Climate effects on the diversification strategy of export firms: Evidence from China

RESEARCH ARTICLE Climate effects on the diversification strategy of export firms: Evidence from China Junmei ZhangID1,2, Lianying Yao1* 1 Global Institute for Zhejiang Merchants Development, Zhejiang University of Technology, Hangzhou, Zhejiang, China, 2 Institute for Industrial System Modernization, Zhejiang University of Technology, Hangzhou, Zhejiang, China * yaolianying@zju.edu.cn Abstract Using a multi-product trade model, this study investigates the impact of temperature on firms’ diversification in the export market. Using export and meteorological data of Chinese firms from 2000 to 2016, the empirical results confirm an inverted U-shaped relationship, implying that extreme temperatures significantly reduce firms’ export product diversification. The analysis shows that extreme temperatures primarily reduce the variety of both new and existing products, with a less robust effect on product exit strategies. General trade firms are more adaptable to extreme temperatures than processing trade firms, and are likely to maintain diversified strategies. Stronger regional financial markets and higher energy con- sumption increase the adaptability of local firms to extreme temperatures. Firms have not yet adapted to local climatic norms. Furthermore, extreme temperatures also partly inhibit diversification of export destinations and relationships. The results of the study show that as climate change intensifies, leading to more frequent extreme temperatures, firms will face significant hurdles in pursuing diversified development strategies, pointing to increasing

challenges ahead. Introduction As climate change accelerates, extreme temperatures are becoming more common around the world. In 2023, Europe experienced its warmest winter on record, while the northeastern United States experienced a severe cold snap. Summer heatwaves swept across densely popu- lated countries such as the United States, China, and India. The World Meteorological Organi- zation’s 2023 Global Climate Status Report declared it the hottest recorded year on record. Extreme temperatures have widespread economic impacts, causing labor and capital outflows due to adverse working conditions [1–3], and hindering human capital accumulation through reduced productivity and health issues [4,5]. Together, these factors erode regional economic competitiveness by disrupting local industries and deterring new investments. Mitigating the physical risks of climate change is an increasingly urgent global challenge. The prevailing response to climate change is centered on ’mitigation’ and ’adaptation’. Aca- demic research emphasizes reducing emissions to mitigate climate warming, yet the analysis of PLOS ONE PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 1 / 28 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Zhang J, Yao L (2024) Climate effects on the diversification strategy of export firms: Evidence from China. PLoS ONE 19(8): e0308391. https://doi.org/10.1371/journal.pone.0308391 Editor: Baogui Xin, Shandong University of Science and Technology, CHINA Received: March 17, 2024 Accepted: July 23, 2024

Published: August 16, 2024 Copyright: © 2024 Zhang, Yao. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All data files are available from the DOI:10.5281/zenodo.12155977. Funding: Zhejiang Office of Philosophy and Social Science #24NDQN165YBM, awarded to JZ; Zhejiang Province Public Welfare Technology Application Research Project #LQ24G030039, awarded to JZ; National Social Science Fund of China #23BZZ012, awarded to LY; This study is a result of a project funded by the Zhejiang Provincial Social Science Planning. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

economic risks and resilience among various stakeholders in the face of warming remains rela- tively poor. This gap in research constrains the formulation of effective adaptation strategies. This study addresses this deficiency, concentrating on the adaptation of Chinese export firms to the escalating frequency of extreme temperatures attributed to climate change, and discern- ing which firm types are more susceptible. Specifically, we examine the sensitivity of these firms’ diversification strategies to extreme temperature. Our findings aim to provide policy insights by elucidating the influence of extreme temperatures on the diversification strategies of export firms in the world’s foremost goods trading entity. Diversification is an important way for firms to expand their market presence [6]. Upon analyzing China’s export data from 2000 to 2016, we found that more than 80% of firms exported more than one product, which together accounted for 94% of total export value. Diversified firms show resilience in the market by maintaining the flexibility to expand in favorable conditions and by mitigating risk through diversified investments [7]. Conversely, in adverse conditions, firms can maintain profitability by strategically focusing resources on their most competitive core products [8,9]. Dynamic internal resource allocation, achieved by intro- ducing new products and discontinuing less profitable ones, enables effective responses to market fluctuations and unforeseen challenges. In the study, we integrate the temperature variable into a multi-product firm trade model.

Through theoretical analysis, we postulate a fundamental hypothesis of an inverted U-shaped relationship between temperature and the export product diversification of firms. Under- pinned by this hypothesis, we extend our methodology to employ a non-parametric economet- ric model. We leverage comprehensive Chinese customs export data, complemented by daily meteorological data. The principal findings of our empirical inquiry are delineated as follows: First, temperature has a significant inverted U-shaped relationship with export product diversification, with both positive and negative deviations from the optimal temperature reducing the diversity of exported products. The more extreme the temperature, the stronger the effect. For temperatures below -2˚C, each additional day results in a decrease of 0.22 in the number of exported product categories. Likewise, for each day with an average temperature exceeding 27˚C, there is a decrease of 0.09 in the number of exported product categories. Second, processing trade firms show lower adaptability to extreme temperatures than gen- eral trade firms, resulting in a greater reduction in the diversity of their export products. A mature financial market is conducive to improving the ability of local firms to adapt to extreme temperature variations, but an increase in the level of economic development does not seem to have the same effect. Firms located in developed cities tend to reduce the diversity of export products under extreme temperatures. However, increased energy consumption can

mitigate this negative impact. Finally, firms have not adapted to the local temperature norms, as those in hotter regions still experience greater negative impacts from high-temperature shocks. Third, temperature shocks primarily affect the diversification of exported products by influ- encing the entry and survival dynamics of products. Specifically, extreme temperatures lead to a significant reduction in the number of new and existing product categories. Finally, further examination of the influence of temperature reveals an inverted U-shaped relationship with the diversification of both export destinations and relationships (i.e. country-product pairs). However, the effects are asymmetric; low temperatures more robustly impact the diversifica- tion of export market destinations, while high temperatures more significantly affect export market relationships. The existing literature relate to our topic primarily investigates two aspects. Firstly, research on the determinants of firms’ diversification strategies. At the theoretical level, Melitz (2003) [10] introduces a foundational model for homogeneous single-product firms, which Bernard PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 2 / 28 Competing interests: The authors have declared that no competing interests exist

et al. (2010) [11] subsequently extend to encompass multi-product firms. Their research indi- cates that firms adjust their diversification levels in response to various exogenous shocks, including technological advancements and fluctuations in international trade. Additionally, empirical studies explore how market environment changes influence firms’ diversification decisions, considering factors such as trade environment improvements [12–14], financial market development [15–17], and shifts in factor markets [18]. However, current research overlooks the impact of the physical environment, such as temperature, on firms’ export diver- sification decisions. Secondly, research on the economic impacts of temperature shocks. This body of research can be broadly divided into two areas of focus. The first area examines macroeconomic impacts, with some scholars arguing that temperature shocks significantly affect only poor regions [19,20], while others contend that such shocks reduce economic performance across both developed and developing countries [21–23]. The second area focuses on micro-level firm studies, using data from various countries to explore the effects of temperature shocks on firm output [24], profit [25,26], productivity [27–29], employment [30], exports [31,32], and product quality [33]. However, these studies have overlooked firms’ diversification behavior. Given the critical impact of diversification on firms’ internal resource allocation and interna- tional competitiveness, a systematic investigation of this area is essential.

This study makes distinct contributions to the existing literature. Theoretically, we innovatively incorporate temperature factors into the multi-product trade theory model framework of Bernard et al. (2010) [11], deducing a mathematical relationship that indicates an inverted U-shaped rela- tionship between temperature and the diversification of a firm’s export products. Empirically, our study belongs to the realm of micro-level research. Unlike existing literature that primarily focuses on firm-level performance, we are more concerned with whether temperature shocks will alter a firm’s export diversification strategy, specifically if firms will respond by changing the number of export products and export relationships. To investigate this, we use data from Chinese exporting firms. Additionally, we reveal the effect of temperature on intra-firm resource allocation from the perspective of the entry and exit of export products. Finally, we explore the climate adaptability of firms from the perspectives of export modes, regional financial development, economic develop- ment, energy usage, and the geographical location of firms. The subsequent sections of the paper are organized as follows. The Theoretical framework section outlines the theoretical model and provides the conceptual framework for our analysis. The Empirical methodology and data description section details the empirical methodology and provides a comprehensive overview of the data used. The Empirical results section pres-

ents the baseline regression and sensitivity analysis. The Climate adaptation section investi- gates the climate adaptability of firms. The Structural factors in export diversification decline section examines the structural factors influencing export product diversification under extreme temperature. The Extensive research section examines the impact of temperature on the diversification of firms’ export destinations and relationships. The Discussion section addresses the contributions of this paper to the existing literature, its limitations, and potential directions for future research. The Conclusion section summarizes the research findings and further proposing policy recommendations. Theoretical framework This study builds on the multi-product firm trade framework of Bernard et al. (2010) [11]. By incorporating temperature elements, it formulates a theoretical model to explore how temper- ature shocks affect firms’ decisions to diversify their exported products. The specific model is presented below. PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 3 / 28

Consumer side Assuming a representative consumer in a country with a utility function characterized by a constant elasticity of substitution (σ>1), the function is expressed as: U ¼ Z o2O ½qðoÞ� s 1 s doÞ ( ) s s 1 ð1Þ Here, ω represent a specific product category, O is the set of all product categories exported, and q(ω) denotes the quantity of product ω exported. The demand function is derived by maximizing consumer utility, as shown in the subse- quent equation. qðoÞ ¼ p s o Ps 1Y ð2Þ Where, pω is the export price of product ω, P is the price index, P ¼ ð Z o2O p1 s o doÞ 1 1 s , and Y is the national income. Production side Assuming a Cobb-Douglas production function, firms produce product ω using both labor (l) and capital (k). The production function is expressed as follows: qoðTÞ ¼ φoðTÞlakb ð3Þ Where, φω is productivity, T is temperature, a and b are the proportions of labor and capital inputs, satisfying a+b = 1. Productivity is modelled following Chatterjee et al. (2013) [16] as: φoðTÞ ¼ φðTÞZ ðn 1Þ φ ; Zφ > 1 ð4Þ Where, ηφ>1 represents product gradient. The parameter n indicates the ranking of the dis- tance of product ω from the firm’s core competitive product (i.e. the number of product cate- gories). if n = 1, representing the core product, φω decreases as n increases, indicating a greater distance from the firm’s core product. For the firm’s productivity, existing research shows an inverted U-shaped relationship with temperature, revealing an optimal temperature T* at

which productivity peaks [24,27]. Specifically, for T<T*. φ0(T)>0, and for T>T*, φ0(T)<0. Optimal export decision Assuming a fixed export cost fe for product ω, the profit function for the exported product is expressed as: Po ¼ ðpo MCÞ � qo fe ð5Þ Where MC is the marginal cost of producing product ω. The marginal cost is derived from cost minimization: MC ¼ apk b � �b þ b apk � �b � pk " # � 1 φo ð6Þ Where labor wages are normalized to 1, pk denotes the price of capital. Applying the profit maximization condition and combining Formulas (2) and (5), we derive the following PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 4 / 28

expression: po ¼ s s 1 � MC ð7Þ This leads to the optimal profit: P∗ o ¼ P � φo � ðs 1Þ Apkbs � �s 1 � Y s fe ð8Þ Where A ¼ a b �b þ b a �a . When a firm adds products, it increases its sunk costs and reduces its profits [34]. There- fore, if a firm produces n products, it is expected that product n* should satisfy P∗ o ¼ 0. If this condition is not met, the firm would continue to produce new products. Combining Eq (4) with the zero-profit condition, we obtain the following equation: Z1 n∗ φ ¼ 1 φðTÞ � Apk b ðs 1ÞP � s Y � � s s 1 fe 1 s 1 ð9Þ Extreme temperatures and export product diversification According to Eq (9), the derivative of the number of exported product categories (n*) with respect to temperature gives the following equation: @n∗ @T ¼ φ0ðTÞ φðTÞ � lnZφ ð10Þ As φ(T) and lnηφ are both positive, the sign of n*0(T) depends on the φ0(T). Thus, for T<T*, indicating temperatures below the optimum, the number of exported product catego- ries n* increases with increasing temperatures, promoting diversification. Conversely, for T>T*, temperatures above the optimum lead to n*0(T)<0. In this scenario, increasing temper- atures lead to a decrease in the number of exported product categories, reducing diversifica- tion. In summary, export diversification has an inverted U-shaped relationship with temperature. Extreme temperatures have an unfavorable effect on the implementation of diversification strategies by firms. Empirical methodology and data description Empirical methodology

Following Deschênes and Greenstone (2007) [35], this study uses a non-parametric model to analyze the impact of daily extreme temperature increases on firms’ export diversification strategies. The use of this model has several advantages: Firstly, the non-parametric model’s flexibility in specification enables the examination of the non-linear relationship between tem- perature and export diversification without necessitating the pre-specification of a functional form. Secondly, the model includes a series of temperature bin variables with interval lengths of 5˚C. These variables cover all temperature bins, ranging from the number of days with tem- peratures below -7˚C, the number of days between -7˚C and -2˚C, and so forth, up to the num- ber of days with temperatures above 28˚C. This illustrates the distribution of daily temperatures over a year. For instance, if the temperature on a particular day is 0˚C, that day will be counted in the -2˚C to 3˚C temperature bin. This methodology facilitates the direct observation of the impacts of extreme temperatures, both high and low. In contrast, when PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 5 / 28

using a parametric model, the annual average temperature can mask the effects of daily extreme temperatures, significantly reducing the information contained in the estimation results. In the following sections, we conduct robustness tests using a parametric regression model. Specifically, we construct the non-parametric model as follows. Diversityicrt ¼ a þ X m6¼5bm 1 Tbinm crt þ d1Xcrt þ gi þ grt þ εicrt ð11Þ The indices i, c, r, and t represent firms, cities, provinces, and years, respectively. The dependent variable, Diversityicrt, indicates the diversity of products exported by firm i, mea- sured by the number of product categories exported by a firm. The core explanatory variables, Tbinm ct , denote the number of days in city c in year t with a daily average temperature within the temperature bin m. Temperature bin variables are defined with a length of 5˚C, for a total of 9 bins. Where, Tbin1 ct represents days with a daily average temperature below -7˚C. For example, if a city experiences temperatures below -7˚C for 10 days in a year, then the value of Tbin1 ct is 10. Tbin2 ct represents days with a temperature in the bin [-7, -2)˚C. Tbin3 ct repre- sents days with a temperature in the bin [-2, 3)˚C, and so on, with the final bin, Tbin9 ct, indi- cating days with a daily average temperature �28˚C. The regression uses the temperature bin (8, 13˚C (m = 5) as the base group. Given that the annual total of days is fixed, the cumulative days across all temperature bins remains constant. To mitigate multicollinearity in the regres-

sion analysis, one temperature bin must be designated as the baseline, thereby excluding it from the set of variables. This baseline selection is critical as it should ensure that the coeffi- cients associated with all other temperature bins are negative, thereby signifying it as the opti- mal temperature range [4,24]. In accordance with this criterion, this study selects the temperature range of 8, 13˚C as the baseline. The coefficient βm 1 is of primary interest, repre- senting the change in product diversity for each additional day with a daily mean temperature in the mth bin relative to the baseline. We expect βm 1 to be consistently negative, reflecting an inverted U-shaped trend based on our theoretical model. Control variables, denoted by X, include city-level and temperature-related factors, such as precipitation (Prec), represented by the annual mean precipitation in the city, wind speed (Wind), represented by the annual mean wind speed in the city, sunshine duration (Sun), rep- resented by the annual mean sunshine duration in the city, humidity (Wet), represented by the annual mean relative humidity in the city, and atmospheric pressure (Atmo), represented by the annual mean atmospheric pressure in the city. To reduce omitted variable bias, we include firm (γi) and province-year (γrt) fixed effects in the model. ε represents the error term. Stan- dard errors are two-way clustered at the firm and city-year levels [36]. Data description This section outlines the use of two key datasets: meteorological data and firm export data.

The meteorological data, obtained from the National Meteorological Information Center, cov- ers daily observations from 1 January 1951 to the present, including variables such as daily average temperature, maximum and minimum temperatures, 20-20-hours precipitation, aver- age relative humidity, and average wind speed. The article focuses on data from 833 continu- ously operating meteorological stations between the years 2000 and 2016. Using a nearest- neighbor approach based on station coordinates and county-level centers, each county is matched with its closest meteorological station. City-level meteorological data are derived by averaging data from all stations within the same city. Firm export data are mainly taken from the China Customs database (2000–2016). It con- tains import/export information at firm-product-destination-year level. The product codes in the customs database have undergone changes in various years, such as 2002, 2007, and 2011. PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 6 / 28

We have standardized the codes to the HS2002 version and defined product categories based on the first six digits of the HS2002 version. For the purposes of this study, only export data is kept and the data are aggregated at the firm-year level. Therefore, we can obtain detailed firm export behavior data at the firm-city-year level, including the number of export product cate- gories, the number of export destinations, and the number of export relationships for each firm. The China Customs database reports the cities where exporting firms are located, and we have also obtained city-level meteorological data. Therefore, we match the two databases using city codes. Table 1 presents descriptive statistics for the variables used in this study. Firms show con- siderable diversity in the number of product categories exported. Some firms export only one product, while others export up to 4285 product categories at the 6-digit HS2002 level. The cli- mate data also show considerable variation between China’s cities. For instance, some cities have average daily temperatures below -7˚C for more than 149 days a year, while others have temperatures consistently above -7˚C. Similarly, in terms of high temperatures, some cities have 163 days a year when temperatures do not fall below 28˚C, while others have average daily temperatures below 28˚C. These variations in both export strategies and climatic condi- tions across different firm locations provide a robust empirical basis for our research. Empirical results Baseline regression

We examine the relationship between temperature and firms’ export product diversity using Eq (11). Table 2 shows the baseline regression results. In column (1), without controlling for additional variables and fixed effects, a U-shaped relationship between temperature and export diversity is observed, but with a modest R2 of 0.11, indicating the presence of an important uncontrolled effect. Introducing firm and province-year fixed effects in column (2) reveals pre- dominantly negative and significant coefficients for temperature bins. The R2 increases to 0.737, highlighting the need to control for the fixed effects in Eq (11). In column (3), when other meteorological variables that are highly correlated with temperature are included in the model, all temperature bins have negative and statistically significant coefficients, showing an Table 1. Summary statistics. Variables Obs Mean SD Min Max Diversity 3,149,535 14.43 49.97 1 4285 <-7˚C 3,126,982 4.872 16.61 0 149.4 -7~-2˚C 3,126,982 8.739 14.89 0 93.82 -2~3˚C 3,126,982 22.41 20.40 0 99.37 3~8˚C 3,126,982 36.86 22.17 0 98.67 13~18˚C 3,126,982 43.44 15.26 0 113.33 13~18˚C 3,126,982 52.76 12.41 0 154.2 18~23˚C 3,126,982 70.25 13.05 0 181.7 23~28˚C 3,126,982 81.91 27.29 0 224 �28˚C 3,126,982 44.04 27.33 0 163 Prec 3,126,982 3.452 1.504 0.0529 8.594 Wind 3,126,982 2.427 0.548 0.546 5.692 Sun 3,126,982 5.268 0.924 2.017 9.694 Wet 3,126,982 4.261 0.0972 3.322 4.515 Atmo 3,126,982 100.4 2.553 59.24 101.7

https://doi.org/10.1371/journal.pone.0308391.t001 PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 7 / 28

inverted U-shaped relationship (see Fig 1). Firms tend to reduce export diversity as tempera- ture deviates from the optimum, and the reduction intensifies with larger deviations. The above results are consistent with the theoretical model estimates. In terms of economic significance, each additional day below -2˚C reduces the number of product categories exported by firms by 0.22 (0.1227+0.0937). On average, Firms face 13.6 days below -2˚C per year, resulting in a reduction of 2.94 categories. Similarly, each additional day �28˚C reduces the number of exported categories by 0.09. Firms experience an average of 44 days �28˚C per year, resulting in a reduction of 4 categories. Extreme temperature shocks due to climate change warrant increased attention from an economic perspective for firms and nations. Table 2. Baseline regression results. Variables (1) (2) (3) Diversity Diversity Diversity <-7˚C 0.1976*** -0.1196*** -0.1227*** (0.003) (0.044) (0.044) -7~-2˚C 0.0571*** -0.0855*** -0.0937*** (0.004) (0.033) (0.034) -2~3˚C -0.0076** -0.0682** -0.0687*** (0.003) (0.027) (0.026) 3~8˚C 0.0477*** -0.0332 -0.0361* (0.004) (0.022) (0.021) 13~18˚C -0.0091** 0.0102 -0.0076 (0.004) (0.021) (0.019) 18~23˚C 0.2002*** -0.0827** -0.0984** (0.003) (0.040) (0.040) 23~28˚C 0.0758*** -0.0600* -0.0855** (0.003) (0.032) (0.033) �28˚C 0.1702*** -0.0374 -0.0885*** (0.003) (0.029) (0.033) Prec 0.0089 (0.283) Wind -0.9089 (0.641) Sun -0.5178 (0.373) Wet 2.1966 (3.656) Atmo 6.9357*** (1.781) Constant -15.8563*** 30.3267*** -663.8836***

(0.855) (7.395) (174.911) Firm FE No Yes Yes Province×Year FE No Yes Yes Observations 3,125,406 3,015,824 3,015,824 R2 0.011 0.737 0.737 Note: *, **, *** represent statistical significance levels of 10%, 5%, and 1%, respectively. The parentheses in columns (2)-(3) contain standard errors clustered at the firm and city-year levels. In the subsequent text, unless otherwise specified, the parentheses will indicate standard errors clustered in two-way style. https://doi.org/10.1371/journal.pone.0308391.t002 PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 8 / 28

Sensitivity analysis Endogeneity issue. This section considers the potential problem of omitted variables, selection bias, and bidirectional causality in the model specification. Firstly, we discuss the potential problem of omitted variables. In the baseline regression, we follow the common prac- tice in the existing literature by controlling for other meteorological variables that are highly correlated with temperature in the model [27]. We also include province-year fixed effects to control for regional economic development levels and policy effects, and firm fixed effects to control for time-invariant firm characteristics. The above model is designed to mitigate the issue of omitted variable problem. To further validate the robustness of the conclusions, we extend the model (11) by adding city-year level economic variables, including: economic development level (Eco), represented by the logarithm of the city’s GDP; population size (Pop), represented by the total population at the end of the year; green innovation capacity (Envinno), represented by the logarithm of the number of green invention patents filed in the city in the current year; city openness (Open), represented by the logarithm of the actual foreign invest- ment in the city in the current year; and local government capacity (Gov), represented by the share of local fiscal budget expenditure in local GDP. These additional control variables cover aspects such as market, government, factor endowment, and technological development, pro-

viding a comprehensive control for the economic development trends at the city-year level. The regression result with these additional variables is shown in column (1) of Table 3. The result shows that the addition of these control variables does not change the conclusions of the baseline regression, indicating that the baseline model has good explanatory power. Given that Dell et al. (2014) [37] point out that the impact of climate on economic markets is multifaceted, controlling for economic variables may lead to “over-control” problems. Fig 1. The U-shaped relationship between temperature and the product diversification of export firms. The graph is created based on the regression coefficients and standard deviation data for each temperature bin from column (3) of Table 2. https://doi.org/10.1371/journal.pone.0308391.g001 PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 9 / 28

Therefore, we alternatively control for the impact of city development trends by adding a city- year interaction term as a control variable in model (11). The result is shown in column (2) of Table 3, where we find that temperature still has an inverted U-shaped relationship with export product diversification. These results suggest that the omitted variable problem does not affect the basic conclusions of this study. Subsequently, we address selection bias, encompassing both sample selection bias and self- selection bias. Our sample comprises all exporting firms registered in Chinese customs data, thereby minimizing sample selection bias. There are two problems with self-selection bias. Firstly, there is bias from unobserved firms that did not enter or exit the export market. Since data on the product categories of firms selling domestically are not available, we cannot inves- tigate this issue empirically. However, logically, firms that do not enter or exit the export mar- ket are typically those that are eliminated due to lower competitiveness and risk resistance. Therefore, extreme temperatures are likely to have a greater negative impact on the diversifica- tion strategies of these firms, potentially underestimating the overall impact of temperature. Therefore, this bias doesn’t affect the conclusion of an inverted U-shaped relationship between temperature and export product diversification. Secondly, self-selection bias arises from firms’ location choices. Temperature-sensitive firms may prefer regions with comfortable tempera-

tures when choosing a location. To address this, we examine the effect indirectly. Firms are likely to consider temperature conditions in different regions when selecting locations, sug- gesting a spillover effect of temperature across regions. To assess this, we use the inverse of city distances as weights to calculate the weighted average of temperature variables for other cities, constructing spatial lag terms for each temperature bin. The results in column (3) of Table 3 Table 3. Endogeneity issue. (1) (2) (3) Variables Diversity Diversity Diversity <-7˚C -0.1453*** -0.1181*** -0.1061* (0.047) (0.044) (0.056) -7~2˚C -0.1001*** -0.0891*** -0.1003** (0.035) (0.033) (0.047) -2~3˚C -0.0704*** -0.0641** -0.0766** (0.027) (0.026) (0.034) 3~8˚C -0.0319 -0.0317 -0.0407 (0.021) (0.020) (0.029) 13~18˚C 0.0015 -0.0087 0.0064 (0.019) (0.019) (0.026) 18~23˚C -0.1020** -0.1057*** -0.0914** (0.040) (0.040) (0.046) 23~28˚C -0.0934*** -0.0962*** -0.0835** (0.033) (0.034) (0.041) �28˚C -0.0920*** -0.1057*** -0.0876** (0.033) (0.035) (0.039) CV Yes Yes Yes FE Yes Yes Yes Observations 2,922,853 3,015,824 3,015,824 R2 0.741 0.737 0.737 Note: CV includes all unreported control variables as well as the constant term. FE represents fixed effects, including both firm-specific and province-year fixed effects. Unless explicitly stated, the following tables follow this note. https://doi.org/10.1371/journal.pone.0308391.t003 PLOS ONE Climate effects on the diversification strategy of export firms

PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 10 / 28

show that after controlling for this spillover effect, the study’s basic conclusion remains unchanged. Finally, we turn our attention to the potential for bidirectional causality. In the time frame of our study (annual temperature distribution), temperature changes are primarily driven by natural factors and exhibit random fluctuations. In the existing literature, temperature is typi- cally treated as an exogenous variable (Dell et al., 2014) [37]. However, caution is warranted regarding the possibility of unobservable factors that simultaneously affect both temperature and firm behavior, such as certain government environmental policies. To further address the impact of unobservable factors, we employ a placebo test by randomly shuffling the core explanatory variable 100 times and observing whether the temperature variable continues to significantly influence firms’ export decisions. A significant result would indicate the presence of unobservable factors, while an insignificant result would indicate the robustness of our con- clusions. We run regressions after shuffling the variables for temperatures �28˚C, 23~28˚C, <-7˚C, and -7~-2˚C, respectively, and examine the distribution of the regression coefficients, as shown in Fig 2. The results show that after shuffling, the regression coefficients for both high and low temperatures are distributed around zero and are no longer significant, which is Fig 2. Placebo test. Fig 2 displays the research results of the four groups, which respectively illustrate the distribution of regression coefficients resulting from

100 random permutations for variables �28˚C, 23~28˚C, <-7˚C, and -7~-2˚C. https://doi.org/10.1371/journal.pone.0308391.g002 PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 11 / 28

significantly different from the baseline regression results. This suggests that unobservable fac- tors do not affect our conclusions, lending credibility to our findings. Model specification adjustment In the baseline regression, we use an ordinary least squares (OLS) regression to examine the relationship between temperature bins and export product diversification. Given the count integer nature of the dependent variable (number of export product categories), we adopt a count model in this section and use negative binomial regression to re-evaluate the relation- ship between temperature bins and export diversification (see Table 4, column 1). The consis- tently negative regression coefficients across all temperature bins show an inverted U-shaped relationship, confirming the baseline findings. Notably, this baseline conclusion remains robust to a change in the econometric methodology. In addition, we explore the non-linear relationship between temperature and export product diversification using a non-parametric model in the benchmark regression setting. In this section, we construct the following com- monly used parametric model to investigate the relationship between temperature and export product diversification. Diversityicrt ¼ a þ b1Tempicrt � Tempicrt þ b2Tempicrt þ d1Xcrt þ gi þ grt þ εicrt ð12Þ Tempirct represents the annual average temperature of city c for firm i in year t, and the meanings of the other variables remain consistent with the previous text. The focus is on the Table 4. Sensitivity analysis. (1) (2) (3) (4) (5)

Variables Diversity Diversity Diversity Diversity Diversity Temp×Temp -0.0920** (0.042) <-7˚C -0.0112*** -0.1082*** -0.1227* -0.1339** (0.000) (0.035) (0.068) (0.064) -7~2˚C -0.0072*** -0.0885*** -0.0937* -0.1213** (0.000) (0.028) (0.055) (0.050) -2~3˚C -0.0058*** -0.0644*** -0.0687* -0.0859** (0.000) (0.022) (0.041) (0.039) 3~8˚C -0.0029*** -0.0353* -0.0361 -0.0492 (0.000) (0.018) (0.034) (0.034) 13~18˚C -0.0028*** -0.0136 -0.0076 -0.0078 (0.000) (0.017) (0.013) (0.023) 18~23˚C -0.0101*** -0.0872** -0.0984** -0.1024** (0.000) (0.036) (0.049) (0.052) 23~28˚C -0.0094*** -0.0695** -0.0855* -0.0942** (0.000) (0.030) (0.044) (0.046) �28˚C -0.0108*** -0.0664** -0.0885** -0.1347** (0.000) (0.030) (0.044) (0.056) Temp 2.1621 (1.327) CV Yes Yes Yes Yes Yes FE Yes Yes Yes Yes Yes Observations 3,015,824 3,015,824 3,015,824 3,015,824 3,015,824 R2 0.737 0.763 0.737 0.737 https://doi.org/10.1371/journal.pone.0308391.t004 PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 12 / 28

regression coefficient β1. When β1 is significantly negative, it indicates an inverted U-shaped relationship between temperature and export product diversification, suggesting that both very low and very high temperatures inhibit firms’ diversification strategies. The regression results are presented in column (2) of Table 4, where β1 is indeed significantly negative, con- firming the robustness of the baseline conclusion. Outlier impact mitigation. To mitigate the potential influence of extreme values, we implement winsorization on the dependent variable, a technique that trims data points at the 0.5% and 99.5% percentiles. This process aims to reduce the influence of firms with extreme levels of diversification. The regression analysis using the modified sample data, based on Eq (11), is presented in Table 4, column (3). The results confirm the baseline conclusion, with sta- tistically significant negative regression coefficients observed consistently across all tempera- ture bins. Changing the cluster level of standard errors. In the baseline regression, we account for the potential autocorrelation present in the error terms at both the firm and city-year levels. Subsequently, we refine our approach by relaxing this assumption and acknowledging the pos- sibility of intra-city correlation among error terms. To address this, we adjust the standard errors to be clustered at the city level, as illustrated in Table 4, column (4). This slightly reduces the statistical significance, but maintains the inverted U-shaped relationship between tempera-

ture and export diversification. Most temperature bins pass statistical significance tests at the 10% level, underscoring the robustness of the baseline regression conclusions. Weekday data examination. In accordance with Chinese labor laws, a two-day weekend system is implemented on Saturdays and Sundays, concentrating production and labor activi- ties mainly on weekdays. Consequently, we focus on weekdays for a more nuanced analysis of temperature effects. Eq (11) is regressed again using temperature bin variables constructed from Monday to Friday data, as shown in Table 4 column (5). The results consistently show negative regression coefficients for all temperature bins, with the majority achieving statistical significance at the 5% level, consistent with previous analyses. Notably, the absolute values of the regression coefficients increase compared to the baseline, emphasizing that temperature predominantly affects export diversification mainly during weekdays. Altering temperature bins length. In the baseline regression, we divide temperature into 5˚C bins. Here, we refine our assumption by assuming a uniform effect on export product diversification within 3˚C temperature bins. Under this premise, we redefine temperature bins ranging from days below -6˚C (m = 1) to days �27˚C (m = 13). Specifically, if m = 1, it repre- sents the number of days in a year with an average daily temperature below -6˚C. m = 2 repre- sents the days with an average daily temperature between -6 and -3˚C, and so on. if m = 13, it

means the days with daily average temperature �27˚C. The regression with 12~15˚C bin as the baseline group shows an inverted U-shaped relationship between temperature and export diversification. The regression coefficients in Fig 3 show a trend of initial increase and subse- quent decrease with increasing temperatures, confirming the baseline conclusions. This con- firms that the length of the temperature bin specified in the model does not affect the research results. Climate adaptation Following the Industrial Revolution, rapid technological advances have provided multiple opportunities for climate adaptation. The widespread adoption of indoor work environments, coupled with innovations such as air conditioning and heating systems, enables individuals to regulate ambient temperatures, thereby increasing comfort and potentially productivity, miti- gating the impact of climate variability. At the same time, improved financial systems, PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 13 / 28

including insurance and credit, facilitate intertemporal risk management, thereby reducing the impact of climate-related risks. However, climate adaptation entails significant costs and necessitates a cognitive framework that aids in understanding and addressing the complexities of climate change. For Chinese exporters, the extent of adaptation to local temperature condi- tions and the factors influencing the ability to adapt to climate change remain key issues. In the following sections, we address these issues through the lens of firms’ export diversification. The impact of a firm’s export trade mode China’s integration into the global value chain through its processing trade regime has been profound, significantly facilitating the transfer of industries from developed nations to develop- ing nations and enhancing the optimal allocation of global resources. This integration has markedly propelled China’s export growth [38]. Within the realm of international trade research, the "productivity puzzle" of Chinese export firms has generated considerable debate. Studies have revealed that, excluding processing trade firms, the productivity of Chinese export firms surpasses that of non-export firms, thereby resolving the productivity puzzle. This sug- gests that Chinese processing trade firms possess unique characteristics; their engagement in exports is not predicated on higher productivity but rather on an order-based production model that revolves around "importing raw materials and exporting finished products." These

firms are typically labor-intensive, exhibit low added value, and are highly substitutable, situat- ing them at the lower end of the international value chain. In contrast, general trade firms tend to be capital-intensive, prioritize brand development, operate on a larger scale, exhibit higher productivity, and face less restrictive financing constraints (Manova and Yu, 2016) [39]. Acknowledging the pronounced differences between these firm types, our analysis catego- rizes the sample into two distinct groups: processing trade firms (Ptrade group) and general trade firms (Gtrade group), conducting separate regressions for each. The outcomes of the group regressions are detailed in columns (1) and (2) of Table 5-I. The analysis shows that Fig 3. Impact of temperature on export diversification. https://doi.org/10.1371/journal.pone.0308391.g003 PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 14 / 28

Table 5. I. Climate adaptation. II. Climate adaptation. Ш. Climate adaptation. (1) (2) (3) (4) Variables Ptrade Gtrade Underdevelope Advance <-7˚C -0.2423*** -0.0083 -0.2419*** -0.0019 (0.076) (0.026) (0.075) (0.032) -7~2˚C -0.1422*** -0.0146 -0.1516** -0.0184 (0.055) (0.018) (0.061) (0.023) -2~3˚C -0.1062** -0.0038 -0.1090** -0.0109 (0.042) (0.015) (0.043) (0.017) 3~8˚C -0.0526* -0.0056 -0.0725** 0.0096 (0.030) (0.011) (0.034) (0.012) 13~18˚C 0.0072 0.0183 0.0003 -0.0277** (0.025) (0.019) (0.030) (0.013) 18~23˚C -0.1533** -0.0271 -0.1593*** -0.0055 (0.063) (0.019) (0.058) (0.018) 23~28˚C -0.1271** -0.0515* -0.1252*** -0.0101 (0.052) (0.027) (0.046) (0.021) �28˚C -0.1258** -0.0630* -0.1325*** -0.0443* (0.054) (0.034) (0.047) (0.024) CV Yes Yes Yes Yes FE Yes Yes Yes Yes Observations 1,630,475 1,105,352 1,502,227 1,464,566 R2 0.799 0.754 0.733 0.772 (1) (2) (3) (4) Variables Developing Developed Lenergy Henergy <-7˚C -0.0290 -0.9541*** -61.4848** -0.1229 (0.038) (0.262) (25.898) (0.184) -7~2˚C -0.0205 -0.6688*** -0.6884 -0.1681 (0.024) (0.199) (0.459) (0.154) -2~3˚C -0.0288* -0.4899*** -0.5962** -0.1130 (0.016) (0.154) (0.237) (0.144) 3~8˚C 0.0085 -0.4106*** -0.5227*** 0.0529 (0.013) (0.134) (0.172) (0.112) 13~18˚C 0.0162 -0.0185 0.0265 -0.2330** (0.012) (0.079) (0.092) (0.113) 18~23˚C 0.0091 -0.2476** -0.2500* -0.2280 (0.016) (0.121) (0.142) (0.148) 23~28˚C 0.0122 -0.2516** -0.2647 -0.2070 (0.018) (0.122) (0.165) (0.134)

�28˚C 0.0039 -0.4037*** -0.4060** -0.1417 (0.019) (0.144) (0.165) (0.124) CV Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Observations 1,505,678 1,463,285 882,233 561,632 R2 0.703 0.760 0.776 0.748 (1) (2) (3) (4) Variables Northern Cities Southern Cities Cold Cities Hot Cities <-7˚C -0.0281 0.0711 -0.0372 3.2633 (0.042) (0.161) (0.037) (2.762) (Continued) PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 15 / 28

fluctuations. Moreover, their minimal investment in R&D could suggest a limited capacity for innovation in climate change adaptation technologies. This limitation can impede the adop- tion of new technologies and the formulation of responsive strategies, as well as restrict the firms’ flexibility to diversify their products and markets. As a result, in the face of increasingly frequent or severe extreme climatic conditions, processing trade firms may struggle to adapt their business strategies to sustain competitiveness. This challenging scenario could compel them to concentrate more on their existing, constrained product lines or markets to minimize the need for new investments and R&D. These findings highlight that processing trade firms represent a critical vulnerability in the climate adaptation of Chinese export firms. Future Table 5. (Continued) (1) (2) (3) (4) Variables Ptrade Gtrade Underdevelope Advance -7~2˚C -0.0143 -0.0636 -0.0213 0.0346 (0.031) (0.064) (0.026) (0.263) -2~3˚C 0.0070 -0.0971*** -0.0174 -0.0882 (0.025) (0.037) (0.020) (0.055) 3~8˚C 0.0236 -0.0564** 0.0007 -0.0497 (0.022) (0.028) (0.016) (0.040) 13~18˚C -0.0000 -0.0142 0.0149 -0.0292 (0.024) (0.024) (0.017) (0.032) 18~23˚C 0.0225 -0.1361*** 0.0298 -0.1737*** (0.030) (0.050) (0.023) (0.061) 23~28˚C 0.0145 -0.1174*** 0.0079 -0.1464*** (0.032) (0.042) (0.025) (0.050) �28˚C 0.0016 -0.1198*** -0.0055 -0.1475*** (0.032) (0.042) (0.027) (0.049) CV Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Observations 823,240 2,192,584 1,615,926 1,395,795

R2 0.671 0.750 0.714 0.751 https://doi.org/10.1371/journal.pone.0308391.t005 PLOS ONE Climate effects on the diversification strategy of export firms PLOS ONE | https://doi.org/10.1371/journal.pone.0308391 August 16, 2024 16 / 28