Structural Change and Economic Dynamics 66 (2023) 383–392

Contents lists available at ScienceDirect

Structural Change and Economic Dynamics

journal homepage: www.elsevier.com/locate/strueco

Temperature and GDP: A review of climate econometrics analysis

Jun-Jie Chang a, c, e, Zhifu Mi d, *, Yi-Ming Wei a, b, c, **
a
Center for Energy and Environmental Policy Research, Beijing Institute of Technology, Beijing 100081, China
b
Beijing Key Lab of Energy Economics and Environmental Management, Beijing 100081, China
c
School of Management and Economics, Beijing Institute of Technology, Beijing 100081, China
d
The Bartlett School of Sustainable Construction, University College London, London WC1E 7HB, UK
e
School of Geography and the Environment, University of Oxford, Oxford OX1 3QY, UK

A R T I C L E I N F O A B S T R A C T

Keywords: Climate econometric analysis of the relationship between temperature and gross domestic product (GDP) is
Climate change increasingly being used to evaluate climate risks and understand economic impacts caused by climate change.
Temperature We review the literature on growth and level effects (i.e., temperature rise respectively affects the growth and
GDP
level of economic output), the setting of temperature variables’ forms and functional forms, and the inherent
Climate econometrics
Integrated assessment model
model specification of climate econometrics. Additionally, we introduce an approach for combining empirical
findings with climate change integrated assessment models (IAMs) to improve damage modelling. Our findings
show that estimates of damage through growth effects are generally much larger than those through level effects.
Diverse impact mechanisms and adaptation effects can be revealed by changing the time resolution of temper­
ature variables, introducing non-linearity into econometrics functions, and specifying temperature deviation.
Combing the cross-sectional and panel model would enable us to examine the economic impacts at different
future times.

1. Introduction strategy to investigate the economic effects of climate change (Hsiang,

2016; Castle and Hendry, 2022). The approach parameterises impacts in
Climate change has already exerted impacts on the natural and socio- a reduced-form empirical model capturing the connection between
economic system and will continue to do so for centuries, and a growing climate and outcome; for example, temperature and GDP which are
body of research is focusing on how it will affect societies and economies estimated referencing historical data to directly reflect real-world set­
(Stern, 2013; Carleton and Hsiang, 2016; Auffhammer, 2018; Carril­ tings. Climate econometrics can expand the research scope of previous
ho-Nunes and Catalão-Lopes, 2022; Fabozzi et al., 2022). Further literature on climate change impacts that have not been previously
research into the economic consequences of climate change is critical for considered; for example, the climate effects on labour productivity,
practical policymaking as it can clarify how to trade off the benefits of capital and total factor productivity (Fankhauser and Tol, 2005; Lecocq
mitigating greenhouse gas emissions(GHGs) with the costs and value of and Shalizi, 2007; Zivin and Neidell, 2010; Zhang et al., 2018; Henseler
climate change mitigation spending compared to other social in­ and Schumacher, 2019).
vestments (Wei et al., 2018; Nordhaus, 2019, 2020; Chakraborty and Research applying climate econometrics to explore climate impacts
Mazzanti, 2021; Chen et al., 2023). can be divided into two categories of sectoral studies and top-down,
Previous research exploring the impact of climate change on the aggregate macro-economic output studies (Hsiang, 2016; Rising et al.,
economy has relied upon process-driven models which replicated 2022). Sectoral studies primarily examine agriculture, energy demand,
distinct physical and market equilibrium processes to simulate the mortality, crime and labour productivity (Hsiang et al., 2013; Dell et al.,
consequences of climate factors on a sector or aggregate output in the 2014; Carleton and Hsiang, 2016; Hsiang et al., 2017; Auffhammer,
socio-economic system (Tol, 2021; Rising et al., 2022). Instead, 2018). Nevertheless, these studies usually measure the socio-economic
benefitting from an increasingly data-rich environment, there is a impacts by physical units such as crop yield, deaths etc., even while
growing trend in recent research to employ the climate econometrics not always (Diaz and Moore, 2017; Sarofim et al., 2019). While, a set of

* Corresponding author at: The Bartlett School of Sustainable Construction, University College London, London WC1E 7HB, UK.
E-mail addresses: [email protected] (Z. Mi), [email protected] (Y.-M. Wei).
**
Co-corresponding author.

https://doi.org/10.1016/j.strueco.2023.05.009
Received 16 March 2023; Received in revised form 21 May 2023; Accepted 25 May 2023
Available online 26 May 2023
0954-349X/© 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
J.-J. Chang et al. Structural Change and Economic Dynamics 66 (2023) 383–392

research has focused on identifying the impacts of climate change on only use different types of data which vary in time and space but can also
macro-economic output from a top-down perspective by estimating the control for the unobservable omitted variables. Moreover, different
relationship between temperature and GDP, which became a critical specifications help to distinguish short- and long-run climate impacts,
topic in climate econometrics research (Dell et al., 2009, 2012; Burke and potential adaptation effects can also be identified using appropriate
et al., 2015; Kalkuhl and Wenz, 2020; Colacito et al., 2021; Newell et al., specifications. Recent hybrid models even can simultaneously consider a
2021). There are two highlights for this strand of research. One is the use series of short- to long-term effects (Kalkuhl and Wenz, 2020). In addi­
of GDP to measure economic impact as a monetary unit for measuring tion, given that current damage functions have been criticised regarding
aggregate macro-economic output. This offers a means to quantify the paucity of empirical evidence and examining the temperature–GDP
human welfare under the assumption that the prices of the commodities relationship can directly provide a monetised indicator for measuring
and services on the market accurately capture the costs associated with economic impact (Pindyck, 2013; Stern, 2013; Stoerk et al., 2018; Stern
their production and use. The second is the top-down perspective, which et al., 2022), estimating the relationship between temperature and GDP
partially avoids the need for explicit representation of impact on indi­ using the climate econometrics model can improve the accuracy of the
vidual sectors and associated criticisms of the omitting of impact types, IAMs that are commonly used to determine the optimal climate policy
interaction effects and adaptation (Dell et al., 2012; Burke et al., 2015; by examining the costs and benefits trade-offs of mitigation.
Moore and Diaz, 2015; Lemoine and Kapnick, 2016). By reviewing the climate econometrics literature on the temperature-
A series of key problems and research challenges for modelling the GDP nexus, this study first identifies the critical differences between the
temperature–GDP relationship using climate econometrics have been growth and level effects of temperature and the determinants of them.
frequently discussed. We summarise them into four relevant research Also, it identifies the debate in current research on both effects and how
strands. First, does temperature affect GDP level or growth? There are state-of-the-art studies combine both effects to investigate the temper­
numerous discussions and disagreements in the research regarding ature effects on GDP. Secondly, this research summarises the charac­
whether temperature impacts GDP level or growth. Findings in recent teristics of impact functions in both linear and non-linear forms and their
literature have demonstrated that the damage estimated through application scenarios, and further identifies the shortcomings of the
modelling level effects is underestimated in contrast to models of growth linear form and how it can be improved by the non-linear form. Thirdly,
effects since the temperature impacts on GDP growth accumulate over this study categorises the forms of temperature variables in existing
time but level effects do not (Dell et al., 2012, 2014; Burke et al., 2015, studies by temporal resolution and shows how current research uses
2018; Colacito et al., 2021; Newell et al., 2021). The literature related to different types of temperature variables to investigate the impact
growth effects has also involved considerable disagreement over the mechanisms between temperatures and GDP. It also identifies the im­
details, with substantial variance in resulting estimates (Dell et al., 2012; provements brought about by the introduction of the form of tempera­
Burke et al., 2015; Carleton and Hsiang, 2016; Hsiang et al., 2017; Pretis ture deviation used in recent research. Fourth, this study reviews the
et al., 2018; Henseler and Schumacher, 2019; Colacito et al., 2021; development of climate econometric models and sorts out the charac­
Kahn et al., 2021; Newell et al., 2021). teristics of different model specifications of the temperature-GDP rela­
Second, how is the functional form of temperature–GDP nexus tionship, comparing their advantages and disadvantages. Next, each
specified? The specification of the functional form of temperature–GDP specification of the climate econometric model is matched with its
varies in previous climate econometrics literature (Dell et al., 2014; appropriate application scenario. Fifth, the study identifies the channels
Newell et al., 2021). The choice of linear or non-linear forms can through which future research can combine climate econometrics with
dramatically influence estimations of temperature impacts. For instance, IAMs. Finally, it illuminates the limitations of existing studies and
linking the aggregate per capita economic growth at the country level to summarize a series of future potential directions.
a linear function with the independent variable of temperature and The literature review contributes significantly to multiple facets
precipitation, Dell et al. (2012) found that temperature shocks only including that firstly, it reveals the lack of consensus in identifying and
negatively impact growth in developing countries. In contrast, Burke interpreting growth and level effects among existing empirical studies
et al. (2015) used a non-linear form to identify a ‘turning point’ of the and suggests that future research should incorporate multiple model
annual average temperature of 13 ◦ C for GDP growth, demonstrating specifications of climate econometrics into a single model to distinguish
that countries with lower temperatures would either have boosted between the characteristics of growth and level effects in the short,
growth or be harmed when temperatures rise. medium, and long term. Secondly, it finds that an increasing number of
Third, how are temperature variables specified? Several studies have studies are utilizing relevant variables that can capture climate vari­
used different temperature variables to explore the impact of tempera­ ability to describe climate change, rather than solely relying on absolute
ture on GDP under different time resolutions, revealing the impact values of climate or weather factors. Future research should focus on
mechanism of non-linearity and identifying the adaptation effects in developing climate econometrics models that incorporate or improve
long-term climate change, among other considerations. For example, these variable forms in order to avoid potential estimation bias caused
studies have used different time resolutions, including annual (Burke by spurious regressions, ignoring the adaptation effects and so on.
et al., 2015; Diffenbaugh and Burke, 2019; Chang et al., 2020), monthly Thirdly, the study finds that projecting future climate change impacts on
(Pretis et al., 2018) and seasonal temperature variables (Chen and Yang, GDP by extrapolating the existing empirical temperature-GDP rela­
2019; Yuan et al., 2020; Colacito et al., 2021). Some studies have tionship may result in estimation bias, as these relationships do not fully
investigated non-linear effects using temperature bins (Deryugina and reflect potential future adaptation effects, leading to biased projections
Hsiang, 2014; Chen and Yang, 2019). In addition, recent research has of future impacts. To improve this, we suggest that future studies should
revealed the inherent limitations of temperature on GDP when exam­ use more comprehensive data and integrate long-difference economet­
ined across a variety of non-climate factors using a daily temperature rics models to identify climate adaptation patterns at a regional level
variable (Kotz et al., 2021), and explored the implicit adaptations over an extended period of time. Fourthly, we note that a substantial
required to mitigate the long-run impacts of temperature on GDP by body of empirical research has already investigated the impact of tem­
constructing a variable of temperature deviations based on respective perature on GDP and its determinants. Therefore, we recommend that
historical norms (Kahn et al., 2021). future studies incorporate these existing empirical foundations into the
Fourth, how is the climate econometrics model specified? It is crucial damage functions of IAMs. Finally, we also have identified a number of
to investigate the model specifications for climate econometrics as they other research directions that need to be further explored, including
have rapidly developed from cross-sectional approaches to panel models identifying the potential empirical nexus between climate change and
over the past two decades (Dell et al., 2009, 2014; Hsiang, 2016; Kolstad non-market sectors such as biodiversity, health and tipping points as
and Moore, 2020; Castle and Hendry, 2022). More advanced models not well as examining the underlying temperature effects in the trading

384
J.-J. Chang et al. Structural Change and Economic Dynamics 66 (2023) 383–392

network. labour productivity, population and exponentiated temperature. The

authors suggested that changing temperatures may have an impact on
2. Does temperature affect GDP level or growth? investment, which could impact productivity growth. According to a
growing number of academics, climate change might permanently harm
Using growth or level effects to model the temperature-GDP rela­ capital stocks and productivity, which will likely have a long-term in­
tionship has been a controversial topic. The level effect assumes that fluence on GDP growth (Dell et al., 2012; Pindyck, 2013; Stern, 2013;
temperature rises only temporarily affect the level of economic output, Hsiang and Jina, 2014; Moyer et al., 2014; Burke et al., 2015; Moore and
and once the temperature returns to the pre-warming value, the level Diaz, 2015, 2018). This research has suggested that global warming
effect disappears. In contrast, the growth effect assumes that tempera­ adversely affects developing countries’ GDP growth (Dell et al., 2012;
ture rise permanently affects economic output growth, and this effect Henseler and Schumacher, 2019; Letta and Tol, 2019). Several studies
will not disappear even if the temperature returns to the pre-warming have determined that more temperate countries face greater risks in
value but will compound over time instead. The premises of these two GDP growth (Burke et al., 2015; Burke and Tanutama, 2019; Kumar and
effects are presented in Fig. 1, considering a period of rising tempera­ Khanna, 2019; Kalkuhl and Wenz, 2020). Similarly, some research has
tures relative to historical norms (ΔT > 0), which is eventually (but not focused on a sub-national scale; for example, Yuan et al. (2020) and
always) reversed. The red line indicates that rising temperatures could Chang et al. (2020) found that southern China, which is a relatively
damage the GDP level; however, once temperatures have stabilised, the warmer region, will suffer more negative impacts on economic growth.
GDP can restart growth at the rate prior to the temperature rise. There is minimal consensus on the identification and interpretation
Although this generates a lower GDP level relative to the initial trajec­ of growth and level effects in existing empirical studies. For example,
tory, the trend rate either remains unchanged (black line), or the dam­ Dell et al. (2012) interpreted the coefficients of the lagged terms of
age would be so significant that it lowers the trend rate of GDP growth temperature as indicating growth effects, assuming that if the summa­
(blue line). If GDP growth is impacted by temperature changes, at some tion of the coefficients (marginal temperature effects) of all temperature
future time, the output will depend on the trajectory of temperature and lags significantly differs from zero and the signs of different terms are
output up to that time rather than just the temperature at that time reversed indicate both level and growth effects of temperature on GDP,
(Pindyck, 2013; Batten, 2018). As growth effect can cause permanent and the results are consistent with this assumption. Similar to Dell’s
damage to the economy rather than just reducing output in a given year assumption, Burke et al. (2015) regressed distributed lag models
as level effect does, the damages compound over time for growth effect including 1–5 lag terms of linear and quadratic temperature variables to
models, resulting in projected long-term economic damage of global investigate temperature effects. However, the cumulative marginal
warming that is orders of magnitude higher than conventional estimates temperature effects on growth indicate that the sign reversed when more
(Dell et al., 2014; Burke et al., 2015; Pretis et al., 2018; Piontek et al., lags were included, but the increasing uncertainty caused these cumu­
2021). lative effects to have no statistical significance; therefore, we can’t reject
There is a vast body of literature concerning the micro-foundations the hypothesis of this representing a genuine growth effect, nor can we
that link temperature to various economic outcomes. For example, reject the hypothesis of its representing a temporary level effect.
temperature impacts macro-economic output through various channels As reflected in the examples above, previous research has not yet
including energy (Auffhammer and Aroonruengsawat, 2011; Davis and fully determined whether climate change has only a level or a growth
Gertler, 2015), total factor productivity (TFP) (Zhang et al., 2018), effect; however, the two effects may simultaneously exist and could
capital (Zhang et al., 2018), agriculture (Schlenker et al., 2005; manifest at different future timing. To further explore this issue, Kalkuhl
Schlenker and Roberts, 2009), industrial production (Hsiang, 2010; and Wenz (2020) divided the climate effects on economic growth into
Chen and Yang, 2019), labor force (Zivin and Neidell, 2010), and capital three channels according to time scales, which included (i) the impacts
(Zhang et al., 2018), which represent widely-accepted determinants of of a temperature change on production in the short-run, (ii) the tem­
GDP. Some econometrics models directly aggregate related temperature porary effect on the growth rate to bring it into line with the economy’s
to GDP levels (Dell et al., 2009; Hsiang, 2010; Deryugina and Hsiang, long-run growth rate and (iii) the long-run growth rate of productivity.
2014). In the model specification, the coefficients of a linear term in climate
However, a series of solid theoretical bases have been proposed, change that solely considers contemporary and one-year-lag changes in
reasoning that GDP growth rate, not simply GDP level, might be weather indicate the immediate effects, the coefficients of a model in
impacted by climate change. The early theoretical model developed by linear and quadratic contemporaneous weather terms referencing Burke
Dell et al. (2012) characterises production as a multiplicative function of et al. (2015) indicate that there exists the short-run or long-run growth
effects and the coefficients of the interaction term and linear term
reflecting weather conditions consider all impact channels.

3. How is the functional form of temperature–GDP specified?

Functional form specification is among the key concerns regarding

which the temperature–GDP literature varies because there are no re­
strictions on particular, estimable, structural links between temperature
and GDP in the present paradigm (Dell et al., 2014; Burke et al., 2015;
Hsiang, 2016; Schlenker and Auffhammer, 2018; Kalkuhl and Wenz,
2020). Choosing a linear or non-linear model might significantly impact
the extent of damage estimations.

3.1. Linear form

Fig. 1. The GDP pathways under different temperature effects.
Note: The black line corresponds to the baseline pathway which assumes a The linear model is constructed assuming a linear relationship be­
world without temperature warming. The red line corresponds to the pathway tween a series of weather variables, including temperature, time trends
in which temperature change shocks GDP via the level effect. The blue line and an explained variable, which is generally the GDP level or the first
corresponds to the pathway in which temperature change affects GDP via the difference of the logarithm of GDP per capita. The model could include
growth effect. some fixed effects (FEs) but usually includes an error term. Numerous

385
J.-J. Chang et al. Structural Change and Economic Dynamics 66 (2023) 383–392

studies have examined the nexus between climate change and the 4. How are temperature variables specified?
economy using linear model specifications. For example, using a linear
model, Hsiang (2010) determined the statistically and economically 4.1. Annual average temperature
significant impacts of annual average temperature on sectoral as well as
total production in the Caribbean and Central America. Dell et al. (2012, Another important dimension is the specification of the forms of
2014) used a distributed lag linear model to explore the weather variables, particularly those for temperature. Average annual
temperature-economic growth relationship, finding that only devel­ temperature has a considerable impact on economic growth, according
oping country growth will be damaged by positive temperature shocks. to recent climate econometrics calculations examining spatiotemporal
However, many micro-foundation studies from a sectoral perspective heterogeneity in climate variables and growth. For example, Dell et al.
have demonstrated that temperature will have both negative and posi­ (2012) examined historical variations in annual temperature within
tive effects (Schlenker and Roberts, 2009; Auffhammer and Mansur, countries to determine the impacts on overall economic output, deter­
2014; Deryugina and Hsiang, 2014; Graff Zivin and Neidell, 2014; mining that higher temperatures significantly reduced economic growth
Hsiang et al., 2017; Graff Zivin et al., 2018; Zhang et al., 2018; Chen and in developing countries. Burke et al. (2015) found global economic
Yang, 2019). production to be non-linearly correlated with annual temperature across
In addition, the estimation procedure for obtaining temperature ef­ 166 countries, reaching a maximum of 13 ◦ C and sharply dropping
fects in a linear model has traditionally been based on temperature de­ above that. Furthermore, Kalkuhl and Wenz (2020) employed annual
viation in a certain year from each region’s average value, and the panel models based on sub-national data sets, revealing robust evidence
temperature effect is finally estimated by averaging each region’s effect, that annual temperature considerably impacts productivity levels. Also,
wherein only the FEs can reflect some potential differences among re­ based on a sub-national data set for China, Chang et al. (2020) found an
gions but temperature effects would not be affected by FEs. This suggests inverted U-shaped temperature-GDP growth relationship at the pro­
that estimations using linear models tend to capture short-term effects vincial level, with the threshold at 12.2 ◦ C.
and temporary temperature shocks (Schlenker and Lobell, 2010). As it is
difficult for this effect to reflect the response of the socio-economic 4.2. Monthly temperature
system to long-run climate change, many studies using the short-run
response to extrapolate climate impacts may have elicited biased esti­ However, annual temperature may not comprehensively reveal the
mations of GDP impacts because the approach does not account for impact mechanism of climate change on economic consequences.
potential adaptations (Kolstad and Moore, 2020). Summer heatwaves, prolonged cold spells, higher temperatures and
precipitation variability increase uncertainty, and growing evidence has
3.2. Non-linear form suggested that changes in the frequency and intensity of climate ex­
tremes and changes in exposure, impact socio-economic systems. This
As noted, non-linear models have been modelled to capture the non- research has acknowledged that many climate-related events which may
linear relationships (usually in a quadratic relationship) between a series have an impact on economic growth happen over shorter time frames.
of weather variables including temperature, time trends and an As a result, only concentrating on the impacts of annual average tem­
explained variable, which is GDP level or the first difference of the perature—as has been the case in the research on climate econome­
logarithm of GDP per capita. This approach allows the exploration of the trics—runs the danger of overlooking the majority of extreme weather
marginal effects of temperature, which is also a function of temperature events. Therefore, Pretis et al. (2018) introduced within-year monthly
(Kolstad and Moore, 2020). This non-linear shape assesses positive or variables, within-year monthly variation variables and maximum and
negative temperature effects on GDP. For example, Burke et al. (2015) minimum within-year monthly variables which are associated with
proposed a temperature-GDP per capita growth nexus with quadratic temperature and precipitation into the climate econometrics model. The
temperature, precipitation and time trends. The model determined that research demonstrated that, beyond global non-linear temperature im­
the optimal temperature is 13 ◦ C, and economic growth in countries pacts, within-year variability of monthly temperatures and precipitation
with baseline temperatures which are higher than that will be negatively had minimal impact on economic growth.
impacted by temperature, whereas others will be positively affected.
The authors also found this relationship to be globally generalizable and 4.3. Seasonal temperature
robust for agricultural and non-agricultural activity in both developed
and developing countries. The non-linear model can make up for the In addition, some studies have determined that the varied impacts of
shortcomings of linear models; for example, by considering adaptations various seasons may be hidden by prior studies’ aggregating of tem­
to implicitly model the marginal effect of temperature to change with perature data into annually temperature averages (Dell et al., 2012). For
climate across countries. This implies that non-linear models can cap­ instance, Colacito et al. (2021) showed that summer and autumn tem­
ture the long-run adaptation effect by determining the changes in peratures oppositely impact the gross state product (GSP) growth in US
marginal response. states. The average summer temperature harms the GSP growth rate, but
the average autumn temperature has a positive impact. Meanwhile,
3.3. Hybrid forms beyond sectors that are traditionally seen as sensitive to changing cli­
matic conditions, rising summer temperatures have an extensive impact
Some studies have led to a trend towards hybrid functional forms on the full cross-section of industries investigated. Liddle (2018) found
which share the advantages of both linear and non-linear functional summer temperature to harm agricultural GDP in the US. It was found
forms. For example, Kalkuhl and Wenz (2020) used a hybrid model that the industrial production in China responded positively to spring
which includes both linear and non-linear models with interaction lag temperature but adversely to summer temperature (Chen and Yang,
terms. This hybrid model can explicitly distinguish between short-term 2019). Yuan et al. (2020) investigated whether seasonal temperatures
temperature shocks (i.e. deviations from a region’s average climate impact the aggregate economic output of the cities in China, revealing
conditions) and long-term climatic changes that socio-economic systems significant negative effects from warm seasonal temperatures but posi­
could adjust to adapt to the new world with global warming. tive effects from cold seasonal temperatures on economic growth.
Compared with annual temperatures, examining seasonal variations in
temperature can capture the complex effect mechanism; hence, it is
essential to explore the impacts of seasonal temperature when investi­
gating how global warming affects GDP.

386
J.-J. Chang et al. Structural Change and Economic Dynamics 66 (2023) 383–392

4.4. Temperature bins econometrics model, and Kahn et al. (2021) used it to determine that
consistent changes in the temperature above or below historical norms
Another variable specification, temperature bins, is defined by the would negatively impact the growth of GDP per capita. This empirical
frequencies at which the weather events fall into different bins. This research revealed that GDP is impacted by consistent changes in tem­
approach endeavours to uncover the non-linear effects of temperature perature, the pace of temperature change and the extent of temperature
on GDP. For instance, several regressors to record the number of days variability.
per year that are inside defined temperature ranges. (e.g. 0–5 ◦ C,
5–10 ◦ C, etc.) can be used to measure temperature. Deschênes and 5. How is the climate econometrics model specified?
Greenstone (2011) presented an early instance of this method, the major
advantage of which is the ability to avoid specifying functional forms of 5.1. Cross-sectional model
temperature and GDP since the econometrics model with this variable
specification is relatively non-parametric. In addition, using tempera­ Early research investigating the temperature–GDP relationship has
ture bin variable specifications, Chen and Yang (2019) identified the largely been constructed referencing expert judgement (Fankhauser,
critical temperature threshold above and below which adverse effects 1995). The response functions utilised in such models, however, have
can occur on industrial output, while Zhang et al. (2018) determined the little empirical basis (Mendelsohn et al., 2000). To compensate for the
significant effects of high temperatures (above 32 ◦ C) on output. To lack of empirical foundation, early methods focused on single points in
advance the establishment of adaptation plans for a potentially unpre­ time, estimating the marginal impact of long-term changes in the dis­
dictable future because of climate change, the identification of such key tribution of temperature and precipitation using changes in climate
temperature thresholds is crucial (Hallegatte, 2009). It should be noted among cross-sections (Mendelsohn and Nordhaus, 1999). The
that this specification requires high-resolution data since if data are cross-sectional model’s econometric specification includes an explained
aggregated across space or time before creating bins, extreme days could variable that measures the economy, which varies throughout space at a
be averaged away so that the data are smoothed, and if non-linearities given time point, along with a function of temperature and other control
are significant, this may result in inaccurate estimations (Dell et al., variables and an error term. In early research, Mendelsohn et al. (1992)
2014). proposed the first climate econometrics method, which was a
cross-sectional model for estimating the long-run effects of climate
4.5. Daily variability change. This method was used to compare the results of different re­
gions, referencing the current climate of hotter regions with analogues
The degree of variability in daily temperature is not reflected in daily for regions that are currently cooler under an altered climate. Sachs and
temperature bins. Kotz et al. (2021) found that daily temperature vari­ Warner (1997) and Nordhaus (2006a) then used cross-sectional
ability from seasonal expectations has significant impacts on crop yields regression analysis to investigate the research on climate and growth.
as well as asset prices resulting from investor expectations. In contrast, One of the critical advantages of utilizing the cross-sectional model
numerous empirical findings have demonstrated that when investigated lies in its ability to capture long-term equilibrium effects which reflect
across a range of non-climate elements and timelines, uncertain­ the net impacts after potential adaptation measures have occurred (re­
ty—measured as variability or volatility—poses an intrinsic limitation sidual damages) (Dell et al., 2014; Burke et al., 2015; Kalkuhl and Wenz,
on macro-economic output. For instance, temporal fluctuations of GDP 2020). However, one important criticism of the cross-sectional approach
have negative effects on GDP, as well as volatility in government is the biased estimation caused by omitted variables (Hsiang, 2016). In
spending and exchange rates (Ramey and Ramey, 1994; Aghion et al., practice, a common approach to avoid such biased estimation is to
2009). In addition, for the determinants of GDP, agricultural output and control for explanatory variables as much as possible, as Nordhaus
welfare would also be negatively affected by food price volatility (2006b) did in his global-level geographic–economic cross-sectional
(Myers, 2006; Haile et al., 2016). Thus, relevant contemporary research model, controlling for a variety of factors such as temperature, rain­
has concluded that macro-economic output and GDP would potentially fall, elevation, ruggedness and soil type. Another option for addressing
be affected by the aggregate effects of daily temperature variability omitted variable bias is to restrict the sub-samples of observations. For
across these fundamental economic factors. For example, to obtain the instance, Dell et al. (2009) examined the impact of temperature on per
measurement of annual daily variability, Kotz et al. (2021) calculated capita income using (prefecture) city-level data for 12 countries, con­
the intra-monthly standard deviation of daily temperature and then trolling for national FEs and state (province) FEs in the regression
averaged these standard deviations across months of a given year. The analysis. It is noteworthy that adding abundant control variables to a
study demonstrated that increases in seasonally adjusted daily temper­ cross-sectional regression model does not necessarily reduce the esti­
ature variability can negatively affect economic growth on a global mation bias of the model. If the control variables are endogenous or
scale, independent of (and in addition to) changes in annual average determined by the climate itself, then their introduction into the model
temperature. would instead result in the so-called ‘bad controls’ and ‘over-controlling’
problems. In addition, some factors that do not vary over time (e.g.
4.6. Deviation from historical norms geography, culture, customs and institutions) may also affect the
dependant variable (some of which may be highly correlated with the
Previous studies indicated a strong upward trend in temperature data climate variable and the control variable) and have often been excluded
for nearly all countries or regions, so it almost seems as if it is possible in cross-sectional regression models because the relevant data may be
that the two trends are correlated in terms of the temperature-GDP difficult to obtain or is not accurately observed (Hsiang, 2016).
growth nexus. This leaves a key problem for climate econometrics in
identifying the temperature-GDP causal links which is that the temper­ 5.2. Panel model
ature trend included in the regression can cause biased estimation
because it introduces a linear trend to GDP growth which is spurious and A recent large body of literature associated with climate economet­
unsupported by data (Kahn et al., 2021). This is an econometric pitfall rics has used panel models with various FEs to identify the relationships
related to trended variables that can be addressed by introducing vari­ between temperature change and GDP (Dell et al., 2014; Kolstad and
able specification of temperature deviation from its respective historical Moore, 2020). In a panel model, the economically explained variables
norm to explore the long-run impacts of climate change on (e.g. GDP or some determinant factors of GDP such as TFP), weather
macro-economic output across countries. This approach allows for variables (temperature etc.), FEs and control variables all vary over
consideration of adaptation and non-linearity in the climate space and time. The model applies a linear or non-linear function

387
J.-J. Chang et al. Structural Change and Economic Dynamics 66 (2023) 383–392

consisting of explanatory variables, FEs (sometimes the model includes medium- and long-run variations in weather and climate, hybrid panel
time trends) and an error term. Compared to cross-sectional regressions, models have been introduced into econometric climate impact studies.
this specification provides a number of advantages in terms of causal Such studies usually use panel data and control for unobservable con­
identification, due to the fact that the extended set of data utilised founding variables, integrating cross-sectional, interannual and decadal
changes through time and space, FEs and time trends may be used to variations in temperature to estimate level and growth effects of tem­
account for a variety of unobserved omitted variables, including perature from weather shocks in the short run and climate equilibrium
time-invariant changes over space, which is frequently a topic of dis­ impacts in the long run. Comparing the responses to these variations can
cussion in this context (Deschênes and Greenstone, 2007). Panel models provide information related to potential adaptation paths and produce a
have been widely utilised to analyse the temperature-economic growth more accurate estimate of climate impacts (Kalkuhl and Wenz, 2020;
nexus (Dell et al., 2009, 2012; Burke et al., 2015; Diffenbaugh and Kolstad and Moore, 2020).
Burke, 2019), crop yields (Schlenker and Roberts, 2009; Chen and Yang, Over the recent years, researchers have developed a hybrid method
2019), energy demand (Auffhammer and Mansur, 2014; Wenz et al., called the long difference approach to quantify the climate adaptation
2017), labor productivity (Deryugina and Hsiang, 2014) and human effects. It has the same econometric expressions as the panel model, but
capital (Graff Zivin and Neidell, 2014; Graff Zivin et al., 2018). These the explained and explanatory variables in the model are averaged over
studies shed light on the possible long-term effects of climate change. In time, and can be averaged over a decade, a few decades or even a cen­
some instances, anthropogenic warming is anticipated to cause signifi­ tury (e.g. 1960–1970 or 1970–2000). Although its econometric ex­
cantly greater economic losses than what prior research employing IAMs pressions are very similar to those of a general panel model, there are
had indicated (Burke et al., 2015). significant differences in the interpretation of the model’s estimated
Although the problem of time-invariant omitted variables in the parameters. Compared to the panel model, the hybrid describes changes
cross-sectional model can be addressed by introducing individual FEs that are closer to climate change rather than weather change; therefore,
into the panel model, there is also a potential problem of time-varying the temperature effects describe medium- to long-term climate effects
omitted variables. One example is the impact of seasonal temperature instead of short-term effects. Dell et al. (2012) used this kind of hybrid
on industrial output, Chen and Yang (2019) used the duration of sun­ model to estimate the effect of a 1 ◦ C rising in temperature on GDP
light as an explanatory variable. Since sunlight is strongly correlated growth occurring over 15 years. The authors find a temperature effect
with temperature and has been shown to have an important effect on that is similar to the effect estimated by a panel model with interannual
human health and labour productivity (Lambert et al., 2002; Patz et al., changes in temperature variables, asserting that adaptation is not
2005; De Witte and Saal, 2010), omitting sunlight from the regression particularly effective throughout this period.
analysis could result in inaccurate estimates of the annual temperature’s Another challenge is distinguishing the effects of short-, medium-
impacts on overall economic output. In addition to controlling for and long-term variations in weather and climate. Few studies in climate
omitted variable bias, the FEs panel regression technique also has some econometrics have identified the effects of temperature on GDP at
weaknesses related to its emphasis on short-run changes in the weather different temporal scales simultaneously. This kind of research can
rather than the long-run variability of climatic conditions. It’s critical to contribute to revealing whether temperature’s effects on GDP are
distinguish between short-run and long-run effects because the persistent (Dell et al., 2012; Burke et al., 2015; Diffenbaugh and Burke,
socio-economic system could be better able to react to long-run changes 2019; Rosen, 2019; Newell et al., 2021), which could impact the design
than short-run ones by investing in adaptative responses. Identifying of optimal future climate policies. Existing estimates of the effects of
only short-term responses to weather variation and only extrapolating weather shocks on GDP have been criticised for methodological short­
the short-run effects of weather variation on economic outcomes would comings, including the failure to consider trends in climate variables and
also tend to overestimate the effects of global warming if future signif­ the impact of average climate conditions on the marginal response to
icant adaptation occurs (Kolstad and Moore, 2020). weather. Concerning spatial scope, temporal scale and particular tech­
Traditionally, the FE estimators utilized in the research with the nological challenges, Kalkuhl and Wenz (2020) addressed a number of
panel model have assumed that climatic variables are strictly exogenous; shortcomings in earlier analyses, explicitly distinguishing between
however, there are the bi-directional feedback effects between temper­ short-term weather shocks and long-term climate effects and offering
ature and economic growth. According to Nordhaus (1992), greater potentially insightful information related to adaptation.
economic growth results in more GHGs, which boosts the global average
temperature. Meanwhile, a rise in the global average temperature would 6. Improve damage functions in IAMs
in turn slow down economic growth. As economic growth in the pre­
vious periods could have feedback effects on future temperature, tem­ IAMs have been created to represent the complex interactions be­
perature may not be strictly exogenous in models that estimate the tween socio-economic and natural systems under the background of
impact of temperature on economic growth (Kahn et al., 2021).Dynamic climate change to provide insights and policy suggestions to confront
panels with a cross-section dimension (N) that is higher than the time global warming. The damage module, which introduces a series of
series dimension (T) exhibit small-T bias due to the FE estimator’s damage functions for calculating the economic impacts of climate
limitations, meaning that there is a biased estimation in the model change, is an essential component of IAMs. Losses in GDP and con­
specification based on a standard FE estimator if one or more dependant sumption are usually employed to measure economic damage (Nord­
variables are not strictly exogenous (Kahn et al., 2021). To avoid this, haus, 2019; Rising et al., 2022). Previously, the parameters of multiple
Kahn et al. (2021) contributed by introducing the half-panel jackknife damage functions in IAMs were calibrated to match the existing esti­
FE (HPJ-FE) estimator proposed by Chudik et al. (2016) into the model mation results (Tol, 2009; Pindyck, 2013; Diaz and Moore, 2017). Over
with the specification of the widely-used FE estimator. When the panel’s the previous few years, numerous empirical research has reassessed the
temporal dimension is modest in terms of N, the HPJ-FE estimator is long-term economic effects of climate change in different sectors (Car­
resilient to potential feedback effects from total economic output to the leton and Hsiang, 2016; Hsiang et al., 2017), and some studies have
variables related to climate change and effectively corrects the bias if the integrated damage assessment results using the meta-analysis approach
regressors are weakly exogenous. (Howard and Sterner, 2017). These more recent empirical results can be
further integrated into the modelling procedure of damage functions.
5.3. Hybrid panel model However, the current damage function approach has been criticised
for being largely static and not considering the dynamic effects by which
To overcome the challenges of general panel models for capturing temperature changes may affect the growth of macro-economic output
long-term climate impacts and to distinguish the impacts of short-, and, consequently, future welfare (Pindyck, 2013; Diaz and Moore,

388
J.-J. Chang et al. Structural Change and Economic Dynamics 66 (2023) 383–392

2017; Batten, 2018; Rising et al., 2022). Projections of future economic investigate the lack of clarity regarding adaptation options. Notably,
losses vary significantly depending on whether growth or level models recent advancements in methodology have allowed for assessing the
are specified as growth effect shocks are considered to permanently benefits and costs of adaptation for some individual sectors that are
damage the economy rather than temporarily reducing output in a given derived empirically from observed variations in weather sensitivity
year as level effects do. (Carleton et al., 2022). Nevertheless, few studies have explored the
Thus, an alternative for improving traditional damage functions is impacts and costs of adaptation in macro-economic systems from a
based on the recent empirical basis from a reduced-form temperature- top-down perspective, using GDP as a measure, as has been done for
GDP growth nexus (Dell et al., 2014; Burke et al., 2015; Moore and Diaz, climate impacts. This is also a research question worth exploring in the
2015). To incorporate these growth shocks into damage assessment in future (Sussman et al., 2014; Chapagain et al., 2020).
an IAM, Dietz and Stern (2015) provided two key examples. The first To identify adaptation effects more accurately, it is crucial to
examined the relationship between temperature impacts and physical differentiate between short- and long-run effects as people and enter­
capital stock which is boosted by capital investment (It ) but diminished prises may adjust to long-run changes in the expected distribution of
by physical depreciation (δ) and damage (DKt ) representing the direct weather differently than they may to short-run, temporary, and unex­
climate impacts on capital stock. pected variations in weather. If this adaptation is significant, then the
( ) effect of weather change may not offer an appropriate analogue for the
Kt+1 = 1 − DKt (1 − δ)Kt + It long-run effects of climate change. According to Dell et al. (2009)
The second examined the relationship between temperature impacts comparison of cross-sectional and panel data, adaptation mitigates the
and productivity, which inherently reflects the growth effect of tem­ detrimental effects of temperature shocks by 50%. If non-linearities
perature change. emerging in the temperature range outside the previous data differ
( ) from those within it, extrapolating from historical temperature data may
At+1 = 1 − DAt At not be rational (Kahn et al., 2021). Therefore, referencing a short-term
response to extrapolate climate effects will suffer a biased estimation
The mechanism by which climate impacts are modelled is critical as
of damages if the model ignores long-term adaptation. To overcome the
even the smallest growth effects would ultimately outweigh large level
bias estimation from extrapolation, future research should further refine
effects. Removing growth effects from the model would significantly
the use of the long difference model (cross-sectional regressions) in
reduce the potential economic losses from global warming. Conse­
growth rates across longer periods to model the potential adaptation
quently, this may have a detrimental impact on model-based policy
effects more thoroughly. Meanwhile, in panel data analyses, the devia­
implications. For instance, Moyer et al. (2014) simulated the impacts of
tion from long-run averages to estimate unbiased weather impacts
temperature on productivity growth using an IAM and discovered even a
should also be focused on for exploring adaptation effects in climate
tiny growth effect to have a significant effect on the optimal climate
econometrics. This change will enhance existing panel models to
policy. Based on the empirical results from Dell et al. (2012), Moore and
explicitly represent climatic variability in the calculation of long-run
Diaz (2015) calibrated climate damage to the growth rate in a simple
damage functions and enable the construction of an implicit model of
IAM, determining that enabling climate change to directly shock GDP
adaptation (Mendelsohn et al., 2016; Kahn et al., 2021; Tol, 2021).
growth would greatly boost the optimal mitigation rate in the near term.
Since GDP does not account for the monetized effects of temperature
The majority of IAMs simply take into account the level effects of
change on non-market sectors, future research should further enrich this
long-term climatic conditions on economic output. Kalkuhl and Wenz
limitation by exploring empirical links between climate change and
(2020) allowed climate change to affect level as well as growth rate of
biodiversity impacts, health losses and tipping points, and monetizing
GDP, separating the effects of climate change on productivity and
these climate impacts (van der Wijst et al., 2023). In addition, temper­
growth rate. Growth is impacted by (i) the short-run impact of a change
ature change may also shock the trade between countries or regions by
in temperature by the impact on the productivity level, (ii) a temporary
some potential channels (Burke et al., 2018; He et al., 2023). Future
impact on growth rate to converge with the long-run growth rate of the
research should also identify how this underlying factor affects the
economy, and (iii) the long-run productivity growth rate. This empirical
empirical relationship between temperature and GDP.
evidence enabled clear identification of the macro-economic effects of
climate change at various periods in the IAM to allow policymakers to
make informed risk management decisions that are tailored to different 8. Conclusions
periods. In addition, the model also contributes to formalising the
depiction of the consequences of climate change in already-existing Exploring the reduced-form temperature-GDP relationship is a crit­
IAMs. Compared with the original DICE-2016 damage function pre­ ical area of climate econometrics. The use of level and growth effects to
dictions, the authors determined that the social cost of carbon is higher evaluate climate change impacts has given rise to a long-standing and
than three times higher when incorporating the damage function cali­ well-known controversy in the literature. We find that there are
brated by their empirical results into the DICE-2016. compelling theoretical justifications for thinking that climate change
might affect the pace of GDP growth, not only its level, and projections
7. Limitations of potential losses due to climate change vary significantly relying on
which effect is specified. However, there is minimal empirical evidence
Modelling adaptation effects based on climate econometrics is a and weak consensus regarding how to interpret this effect or specifica­
profound and challenging research topic; however, it is one of the least tions. Recent studies have led the trend to explore whether these two
researched subfields in climate economics (Burke and Emerick, 2016). effects would both exist and how they manifest at different times,
This is the result of one of the most important barriers—a dearth of data implying that future research should incorporate multiple model spec­
regarding the aggregated costs and benefits of adaptations. In addition, ifications of climate econometrics into a single model to distinguish
current knowledge regarding the available options for future adaptation between the two effects in the short, medium, and long term.
measures and adaptive capacity is extremely limited, and the actual In terms of variable forms, in addition to annual average tempera­
costs and impacts of different adaptation options remain unknown tures, the introduction of daily, monthly and seasonal temperature
(Sussman et al., 2014; Trnka et al., 2017; Chapagain et al., 2020; Rising variables and related variants (e.g. temperature bins) would enrich the
et al., 2022). Thus, it would be very difficult to model climate adaptation exploration of the mechanisms by which temperature affects GDP from
explicitly using climate econometric models in comparison with implicit multiple perspectives. An increasing number of studies are utilizing
modelling of climate adaptation. Future research should further relevant variables that can capture climate variability to describe
climate change, rather than solely relying on absolute values of climate

389
J.-J. Chang et al. Structural Change and Economic Dynamics 66 (2023) 383–392

or weather factors. Future research should focus on developing climate the work reported in this paper.
econometrics models that incorporate or improve these variable forms
in order to avoid potential estimation bias caused by spurious re­ CRediT authorship contribution statement
gressions, ignoring the adaptation effects and so on. For instance, using a
temperature variable that captures the deviations of temperature from Jun-Jie Chang: Investigation, Data curation, Methodology,
respective historical norms could further expand the long-term climate Conceptualization, Formal analysis, Visualization, Writing – original
impacts. draft, Writing – review & editing. Zhifu Mi: Resources, Supervision,
In terms of functional forms, the choice of linear or non-linear Funding acquisition, Validation, Conceptualization, Data curation,
models significantly impacts the magnitude of damage estimates. The Formal analysis, Investigation, Methodology, Writing – review & edit­
linear specification is suitable for estimating short-run weather impacts ing. Yi-Ming Wei: Resources, Supervision, Funding acquisition, Vali­
but it would result in a biased estimation of economic losses as it doesn’t dation, Conceptualization, Data curation, Formal analysis,
take long-term adaptations into consideration. In contrast to the linear Investigation, Methodology, Project administration, Writing – review &
specification, the estimation results from the non-linear panel model can editing.
reflect both short- and long-term impacts. Therefore, future studies
should use more comprehensive data and integrate long-difference non- Data availability
linear models to identify climate adaptation patterns at a regional level
over an extended period of time. This potential improvement would Data will be made available on request.
reduce the bias of projecting future climate impacts by climate econo­
metrics models. Acknowledgments
Furthermore, concerning the reduced form of the temperature–GDP
relationship, this study provides a careful review of the development of The authors are grateful for financial support from the National
the specifications of the econometrics model along historical lines. In Natural Science Foundation of China (NSFC) (Nos.72293600 and
initial models, the temperature–GDP relationships lacked spatial and 72293605). We thank all colleagues from the center for Energy and
structural detail and were largely constructed on expert judgement, with Environmental Policy Research, Beijing Institute of Technology and the
minimal empirical foundations. To compensate for the lack of empirical School of Geography and Environment, University of Oxford for
foundations, the initial approach used a cross-sectional model to esti­ providing helpful suggestions.
mate the long-run equilibrium effects of climate change, rather than
weather effects, by accounting for the net benefits of potential adapta­ References
tion measures; however, this approach faced the challenge of biased
estimation caused by omitted variables. To overcome this, a more recent Aghion, P., Bacchetta, P., Rancière, R., Rogoff, K., 2009. Exchange rate volatility and
productivity growth: the role of financial development. J. Monet. Econ. 56, 494–513.
strand of literature applying the panel model has offered significant Auffhammer, M., 2018. Quantifying economic damages from climate change. J. Econ.
advantages over cross-sectional model in terms of causal identification. Perspect. 32, 33–52.
The expanded set of data utilised changes with time, space and FEs, as Auffhammer, M., Aroonruengsawat, A., 2011. Simulating the impacts of climate change,
prices and population on California’s residential electricity consumption. Clim.
well as the terms of time trends, controlling for various omitted factors Chang. 109, 191–210.
that are not observed, especially time-invariant change across space. Auffhammer, M., Mansur, E.T., 2014. Measuring climatic impacts on energy
Other recent studies have improved the FE setting in traditional panel consumption: a review of the empirical literature. Energy Econ. 46, 522–530.
Batten, S., 2018. Climate change and the macro-economy: a critical review. Bank of
models to further reduce estimation bias due to the bi-directional effect England Working Papers No.706. Bank of England.
between temperature and economic growth. Based on new hybrid panel Burke, M., Davis, W.M., Diffenbaugh, N.S., 2018. Large potential reduction in economic
models including mixed characteristics of the above model specifica­ damages under UN mitigation targets. Nature 557, 549–553.
Burke, M., Emerick, K., 2016. Adaptation to climate change: evidence from US
tions, recent research trends have been distinguishing the impacts of
agriculture. Am. Econ. J. Econ. Policy 8, 106–140.
short-, medium- and long-run variations in weather and climate to Burke, M., Hsiang, S.M., Miguel, E., 2015. Global non-linear effect of temperature on
further capture the adaptation effects. economic production. Nature 527, 235–239.
We find that one of the reasons that climate econometrics is an Burke, M., Tanutama, V., (2019) Climatic constraints on aggregate economic output.
National Bureau of Economic Research Working Papers No. 25779 DOI: 10.33
essentially academic pursuit is that such research can provide empirical 86/w25779.
evidence for improving damage functions in IAMs. The literature review Carleton, T., Jina, A., Delgado, M., Greenstone, M., Houser, T., Hsiang, S., Hultgren, A.,
reveals that a substantial amount of empirical studies is available to Kopp, R.E., McCusker, K.E., Nath, I., 2022. Valuing the global mortality
consequences of climate change accounting for adaptation costs and benefits. Q. J.
update the parameters of the damage functions and a meta-analysis Econ. 137, 2037–2105.
approach could produce a damage function that reflects the latest Carleton, T.A., Hsiang, S.M., 2016. Social and economic impacts of climate. Science 353,
empirical findings more fully. Meanwhile, according to the empirical aad9837.
Carrilho-Nunes, I., Catalão-Lopes, M., 2022. The effects of environmental policy and
temperature-GDP growth nexus, future research can further model technology transfer on GHG emissions: the case of Portugal. Struct. Chang. Econ.
temperature shocks on GDP growth in IAMs by measuring the rela­ Dyn. 61, 255–264.
tionship between the impacts of temperature on macro-economic fac­ Castle, J.L., Hendry, D.F., 2022. Econometrics for modelling climate change. Oxford
Research Encyclopedia of Economics and Finance. 28. Oxford University Press.
tors, such as physical capital stock and productivity. https://doi.org/10.1093/acrefore/9780190625979.013.675.
Also, it is worth noting several other directions for future research in Chakraborty, S.K., Mazzanti, M., 2021. Renewable electricity and economic growth
this field, including further investigating the effects and costs of climate relationship in the long run: panel data econometric evidence from the OECD. Struct.
Chang. Econ. Dyn. 59, 330–341.
adaptation in the macroeconomic system from a top-down perspective
Chang, J.J., Wei, Y.M., Yuan, X.C., Liao, H., Yu, B.-Y., 2020. The nonlinear impacts of
based on innovative econometric models and increasingly abundant global warming on regional economic production: an empirical analysis from China.
data, identifying the potential empirical nexus between climate change Weather Clim. Soc. 12, 759–769.
and some non-market sectors such as biodiversity, health and tipping Chapagain, D., Baarsch, F., Schaeffer, M., D’Haen, S., 2020. Climate change adaptation
costs in developing countries: insights from existing estimates. Clim. Dev. 12, 1–9.
points as well as examining the underlying temperature effects in the Chen, F., Zhang, X., Chen, Z., 2023. Behind climate change: extreme heat and health cost.
trading network. Struct. Chang. Econ. Dyn. 64, 101–110.
Chen, X., Yang, L., 2019. Temperature and industrial output: firm-level evidence from
China. J. Environ. Econ. Manag. 95, 257–274.
Declaration of Competing Interest Chudik, A., Pesaran, H., Yang, J.C., 2016. Half-panel jackknife fixed-effects estimation of
linear panels with weakly exogenous regressors. J. Appl. Econom. 33 (6), 816–836.
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence

390
J.-J. Chang et al. Structural Change and Economic Dynamics 66 (2023) 383–392

Colacito, R., Hoffmann, B., Phan, T., 2021. The impact of rising temperature on US Mendelsohn, R., Nordhaus, W., 1999. The impact of global warming on agriculture: a
economic growth. World Scientific Encyclopedia of Climate Change: Case Studies of ricardian analysis: reply. Am. Econ. Rev. 89, 1046–1048.
Climate Risk, Action, and Opportunity Volume 3. World Scientific, pp. 133–138. Mendelsohn, R., Nordhaus, W., Shaw, D., 1992. The impact of climate on agriculture: a
Davis, L.W., Gertler, P.J., 2015. Contribution of air conditioning adoption to future Ricardian approach. Cowles Foundation Discussion Papers 84. Cowles Foundation,
energy use under global warming. Proc. Natl. Acad. Sci. 112 (19), 5962–5967. Yale University.
De Witte, K., Saal, D.S., 2010. Is a little sunshine all we need? On the impact of sunshine Mendelsohn, R., Prentice, I.C., Schmitz, O., Stocker, B., Buchkowski, R., Dawson, B.,
regulation on profits, productivity and prices in the Dutch drinking water sector. 2016. The ecosystem impacts of severe warming. Am. Econ. Rev. 106, 612–614.
J. Regul. Econ. 37, 219–242. Moore, F.C., Diaz, D.B., 2015. Temperature impacts on economic growth warrant
Dell, M., Jones, B.F., Olken, B.A., 2009. Temperature and income: reconciling new cross- stringent mitigation policy. Nat. Clim. Chang. 5, 127–131.
sectional and panel estimates. Am. Econ. Rev. 99, 198–204. Moyer, E.J., Woolley, M.D., Matteson, N.J., Glotter, M.J., Weisbach, D.A., 2014. Climate
Dell, M., Jones, B.F., Olken, B.A., 2012. Temperature shocks and economic growth: impacts on economic growth as drivers of uncertainty in the social cost of carbon.
evidence from the last half century. Am. Econ. J. Macroecon. 4, 66–95. J. Legal Stud. 43, 401–425.
Dell, M., Jones, B.F., Olken, B.A., 2014. What do we learn from the weather? The new Myers, R.J., 2006. On the costs of food price fluctuations in low-income countries. Food
climate-economy literature. J. Econ. Lit. 52, 740–798. Policy 31, 288–301.
Deryugina, T., Hsiang, S. (2014) Does the environment still matter? Daily temperature Newell, R.G., Prest, B.C., Sexton, S.E., 2021. The GDP-Temperature relationship:
and income in the United States. National Bureau of Economic Research Working implications for climate change damages. J. Environ. Econ. Manag. 108, 102445.
Papers No. 20750 DOI: 10.3386/w20750. Nordhaus, W., 2019. Climate change: the ultimate challenge for economics. Am. Econ.
Deschênes, O., Greenstone, M., 2007. The economic impacts of climate change: evidence Rev. 109, 1991–2014.
from agricultural output and random fluctuations in weather. Am. Econ. Rev. 97, Nordhaus, W.D., 1992. An optimal transition path for controlling greenhouse gases.
354–385. Science 258, 1315–1319.
Deschênes, O., Greenstone, M., 2011. Climate change, mortality, and adaptation: Nordhaus, W.D., (2006a) The economics of hurricanes in the United States. National
evidence from annual fluctuations in weather in the US. Am. Econ. J. Appl. Econ. 3, Bureau of Economic Research Working Papers No. 12813 DOI: 10.3386/w12813.
152–185. Nordhaus, W.D., 2006b. Geography and macroeconomics: new data and new findings.
Diaz, D., Moore, F., 2017. Quantifying the economic risks of climate change. Nat. Clim. Proc. Natl. Acad. Sci. 103, 3510–3517.
Chang. 7, 774–782. Patz, J.A., Campbell-Lendrum, D., Holloway, T., Foley, J.A., 2005. Impact of regional
Dietz, S., Stern, N., 2015. Endogenous growth, convexity of damage and climate risk: climate change on human health. Nature 438, 310–317.
how Nordhaus’ framework supports deep cuts in carbon emissions. Econ. J. 125, Pindyck, R.S., 2013. Climate change policy: what do the models tell us? J. Econ. Lit. 51,
574–620. 860–872.
Diffenbaugh, N.S., Burke, M., 2019. Global warming has increased global economic Piontek, F., Drouet, L., Emmerling, J., Kompas, T., Méjean, A., Otto, C., Rising, J.,
inequality. Proc. Natl. Acad. Sci. 116, 9808–9813. Soergel, B., Taconet, N., Tavoni, M., 2021. Integrated perspective on translating
Fabozzi, F.J., Focardi, S., Ponta, L., Rivoire, M., Mazza, D., 2022. The economic theory of biophysical to economic impacts of climate change. Nat. Clim. Chang. 11, 563–572.
qualitative green growth. Struct. Chang. Econ. Dyn. 61, 242–254. Pretis, F., Schwarz, M., Tang, K., Haustein, K., Allen, M.R., 2018. Uncertain impacts on
Fankhauser, S., 1995. Protection versus retreat: the economic costs of sea-level rise. economic growth when stabilizing global temperatures at 1.5◦ C or 2◦ C warming.
Environ. Plann. A 27, 299–319. Philos. Trans. R. Soc. A 376, 20160460.
Fankhauser, S., Tol, R.S., 2005. On climate change and economic growth. Resour. Energy Ramey, G., Ramey, V.A., (1994) Cross-country evidence on the link between volatility
Econ. 27, 1–17. and growth. National Bureau of Economic Research Working Papers No.4959 DOI:
Graff Zivin, J., Hsiang, S.M., Neidell, M., 2018. Temperature and human capital in the 10.3386/w4959.
short and long run. J. Assoc. Environ. Resour. Econ. 5, 77–105. Rising, J.A., Taylor, C., Ives, M.C., Ward, R.E., 2022. Challenges and innovations in the
Graff Zivin, J., Neidell, M., 2014. Temperature and the allocation of time: implications economic evaluation of the risks of climate change. Ecol. Econ. 197, 107437.
for climate change. J. Labor. Econ. 32, 1–26. Rosen, R.A., 2019. Temperature impact on GDP growth is overestimated. Proc. Natl.
Haile, M.G., Kalkuhl, M., von Braun, J., 2016. Worldwide acreage and yield response to Acad. Sci. 116, 16170.
international price change and volatility: a dynamic panel data analysis for wheat, Sachs, J.D., Warner, A.M., 1997. Fundamental sources of long-run growth. Am. Econ.
rice, corn, and soybeans. Am. J. Agric. Econ. 98 (1), 172–190. Rev. 87, 184–188.
He, K., Mi, Z., Zhang, J., Li, J., Coffman, D.M., 2023. The polarizing trend of regional co2 Sarofim, M.C., Martinich, J., Neumann, J.E., Willwerth, J., Kerrich, Z., Kolian, M.,
emissions in China and its implications. Environ. Sci. Technol. 57 (11), 4406–4414. Fant, C., Hartin, C., 2021. A temperature binning approach for multi-sector climate
Henseler, M., Schumacher, I., 2019. The impact of weather on economic growth and its impact analysis. Clim. Chang. 165, 1–18.
production factors. Clim. Chang. 154, 417–433. Schlenker, W., Auffhammer, M., 2018. The cost of a warming climate. Nature 557,
Howard, P.H., Sterner, T., 2017. Few and not so far between: a meta-analysis of climate 498–499.
damage estimates. Environ. Resour. Econ. 68, 197–225. Schlenker, W., Hanemann, W.M., Fisher, A.C., 2005. Will US agriculture really benefit
Hsiang, S., 2016. Climate econometrics. Annu. Rev. Resour. Econ. 8, 43–75. from global warming? Accounting for irrigation in the hedonic approach. Am. Econ.
Hsiang, S., Kopp, R., Jina, A., Rising, J., Delgado, M., Mohan, S., Rasmussen, D.J., Muir- Rev. 95, 395–406.
Wood, R., Wilson, P., Oppenheimer, M., Larsen, K., Houser, T., 2017. Estimating Schlenker, W., Lobell, D.B., 2010. Robust negative impacts of climate change on African
economic damage from climate change in the United States. Science 356, agriculture. Environ. Res. Lett. 5, 014010.
1362–1369. Schlenker, W., Roberts, M.J., 2009. Nonlinear temperature effects indicate severe
Hsiang, S.M., 2010. Temperatures and cyclones strongly associated with economic damages to U.S. crop yields under climate change. Proc. Natl. Acad. Sci. 106,
production in the Caribbean and Central America. Proc. Natl. Acad. Sci. 107, 15594–15598.
15367–15372. Stern, N., 2013. The structure of economic modeling of the potential impacts of climate
Hsiang, S.M., Burke, M., Miguel, E., 2013. Quantifying the influence of climate on human change: grafting gross underestimation of risk onto already narrow science models.
conflict. Science 341, 1235367. J. Econ. Lit. 51, 838–859.
Hsiang, S.M., Jina, A.S., (2014) The causal effect of environmental catastrophe on long- Stern, N., Stiglitz, J., Taylor, C., 2022. The economics of immense risk, urgent action and
run economic growth: evidence from 6,700 cyclones. National Bureau of Economic radical change: towards new approaches to the economics of climate change.
Research Working Papers No. 20352 DOI: 10.3386/w20352. J. Econ. Methodol. 29, 181–216.
Kahn, M.E., Mohaddes, K., Ng, R.N.C., Pesaran, M.H., Raissi, M., Yang, J.C., 2021. Long- Stoerk, T., Wagner, G., Ward, R.E., 2018. Policy brief—recommendations for improving
term macroeconomic effects of climate change: a cross-country analysis. Energy the treatment of risk and uncertainty in economic estimates of climate impacts in the
Econ. 104, 105624. sixth intergovernmental panel on climate change assessment report. Rev. Environ.
Kalkuhl, M., Wenz, L., 2020. The impact of climate conditions on economic production. Econ. Policy 12 (2), 371–376.
Evidence from a global panel of regions. J. Environ. Econ. Manag. 103, 102360. Sussman, F., Krishnan, N., Maher, K., Miller, R., Mack, C., Stewart, P., Shouse, K.,
Kolstad, C.D., Moore, F.C., 2020. Estimating the economic impacts of climate change Perkins, B., 2014. Climate change adaptation cost in the US: what do we know?
using weather observations. Rev. Environ. Econ. Policy 14 (1), 1–24. Climat. Policy 14, 242–282.
Kotz, M., Wenz, L., Stechemesser, A., Kalkuhl, M., Levermann, A., 2021. Day-to-day Tol, R.S.J., 2009. The economic effects of climate change. J. Econ. Perspect. 23, 29–51.
temperature variability reduces economic growth. Nat. Clim. Chang. 11, 319–325. Tol, R.S.J., 2021. Do climate dynamics matter for economics? Nat. Clim. Chang. 11,
Kumar, S., Khanna, M., 2019. Temperature and production efficiency growth: empirical 802–803.
evidence. Clim. Chang. 156, 209–229. Trnka, M., Hlavinka, P., Wimmerová, M., Pohanková, E., Rötter, R., Olesen, J.E.,
Lambert, G.W., Reid, C., Kaye, D.M., Jennings, G.L., Esler, M.D., 2002. Effect of sunlight Kersebaum, K.C., Semenov, M., 2017. Paper on model responses to selected adverse
and season on serotonin turnover in the brain. Lancet 360, 1840–1842. weather conditions. Facce Macsur Rep. 10, 1–2.
Lecocq, F., Shalizi, Z., 2007. How might climate change affect economic growth in van der Wijst, K.I., Bosello, F., Dasgupta, S., Drouet, L., Emmerling, J., Hof, A.,
developing countries? A review of the growth literature with a climate lens. Policy Leimbach, M., Parrado, R., Piontek, F., Standardi, G., van Vuuren, D., 2023. New
Research Working Paper; No. 4315. The World Bank. damage curves and multimodel analysis suggest lower optimal temperature. Nat.
Lemoine, D., Kapnick, S., 2016. A top-down approach to projecting market impacts of Clim. Chang. 13, 434–441.
climate change. Nat. Clim. Chang. 6, 51–55. Wei, Y.M., Han, R., Liang, Q.M., Yu, B.Y., Yao, Y.F., Xue, M.M., Zhang, K., Liu, L.J.,
Letta, M., Tol, R.S.J., 2019. Weather, climate and total factor productivity. Environ. Peng, J., Yang, P., Mi, Z., Du, Y.F., Wang, C., Chang, J.J., Yang, Q.R., Yang, Z.,
Resour. Econ. 73, 283–305. Shi, X., Xie, W., Liu, C., Liao, H., 2018. An integrated assessment of INDCs under
Liddle, B., 2018. Warming and income growth in the United States: a Heterogenous, Shared Socioeconomic Pathways: an implementation of C3IAM. Nat. Hazards 92,
comon factor dynamic panel analysis. Clim. Chang. Econ. 9, 1–14. 585–618.
Mendelsohn, R., Morrison, W., Schlesinger, M.E., Andronova, N.G., 2000. Country- Wei, Y.M., Han, R., Wang, C., Yu, B., Liang, Q.-M., Yuan, X.C., Chang, J., Zhao, Q.,
specific market impacts of climate change. Clim. Chang. 45, 553–569. Liao, H., Tang, B., Yan, J., Cheng, L., Yang, Z., 2020. Self-preservation strategy for

391
J.-J. Chang et al. Structural Change and Economic Dynamics 66 (2023) 383–392

approaching global warming targets in the post-Paris Agreement era. Nat. Commun. Zhang, P., Deschenes, O., Meng, K., Zhang, J., 2018. Temperature effects on productivity
11, 1624. and factor reallocation: evidence from a half million chinese manufacturing plants.
Wenz, L., Levermann, A., Auffhammer, M., 2017. North–south polarization of European J. Environ. Econ. Manag. 88, 1–17.
electricity consumption under future warming. Proc. Natl. Acad. Sci. 114, Zivin, J., Neidell, M. (2010) Temperature and the allocation of time: implications for
E7910–E7918. climate change. National Bureau of Economic Research Working Papers No. 15717
Yuan, X.C., Yang, Z., Wei, Y.M., Wang, B., 2020. The economic impacts of global DOI:10.3386/w15717.
warming on Chinese cities. Clim. Chang. Econ. 11 (02), 2050007.

392