = 1.53, 2100 = 4.66
    - RCP 2.6: 2030 = −0.08, 2050 = −0.13, 2100 = 0.09
  - India:
    - RCP 8.5: 2030 = 1.16, 2050 = 3.62, 2100 = 8.90
    - RCP 2.6: 2030 = 0.26, 2050 = 0.81, 2100 = 2.57
  - Russia:
    - RCP 8.5: 2030 = 1.03, 2050 = 3.08, 2100 = 8.93
    - RCP 2.6: 2030 = −0.14, 2050 = −0.34, 2100 = −0.71
  - United States:
    - RCP 8.5: 2030 = 1.20, 2050 = 3.77, 2100 = 10.52
    - RCP 2.6: 2030 = 0.20, 2050 = 0.60, 2100 = 1.88
- Country-group summaries (Table 7, m = 30):
  - Rich countries:
    - RCP 8.5: 2030 = 0.84, 2050 = 2.67, 2100 = 7.76
    - RCP 2.6: 2030 = 0.05, 2050 = 0.23, 2100 = 1.09
  - Poor countries:
    - RCP 8.5: 2030 = 0.72, 2050 = 2.18, 2100 = 6.05
    - RCP 2.6: 2030 = −0.16, 2050 = −0.18, 2100 = 0.99
  - Hot countries:
    - RCP 8.5: 2030 = 0.76, 2050 = 2.37, 2100 = 6.65
    - RCP 2.6: 2030 = 0.00, 2050 = 0.15, 2100 = 1.11
  - Cold countries:
    - RCP 8.5: 2030 = 0.81, 2050 = 2.56, 2100 = 7.40
    - RCP 2.6: 2030 = −0.02, 2050 = 0.05, 2100 = 1.05
- Cross-country heterogeneity:
  - Under RCP 8.5 (m = 30), country losses in 2030 vary between 0:50 and 1:20 percent; in 2050 between 1:53 and 3:77 percent; in 2100 between 4:35 and 10:52 percent (textual range reported for focal countries).
  - Under RCP 2.6 (m = 30), world could benefit in year 2030 relative to reference trend; losses limited to 0:11 and 1:07 percent in year 2050 and 2100 respectively (as PPP-weighted averages reported above).

### Role of adaptation (m) and climate variability
- Adaptation measured via m (length of moving average norm):
  - Shorter m (m = 20) implies faster agents treat higher temperatures as new norm → lower per-capita output losses.
  - Longer m (m = 40) implies slower adaptation → significantly higher losses.
  - Table 7 exhibits losses under m = 20, 30, 40 demonstrating this pattern across world, countries, and groups.
- Climate variability (Table 8): allowing temperature increases to affect volatility (σ1Ti = (μ1Ti;j / μ0Ti) σ0Ti) under RCP 8.5 and m = 30:
  - PPP-GDP weighted average losses (Table 8):
    - World: 2030 = 2.02; 2050 = 5.18; 2100 = 13.11
    - China: 2030 = 0.78; 2050 = 1.99; 2100 = 5.02
    - European Union: 2030 = 1.45; 2050 = 3.71; 2100 = 9.37
    - India: 2030 = 2.62; 2050 = 6.70; 2100 = 16.92
    - Russia: 2030 = 2.00; 2050 = 5.13; 2100 = 12.94
    - United States: 2030 = 2.66; 2050 = 6.81; 2100 = 17.19
    - Rich countries: 2030 = 2.24; 2050 = 5.74; 2100 = 14.51
    - Poor countries: 2030 = 1.52; 2050 = 3.92; 2100 = 9.91
    - Hot countries: 2030 = 1.76; 2050 = 4.54; 2100 = 11.52
    - Cold countries: 2030 = 2.10; 2050 = 5.39; 2100 = 13.62
  - Implication: allowing volatility to rise with mean temperature roughly doubles global GDP per capita losses by 2100 (from 7.22 percent in Table 7, RCP 8.5, m = 30, to 13.11 percent in Table 8).
  - For the United States, losses are ~70 percent higher when climate variability increases compared to the baseline counterfactual.

### Interpretation and policy implications
- Main findings:
  - Temperature deviations from historical norms have robust and statistically significant long-run negative effects on per-capita GDP growth across countries.
  - Precipitation deviations are not statistically significant in this specification.
  - Cross-country heterogeneity is large; under no-mitigation (RCP 8.5) losses are substantial and widespread.
  - Abiding by the Paris Agreement (RCP 2.6) would substantially reduce economic losses; in some horizons (2030) the world could marginally benefit relative to the historical-trend baseline.
  - Faster adaptation (shorter m) reduces but does not eliminate losses; adaptation alone is unlikely to offset climate-change damages fully.
  - Increasing climate variability amplifies economic losses materially.
- Channels through which increased variability raises losses:
  - Reduced labor productivity.
  - Increased health problems (e.g., heat-related illness, drought-related shortages).
  - Damage to infrastructure (e.g., flooding, landslides).
  - Supply chain disruptions.
- Policy recommendation (implicit from results):
  - More forceful mitigation policies are needed to limit damages from climate change, alongside adaptation efforts to reduce but not fully offset those damages.

### Evidence from an advanced economy: United States (prelude)
- Motivation:
  - Cross-country averages can mask within-country heterogeneity; large countries (Brazil, China, India, Russia, United States) have substantial geographic and climate diversity.
  - The contiguous U.S. (48 states, 1963–2016) provides within-country heterogeneity and rich data to test whether "hot" or "dry" states respond differently than "cold" or "wet" states.
- Data and approach:
  - Use state-level climate variables and macro variables: Gross State Product (GSP), GSP per capita, GSP per employed, and output for ten economic sectors (Agriculture, Mining, Construction, Manufacturing, Transport/Communications/Public Utilities, Wholesale Trade, Retail Trade, Finance/Insurance/Real Estate, Services, Government).
- Question:
  - Do the cross-country findings (universal long-run relationship between climate change and growth regardless of rich/poor, hot/cold) hold within an advanced economy with high development and adaptation capacity?
- Notes:
  - Advanced economies may exhibit partial insulation because more activity occurs indoors, better diffusion of adaptive technology, more flexible labor markets, and greater ability to invest in adaptation.

*Source: IMF Working Paper (section 3.3 and related counterfactual analysis, tables and figures as reproduced in the supplied content).*

### 5.1  Long-Run Impact of Climate Change on U.S. Economic Growth

### 5.1  Long-Run Impact of Climate Change on U.S. Economic Growth

### Temperature trends across the 48 U.S. states (1963–2016)
- Estimated average annual temperature increase across the 48 U.S. states: 0.026 degrees Celsius (C) per year over 1963ñ2016; estimated coefficient ^
b
T
= 0.0260 (0.0007) with the standard error in brackets; statistically significant at the 1% level.
- The U.S. average per annum temperature increase of 0.026 was higher than the world average rise of 0.018 per annum.
- All 48 U.S. states experienced statistically significant increases in temperature over time. Selected state annual trend estimates ( ^
b
;i ) from Table 9:
  - Alabama 0.0212
  - Arizona 0.0318
  - California 0.0270
  - Colorado 0.0271
  - Connecticut 0.0316
  - Delaware 0.0355
  - Florida 0.0228
  - Georgia 0.0228
  - Idaho 0.0245
  - Illinois 0.0223
  - Indiana 0.0236
  - Iowa 0.0198
  - Kansas 0.0186
  - Kentucky 0.0250
  - Louisiana 0.0210
  - Maine 0.0288
  - Maryland 0.0299
  - Massachusetts 0.0311
  - Michigan 0.0285
  - Minnesota 0.0320
  - Mississippi 0.0205
  - Missouri 0.0179
  - Montana 0.0292
  - Nebraska 0.0222
  - Nevada 0.0273
  - New Hampshire 0.0299
  - New Jersey 0.0343
  - New Mexico 0.0300
  - New York 0.0308
  - North Carolina 0.0257
  - North Dakota 0.0263
  - Ohio 0.0263
  - Oklahoma 0.0171
  - Oregon 0.0198
  - Pennsylvania 0.0280
  - Rhode Island 0.0320
  - South Carolina 0.0250
  - South Dakota 0.0234
  - Tennessee 0.0234
  - Texas 0.0245
  - Utah 0.0291
  - Vermont 0.0318
  - Virginia 0.0266
  - Washington 0.0186
  - West Virginia 0.0268
  - Wisconsin 0.0307
  - Wyoming 0.0279
- Note: z indicates statistical significance at the 1% level.

### Long-run effects of climate deviations on state-level output (1963–2016)
- Method: panel ARDL model (HPJ-FE estimates) with climate variables constructed using 20, 30, and 40 year moving averages; 30-year MA treated as central estimate. Dependent variables: log real GSP and log real GSP per capita. Lag order p = 4. N = 48 states; T = 48 years; N × T = 2304.
- Long-run coefficient estimates from Table 10 (coefficients followed by standard errors in parentheses):
  - b
(T_it − T̄_i,t−1)+ on Real GSP:
    - 20 Year MA: -0.0245*** (0.0081)
    - 30 Year MA: -0.0152** (0.0077)
    - 40 Year MA: -0.0074 (0.0073)
  - b
(T_it − T̄_i,t−1)+ on Real GSP per Capita:
    - 20 Year MA: -0.0143** (0.0064)
    - 30 Year MA: -0.0073 (0.0061)
    - 40 Year MA: -0.0014 (0.0058)
  - b
(T_it − T̄_i,t−1)− on Real GSP:
    - 20 Year MA: -0.0672*** (0.0162)
    - 30 Year MA: -0.0697*** (0.0166)
    - 40 Year MA: -0.0485*** (0.0169)
  - b
(T_it − T̄_i,t−1)− on Real GSP per Capita:
    - 20 Year MA: -0.0444*** (0.0124)
    - 30 Year MA: -0.0454*** (0.0124)
    - 40 Year MA: -0.0275** (0.0127)
  - b
(P_it − P̄_i,t−1)+ on Real GSP (precipitation above norm):
    - 20 Year MA: -0.0001*** (0.0000)
    - 30 Year MA: -0.0001*** (0.0000)
    - 40 Year MA: -0.0001*** (0.0000)
  - b
(P_it − P̄_i,t−1)+ on Real GSP per Capita:
    - 20 Year MA: -0.0001*** (0.0000)
    - 30 Year MA: -0.0001*** (0.0000)
    - 40 Year MA: -0.0001*** (0.0000)
  - b
(P_it − P̄_i,t−1)− on Real GSP (precipitation below norm):
    - 20 Year MA: -0.0001** (0.0001)
    - 30 Year MA: -0.0001 (0.0000)
    - 40 Year MA: -0.0002*** (0.0001)
  - b
(P_it − P̄_i,t−1)− on Real GSP per Capita:
    - 20 Year MA: -0.0001 (0.0000)
    - 30 Year MA: -0.0001** (0.0000)
    - 40 Year MA: -0.0001** (0.0000)
  - Persistence parameter b:
    - 20 Year MA Real GSP: -0.7263*** (0.0491)
    - 20 Year MA Real GSP per Capita: -0.8896*** (0.0530)
    - 30 Year MA Real GSP: -0.7210*** (0.0494)
    - 30 Year MA Real GSP per Capita: -0.8875*** (0.0532)
    - 40 Year MA Real GSP: -0.7109*** (0.0495)
    - 40 Year MA Real GSP per Capita: -0.8734*** (0.0533)
- Interpretation points from text:
  - Deviations of temperature and precipitation from historical norms generally have negative and statistically significant long-run effects on state-level output growth for most specifications.
  - b
(T_it − T̄_i,t−1)+ (excess temperature above norm) was not statistically significant in three out of six specifications in Table 10.
  - Deviations of precipitation (above and below norms) affect state-level economic activity in the U.S.; significance likely due to within-country heterogeneity in precipitation patterns.

### Long-run effects including labour productivity and employment (1976–2016)
- Expanded outcome variables: (i) real GSP, (ii) real GSP per capita, (iii) real GSP per employed (labour productivity), (iv) employment. Sample period 1976ñ2016. N = 48; T = 36; N × T = 1728.
- Selected long-run coefficient estimates from Table 11 (coefficients followed by standard errors in parentheses):
  - b
(T_it − T̄_i,t−1)+:
    - Real GSP: -0.0391*** (0.0108)
    - Real GSP per Capita: -0.0279*** (0.0084)
    - Real GSP per Employed: -0.0188*** (0.0067)
    - Employment: -0.0136** (0.0065)
    - 30 Year MA Real GSP: -0.0379*** (0.0107)
    - 30 Year MA Real GSP per Capita: -0.0273*** (0.0083)
    - 30 Year MA Real GSP per Employed: -0.0190*** (0.0067)
    - 30 Year MA Employment: -0.0135** (0.0062)
    - 40 Year MA Real GSP: -0.0371*** (0.0105)
    - 40 Year MA Real GSP per Capita: -0.0271*** (0.0082)
    - 40 Year MA Real GSP per Employed: -0.0188*** (0.0065)
    - 40 Year MA Employment: -0.0125** (0.0062)
  - b
(T_it − T̄_i,t−1)−:
    - 20 Year MA Real GSP: -0.1554*** (0.0315)
    - 20 Year MA Real GSP per Capita: -0.1199*** (0.0233)
    - 20 Year MA Real GSP per Employed: -0.0604*** (0.0166)
    - 20 Year MA Employment: -0.0723*** (0.0183)
    - 30 Year MA Real GSP: -0.1951*** (0.0358)
    - 30 Year MA Real GSP per Capita: -0.1505*** (0.0259)
    - 30 Year MA Real GSP per Employed: -0.0779*** (0.0191)
    - 30 Year MA Employment: -0.0954*** (0.0205)
    - 40 Year MA Real GSP: -0.2089*** (0.0388)
    - 40 Year MA Real GSP per Capita: -0.1616*** (0.0283)
    - 40 Year MA Real GSP per Employed: -0.0847*** (0.0207)
    - 40 Year MA Employment: -0.1027*** (0.0221)
  - b
(P_it − P̄_i,t−1)+ (precipitation above norm):
    - Selected: -0.0001* (0.0001); -0.0001* (0.0001); -0.0000 (0.0000); -0.0000 (0.0000)
    - 30 Year MA: -0.0002** (0.0001); -0.0001*** (0.0000); -0.0000 (0.0000); -0.0001** (0.0000)
    - 40 Year MA: -0.0001** (0.0001); -0.0001** (0.0000); -0.0001** (0.0000); -0.0000 (0.0000)
  - b
(P_it − P̄_i,t−1)− (precipitation below norm):
    - 20 Year MA: -0.0003*** (0.0001); -0.0002*** (0.0001); -0.0001 (0.0000); -0.0002*** (0.0001)
    - 30 Year MA: -0.0003*** (0.0001); -0.0002*** (0.0001); -0.0001 (0.0000); -0.0003*** (0.0001)
    - 40 Year MA: -0.0004*** (0.0001); -0.0003*** (0.0001); -0.0001 (0.0000); -0.0003*** (0.0001)
  - Persistence parameter b (selected):
    - 20 Year MA Real GSP: -0.5092*** (0.0503)
    - 20 Year MA Real GSP per Capita: -0.6327*** (0.0529)
    - 20 Year MA Real GSP per Employed: -0.7219*** (0.0756)
    - 20 Year MA Employment: -0.4371*** (0.0445)
    - 30 Year MA Real GSP: -0.5016*** (0.0504)
    - 30 Year MA Real GSP per Capita: -0.6301*** (0.0530)
    - 30 Year MA Real GSP per Employed: -0.7035*** (0.0756)
    - 30 Year MA Employment: -0.4382*** (0.0444)
    - 40 Year MA Real GSP: -0.4975*** (0.0502)
    - 40 Year MA Real GSP per Capita: -0.6215*** (0.0530)
    - 40 Year MA Real GSP per Employed: -0.6942*** (0.0749)
    - 40 Year MA Employment: -0.4290*** (0.0445)
- Key interpretation from text:
  - Across specifications, b
(T_it − T̄_i,t−1)− and b
(P_it − P̄_i,t−1)− are negative and statistically significant at the 1% level for almost all outcome variables: when temperature and precipitation fall below historical norms, state-level economic activity suffers, employment declines, and labour productivity growth falls.
  - For the 1976ñ2016 sub-sample, b
(T_it − T̄_i,t−1)+ is negative and statistically significant for several dependent variables (real GSP, real GSP per capita, real GSP per employed, and employment), indicating excess temperature above norms also damages economic outcomes in these specifications.
  - The size of b
(T_it − T̄_i,t−1)+ estimates in the U.S. is smaller in absolute value than cross-country estimates, suggesting a higher degree of adaptation in the U.S., though effects remain non-negligible.

### Evidence on adaptation and temporal evolution
- Procedure: re-estimate model over successive rolling windows starting with 1963ñ2016 and dropping one starting year at a time down to 1983ñ2016; dependent variable: real GSP per capita growth.
- Finding: estimated coefficients (long-run effects) become larger in absolute value over time (Figure 9), i.e., the relationship between climate deviations and per capita GSP growth has strengthened rather than weakened over the sample windows examined.
- Interpretive points raised in text for why adaptation effects may appear limited:
  - Adaptation may be concentrated in certain sectors only.
  - Adaptation may not be keeping pace with the increasing pace of global temperature rise over the past 40 years.
  - Structural shifts in the U.S. economy may have increased exposure (shift of value added to more climate-exposed sectors).
  - Firms may underestimate likelihood or severity of future events and underinvest in adaptation; private sector surveys find few firms have taken sufficient steps to assess and manage climate risks.
  - Adaptation may be available but take-up limited by cost; firms tend to underinvest in adaptation owing to high cost.
- Overall assessment in text: there has been some adaptation in the U.S. (estimates in Tables 10 and 11 generally smaller than cross-country estimates), but evidence suggests adaptation has had limited impact so far in dampening the negative effects of climate change in the United States.

*Source: IMF working paper chapter "5.1 Long-Run Impact of Climate Change on U.S. Economic Growth" (data and tables as provided).*

### 5.2  Further Evidence from U.S. Sector Level Data

### 5.2 Further Evidence from U.S. Sector Level Data

### Short summary and channels of impact
- Adaptation and mitigation can occur via short-term resource reallocation and long-term investment in R&D, innovation, or structural shifts toward less vulnerable industries.
- Adaptation is relatively easier and more effective in some industries than others; agricultural adaptation (e.g., greenhouses, crop mix changes) tends to be costlier than simple measures (e.g., air conditioning for offices).
- Sectoral analysis helps identify channels: direct production losses (e.g., heat damage to crops), labor productivity losses (e.g., due to heat, rain, wind), supply-chain interruptions, reduced consumer spending (during unusually cold periods), and access disruptions (e.g., heavy rain restricting access to mining sites).

### Long-run sectoral effects (U.S. states, 1963–2016) — key estimates from Table 12
- Coefficients shown are long-run effects () from HPJ-FE estimates; standard errors in parentheses; statistical significance: 1% (***), 5% (**), 10% (*).
- Effects of positive temperature deviations (b ((T_it − T¯_i,t−1)+ )):
  - Agriculture: -0.0323* (0.0178)
  - Transport: -0.0030 (0.0400)
  - Finance: -0.0587*** (0.0204)
  - Forestry: -0.0666*** (0.0163)
  - Mining: -0.0261** (0.0112)
  - Construction: -0.0628*** (0.0209)
  - Manufacturing: -0.0531*** (0.0126)
  - Communication: 0.0730*** (0.0161)
  - Wholesale Trade: -0.0336*** (0.0110)
  - Retail Trade: 0.0182* (0.0107)
- Effects of negative temperature deviations (b ((T_it − T¯_i,t−1)− )):
  - Agriculture: -0.0184 (0.0315)
  - Transport: -0.1887*** (0.0728)
  - Finance: -0.1777*** (0.0400)
  - Forestry: -0.1136*** (0.0270)
  - Mining: -0.0633*** (0.0175)
  - Construction: -0.2365*** (0.0389)
  - Manufacturing: -0.1674*** (0.0237)
  - Communication: 0.0050 (0.0309)
  - Wholesale Trade: -0.1201*** (0.0252)
  - Retail Trade: -0.0446** (0.0199)
- Effects of positive precipitation deviations (b ((P_it − P¯_i,t−1)+ )):
  - Agriculture: -0.0003*** (0.0001)
  - Transport: -0.0006*** (0.0002)
  - Finance: -0.0002** (0.0001)
  - Forestry: -0.0002*** (0.0001)
  - Mining: 0.0001** (0.0001)
  - Construction: -0.0001 (0.0001)
  - Manufacturing: -0.0001 (0.0001)
  - Communication: -0.0000 (0.0001)
  - Wholesale Trade: -0.0001* (0.0001)
  - Retail Trade: 0.0000 (0.0001)
- Effects of negative precipitation deviations (b ((P_it − P¯_i,t−1)− )):
  - Agriculture: -0.0005*** (0.0001)
  - Transport: -0.0003 (0.0003)
  - Finance: -0.0002 (0.0001)
  - Forestry: -0.0003*** (0.0001)
  - Mining: -0.0001 (0.0001)
  - Construction: -0.0003** (0.0001)
  - Manufacturing: -0.0002*** (0.0001)
  - Communication: 0.0000 (0.0001)
  - Wholesale Trade: -0.0001* (0.0001)
  - Retail Trade: 0.0000 (0.0001)
- Long-run intercept / baseline (b ):
  - Agriculture: -1.8892*** (0.0843)
  - Transport: -0.8133*** (0.0459)
  - Finance: -0.6412*** (0.0331)
  - Forestry: -0.9599*** (0.0726)
  - Mining: -0.8536*** (0.0581)
  - Construction: -0.4830*** (0.0394)
  - Manufacturing: -0.5787*** (0.0312)
  - Communication: -0.5944*** (0.0546)
  - Wholesale Trade: -0.4135*** (0.0414)
  - Retail Trade: -0.3902*** (0.0458)
- Sample sizes:
  - No. of states (N): 48 for most sectors; Transport: 47
  - T: 48 for all sectors
  - N × T: typically 2304 (Transport 2256)

### Interpretation of sectoral results
- Broad-based impacts: each of the 10 sectors is affected by at least one of the four climate variables.
- Agriculture: negatively impacted by temperature above historical norm and by precipitation above and below the norm; adaptation costs are high in agriculture.
- Manufacturing: deviations of all four climate variables from historical norms adversely affect manufacturing output growth.
- Mining, construction, transport, retail trade, wholesale trade, services, and government: negatively affected by unusually cold days (reduced consumer spending, supply-chain interruptions, delayed construction).
- Communication and Retail Trade show notable positive coefficients for positive temperature deviations in specific estimates, but these may reflect sector-specific demand or measurement features.
- Government and Finance/Insurance/Real Estate show a positive and statistically significant b ((T_it − T¯_i,t−1)+ ) for some specifications (government services at 10% level), likely reflecting increased government relief spending and higher insurance premiums in response to climate events.

### Variability, extremes, and implications
- Rise in average temperatures and increased variability both matter: larger swings increase the frequency of extreme hot/cold and wet/dry conditions, amplifying damage.
- Persistent positive temperature deviations are likely given projected global warming; these could lead to permanent negative impacts on state-level output growth across most sectors (exceptions noted: mining, government, finance/insurance/real estate).
- Even if average precipitation or below-norm temperatures are less likely to persist, unprecedented increases in variability could produce sizable and long-lasting negative impacts.

### Robustness and broader conclusions (from related sections summarized here)
- Cross-country and U.S. within-country analyses show that deviations of temperature below historical norms and deviations of precipitation from historical norms are detrimental to long-run output growth.
- For deviations of temperature above historical norms, recent samples show negative estimates across most economic sectors except mining, government, and finance/insurance/real estate.
- Evidence suggests adaptation has reduced negative effects in certain sectors but has not fully offset them at the macro level.
- Key quantified long-run results (from the broader study):
  - If temperature deviates from its historical norm by 0:01°C annually, long-term income growth will be lower by 0:0543 percentage points per year.
  - Keeping global average temperature increase below 2 degrees Celsius above pre-industrial levels (Paris Agreement) will reduce global income loss to 1:07percent by 2100.
  - An increase in average global temperatures of 0:04°C (RCP 8.5 scenario in the study) reduces world real GDP per capita by 7:22percent by 2100.
  - Global per capita GDP losses under RCP 8.5 would almost double if country-specific climate variability rose commensurate with temperature increases.
- Policy implication: findings call for more forceful policy responses, including more ambitious mitigation and adaptation efforts, since adaptation alone is unlikely to fully offset long-run negative growth effects.

*Italic: Source — wpiea2019215-print-pdf, section 5.2 (tables and estimates reproduced verbatim from the chapter).*

### References

### References

### Climate impacts and adaptation studies
- Agrawala, S., M. Carraro, N. Kingsmill, E. Lanzi, M. Mullan, and G. Prudent-Richard (2011). Private Sector Engagement in Adaptation to Climate Change: Approaches to Managing Climate Risks. OECD Environment Working Papers No 39.
- Barreca, A., K. Clay, O. Deschenes, M. Greenstone, and J. S. Shapiro (2016). Adapting to Climate Change: The Remarkable Decline in the US Temperature-Mortality Relationship over the Twentieth Century. Journal of Political Economy 124(1), 105ñ159.
- Behrer, P. and J. Park (2017). Will We Adapt? Temperature Shocks, Labor and Adaptation to Climate Change. Working Paper.
- Burke, M. and K. Emerick (2016). Adaptation to Climate Change: Evidence from US Agriculture. American Economic Journal: Economic Policy 8(3), 106ñ140.
- Burke, M., S. M. Hsiang, and E. Miguel (2015). Global Non-Linear E§ect of Temperature on Economic Production. Nature 527, 235ñ239.
- Burke, M. and V. Tanutama (2019). Climatic Constraints on Aggregate Economic Output. Working Paper 25779, National Bureau of Economic Research Working Paper 25779.
- Heutel, G., N. Miller, and D. Molitor (2016). Adaptation and the Mortality E§ects of Temperature across US Climate Regions. NBER Working Paper No. 23271.
- IPCC, . (2013). Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge.
- IPCC, . (2014). Climate Change 2014: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, Volume 1. Cambridge University Press, Cambridge.
- Pacala, S. and R. Socolow (2004). Stabilization Wedges: Solving the Climate Problem for the Next 50 Years with Current Technologies. Science 305(5686), 968ñ972.
- Stern, N. (2007). The Economics of Climate Change: The Stern Review. Cambridge University Press, Cambridge.
- Tol, R. S. J. (2009). The Economic E§ects of Climate Change. Journal of Economic Perspectives 23(2), 29ñ51.
- Tol, R. S. J. (2014, May). Correction and Update: The Economic E§ects of Climate Change. Journal of Economic Perspectives 28(2), 221ñ26.
- Weitzman, M. L. (2009, February). On Modeling and Interpreting the Economics of Catastrophic Climate Change. The Review of Economics and Statistics 91(1), 1ñ19.

### Climate data, observations, and variability
- Arguez, A., I. Durre, S. Applequist, R. S. Vose, M. F. Squires, X. Yin, R. R. Heim, and T. W. Owen (2012). NOAAís 1981-2010 U.S. Climate Normals: An Overview. Bulletin of the American Meteorological Society 93(11), 1687ñ1697.
- Brown, P. T., W. Li, J. H. Jiang, and H. Su (2016). Unforced Surface Air Temperature Variability and Its Contrasting Relationship with the Anomalous TOA Energy Flux at Local and Global Spatial Scales. Journal of Climate 29(3), 925 ñ940.
- Matsuura, K. and C. J. Willmott (2015). Terrestrial Air Temperature and Precipitation: Monthly and Annual Time Series (1900 - 2014), v 4.01.
- Swain, D. L., B. Langenbrunner, J. D. Neelin, and A. Hall (2018). Increasing Precipitation Volatility in Twenty-Örst-century California. Nature Climate Change 8(5), 427ñ433.
- Vose, R. S., S. Applequist, M. Squires, I. Durre, M. J. Menne, C. N. Williams, C. Fenimore, K. Gleason, and D. Arndt (2014). Improved Historical Temperature and Precipitation Time Series for U.S. Climate Divisions. Journal of Applied Meteorology and Climatology 53(5), 1232ñ1251.
- CIESIN (2016). Gridded Population of the World, Version 4 (GPWv4): Population Density. Columbia University, Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC).
- NOAA and NASA observational series referenced in Figure A.1: "Global Land-Surface Air and Sea-Surface Water Temperatures" and "Global Land Temperatures, 1900--2014" (1960 standardized to zero).

### Empirical macroeconomic and sectoral studies of weather and temperature
- Bloesch, J. and F. Gourio (2015). The E§ect of Winter Weather on U.S. Economic Activity. Economic Perspectives, Federal Reserve Bank of Chicago 39(1), 1ñ20.
- Cachon, G. P., S. Gallino, and M. Olivares (2012). Severe Weather and Automobile Assembly Productivity. Columbia Business School Research Paper No. 12/37.
- Cashin, P., K. Mohaddes, and M. Raissi (2017). Fair Weather or Foul? The Macroeconomic E§ects of El NiÒo. Journal of International Economics 106, 37ñ54.
- Dell, M., B. F. Jones, and B. A. Olken (2009). Temperature and Income: Reconciling New Cross-Sectional and Panel Estimates. American Economic Review 99(2), 198ñ204.
- Dell, M., B. F. Jones, and B. A. Olken (2012). Temperature Shocks and Economic Growth: Evidence from the Last Half Century. American Economic Journal: Macroeconomics 4(3), 66ñ95.
- Dell, M., B. F. Jones, and B. A. Olken (2014). What Do We Learn from the Weather? The New Climate-Economy Literature. Journal of Economic Literature 52(3), 740ñ798.
- Deryugina, T. and S. M. Hsiang (2014). Does the Environment Still Matter? Daily Temperature and Income in the United States. NBER Working Paper No. 20750.
- Jones, B. F. and B. A. Olken (2010). Climate Shocks and Exports. American Economic Review 100(2), 454ñ59.
- Mejia, S., M. Mrkaic, N. Novta, E. Pugacheva, and P. Topalova (2018). The E§ects of Weather Shocks on Economic Activity: What are the Channels of Impact? IMF Working Paper WP/18/144.
- International Monetary Fund, . (2017). The E§ects of Weather Shocks on Economic Activity: How Can Low-income Countries Cope? World Economic Outlook Chapter 3, 117ñ183.

### Econometric methods, growth, and theoretical foundations
- Binder, M. and M. Pesaran (1999). Stochastic Growth Models and Their Econometric Implications. Journal of Economic Growth 4, 139ñ183.
- Brock, W. A. and L. J. Mirman (1972). Optimal Economic Growth and Uncertainty: The Discounted Case. Journal of Economic Theory 4(3), 479 ñ513.
- Donaldson, J. B. and R. Mehra (1983). Stochastic Growth with Correlated Production Shocks. Journal of Economic Theory 29(2), 282 ñ312.
- Marimon, R. (1989). Stochastic Turnpike Property and Stationary Equilibrium. Journal of Economic Theory 47(2), 282 ñ306.
- Merton, R. C. (1975). An Asymptotic Theory of Growth Under Uncertainty. The Review of Economic Studies 42(3), 375ñ393.
- Nickell, S. (1981). Biases in Dynamic Models with Fixed E§ects. Econometrica 49(6), 1417ñ1426.
- Nordhaus, W. D. (1992). The "DICE" Model: Background and Structure of a Dynamic Integrated Climate-Economy Model of the Economics of Global Warming. Technical Report 1009, Cowles Foundation Discussion Paper No. 1009.
- Nordhaus, W. D. (2006). Geography and Macroeconomics: New Data and New Findings. Proceedings of the National Academy of Sciences of the United States of America 103(10), 3510ñ3517.
- Nordhaus, W. D. (2017). Revisiting the Social Cost of Carbon. Proceedings of the National Academy of Sciences of the United States of America 114(7), 1518ñ1523.
- Pesaran, M. H. (1997). The Role of Economic Theory in Modelling the Long Run. Economic Journal 107, 178ñ191.
- Pesaran, M. H. and Y. Shin (1999). An Autoregressive Distributed Lag Modelling Approach to Cointegration Analysis. In S. Strom (Ed.), Econometrics and Economic Theory in 20th Century: The Ragnar Frisch Centennial Symposium, Chapter 11, pp. 371ñ413. Cambridge: Cambridge University Press.
- Pesaran, M. H. and R. Smith (1995). Estimating Long-run Relationships from Dynamic Heterogeneous Panels. Journal of Econometrics 68(1), 79ñ113.
- Chudik, A., K. Mohaddes, M. H. Pesaran, and M. Raissi (2013). Debt, Ináation and Growth: Robust Estimation of Long-Run E§ects in Dynamic Panel Data Models. Federal Reserve Bank of Dallas, Globalization and Monetary Policy Institute Working Paper No. 162.
- Chudik, A., K. Mohaddes, M. H. Pesaran, and M. Raissi (2016). Long-Run E§ects in Large Heterogeneous Panel Data Models with Cross-Sectionally Correlated Errors. In R. C. Hill, G. Gonzalez-Rivera, and T.-H. Lee (Eds.), Advances in Econometrics (Volume 36): Essays in Honor of Aman Ullah, Chapter 4, pp. 85ñ135. Emerald Publishing.
- Chudik, A., K. Mohaddes, M. H. Pesaran, and M. Raissi (2017). Is There a Debt-threshold E§ect on Output Growth? Review of Economics and Statistics 99(1), 135ñ150.
- Chudik, A., M. H. Pesaran, and J.-C. Yang (2018). Half-Panel Jackknife Fixed E§ects Estimation of Panels with Weakly Exogenous Regressors. Journal of Applied Econometrics 33(6), 816ñ836.
- Phillips, P. C. B., T. Leirvik, and T. Storelvmo (2019). Econometric Estimates of Earth‚ º A ¥ Zs Transient Climate Sensitivity. Journal of Econometrics, forthcoming.
- Hsiang, S. M. (2016). Climate Econometrics. Annual Review of Resource Economics 8(1).

### Other empirical and regional studies
- Gallup, J. L., J. D. Sachs, and A. D. Mellinger (1999). Geography and Economic Development. International Regional Science Review 22(2), 179ñ232.
- Jones, B. F. and B. A. Olken (2010). Climate Shocks and Exports. American Economic Review 100(2), 454ñ59.
- Kort, J. R. (2001). The North American Industry ClassiÖcation System in BEAís Economic Accounts. Survey of Current Business.
- Sachs, J. D. and A. M. Warner (1997). Sources of Slow Growth in African Economies. Journal of African Economies 6(3), 335ñ376.
- Obama, B. (2017). The Irreversible Momentum of Clean Energy. Science.

### Data and appendix references cited in annexes
- Table and figure notes reference NOAA, NASA, NCEI, GPWv4, and country-level estimates and projections shown in Appendix A (Tables A.1 and A.2, Figure A.1). Table A.1 reports OLS estimates of b_Ti from country-specific regressions T_it = a_Ti + b_Ti t + v_T;it with significance indicated at the 1% (***), 5% (**), and 10% (*) levels. Table A.2 reports percent loss in GDP per capita by 2030, 2050, and 2100 under the RCP 2.6 and RCP 8.5 scenarios; notes state h = 16, 36, and 86 (corresponding to the year 2030, 2050, and 2100, respectively) and m = 30, and losses are based on equation (31).
- Table B.1 lists Division (SIC) and Sector (NAICS) Classifications used in the U.S. data appendix.

*Source: References and appendices from wpiea2019215-print-pdf*

### 1963. We deáate the nominal GSP series using the consumer price index (CPI) for each state, and

### wpiea2019215-print-pdf - 1963. We deáate the nominal GSP series using the consumer price index (CPI) for each state, and

### Construction of real GSP series (1963–2016)
- Deáation approach: "We deáate the nominal GSP series using the consumer price index (CPI) for each state, and splice the resulting data over 1963ñ1977 with the real GSP from 1977 using annual growth rates, to construct a real GSP series for 1963ñ2016."
- Time coverage: 1963ñ2016.
- Splicing procedure: splice deáated nominal series (1963ñ1977) with real GSP from 1977 using annual growth rates.

### Sectoral output data and classification issues
- Source: BEA provides output by sector at the state level from 1963.
- Classification change: from 1963 to 1997 the Standard Industrial ClassiÖcation (SIC) consists of ten divisions; from 1997 onwards the North American Industry ClassiÖcation System (NAICS) gradually replaces the SIC, further branching the ten divisions into Öfteen sectors.
- Real sectoral output availability:
  - Only nominal sectoral output data (by SIC divisions) is available before 1977.
  - Real sectoral output is available in both SIC and NAICS classiÖcation in 1997.
- Two-step construction of real sectoral output series (1963ñ2016):
  - (i) reconciling SIC and NAICS classiÖcations (see Table B.1).
  - (ii) splicing the most recent real series (1997ñ2016) backwards using growth rates from the deáated nominal series (1963ñ1997).
- Data gaps/limitations noted:
  - "Agriculture, Forestry, Fishing & Hunting" and "Mining" data is not available for Rhode Island in 2016 and agricultural data in 2016 is also unavailable for Delaware.
  - The mining industry of Delaware is excluded from our sample due to multiple irregular missing entries.

### Deáation methodology for sectoral outputs
- Deflator used pre-1977: "We use BEAís producer price index (PPI) data to deáate the nominal industry outputs under SIC for the years 1963ñ1976."
- SIC–NAICS mapping and PPI:
  - "As the PPI data is constructed based on NAICS, we use the SIC-NAICS matching in Table B.1 for the PPI deáator."
  - Where multiple NAICS sectors match a SIC division: "we take a simple arithmetic average of the PPI of all matched NAICS sectors."
- Post-1997 approach:
  - "From 1997 onwards, real output by sector is available based on NAICS classiÖcation."
  - "We, therefore, aggregate the NAICS real output by industry to SIC divisions using our matching scheme, and splice these series backwards using the growth rates of real sectoral output under SIC in 1963ñ1997."
- Result: real output by sector and state for the period 1963 to 2016.

### Climate and employment data
- Climate data:
  - Source: NOAAís National Centres for Environmental Information (NCEI).
  - Frequency and aggregation: monthly state-level, area-weighted climate data from aggregates of climate readings across weather stations, adjusting for the distribution of stations and terrain.
  - Variables reported: monthly average temperature and precipitation.
  - Measurement units in raw NCEI data: temperature in degrees Fahrenheit and precipitation in inches.
  - Conversion performed: "We convert them into degrees Celsius and millimeters, respectively (to match our cross-country data)."
  - Annualization: "The monthly averages in each year within the sample period are then used to obtain annual averages."
  - Note on precipitation measurement: "Snow is included as melted precipitation in rain gauges under NOAA methodology."
- Employment data:
  - Source: Bureau of Labor Statistics (BLS).
  - Variable: annual, state-level number of employed persons.
  - Coverage definition: encompasses "persons 16 years and over in the civilian noninstitutional population" under a wide range of employment conditions.

*Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019215-print-pdf.pdf*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019215-print-pdf.pdf_