The Effect of Climate Change on Firm Performance and the Role
of Internationalization
Abstract
This paper investigates the effect of climate change on firm performance and the moderating effect of internationalization on this relation. Previous studies lacked a conclusive answer whether climate change has an effect on firm performance or not. Besides that, previous researches do not include the effect of internationalization into their studies. To test the effect of climate change, a panel data analysis is done with a sample consisting of 12,917 firm-year observations of 2,057 U.S. firms over the period 1996-2016. Using damages caused by extreme natural events as a proxy for climate change, we find that climate change has a negative effect on the level of firm performance. The effect is mitigated by
internationalization. Additionally, the results show that the negative effect of climate change is not the same across industries.
Keywords: climate change, finance, firm performance and internationalization
Julian Lo (S2727870) Supervisor: dr. A. Dalò
1. Introduction
In 1999, U.S. was hit by a drought causing over US$ 1 billion in losses and killing over 700 people. A few years later, in 2005, the devastating hurricane Katrina struck the U.S., leaving over US$ 100 billion in damages. In addition, 1,833 people were killed in the event. These events are just two examples of the tremendous impact of such a disaster on the community and the economy. During the last decades, extreme natural events increasingly occur in all parts of the world. An extreme natural event (hereafter ENE) is defined as a rare and severe natural process that could affect people and property. There were globally 77 ENEs reported in 1968. Half a century later, this number increased 3,5 times (Ritchie and Roster, 2019). The increase in ENEs is linked to the rising global temperature, which refers to global warming (Environmental Defense Fund, 2019). According to the United Nations (2019), the main driver of global warming is human activities, deforestation and agriculture, and in particular, the emission of greenhouse gases. Greenhouse gases remain in the atmosphere, which increases the ability of the Earth to contain its warmth. Hence, the global temperature rises (United Nations, 2019). Global warming causes increasing maximum and minimum temperature, increasing frequency and intensity of precipitation, increasing frequency of tropical cyclones and increasing wind intensities (Harvard Medical School, 2005). As a consequence, climates are changing, and the number of ENEs is rising.
As response on the changing climates, in 2016, 195 states have signed the Paris Agreement to collectively reduce the emission of greenhouse gases. The aim of the agreement is to keep the increase in the global temperature well below two degrees Celsius above pre-industrial levels, and in the future even limits the increase to 1.5 degrees Celsius. To reach these goals, investments in climate finance are
alone does not have enough resources to cover all the expected costs. Therefore, the public sector encourages and increasingly funds private capital investments in climate change projects in developing countries (World Resources Institute, 2019). Yet, it is high time that the private sector takes the initiative. Fortunately, acknowledgment and awareness of global warming are increasing, and institutions and agreements are increasingly established to decrease the human footprint on the environment (UN, 2019). Nevertheless, little research is done regarding the effect of climate change on firm performance.
Research shows that ENEs negatively affect physical assets and its productivity (Hallegatte, Hourcade, and Dumas, 2005). Preston (2012) also suggests that losses from ENEs will grow by a factor 1.3 to 1.7 by 2025 and 1.8 to 3.9 by 2050 based on U.S. annualized loss of approximately $10-13 billion. Similarly, Klomp and Vackx (2013) indicate a growing impact of climate change on firm performance. The destruction of assets will lower firm’s capital and consequently, decreases its liquidity. When many firms experience the same decrease in liquidity, it can be harmful for financial stability of nations and in turn, to the world economy (Dafermos, Nikolaidi, and Galanis, 2018). On the contrary, other studies (e.g. Noth and Rehbein, 2019; Strulik and Trimborn, 2019) do not show a negative effect of this on the firm performance in the long run. Some even show a positive effect of climate change on performance.
While many studies (e.g. Hallegatte et al., 2005) have emphasized the negative effects of climate change on firm performance, some other studies (e.g. Strulik and Trimborn, 2019) find little to no negative effects. Therefore, the main research question is:
What is the effect of climate change on the firm performance?
observations of 2,057 U.S. firms over the period 1996-2016. In addition, the study is focusing on ENEs that can damage property or harm people. According to the Harvard Medical School (2005), several ENEs are expected to increase in frequency and magnitude due to climate change. Hence, the following ENEs are included: Tropical storm, Droughts, Severe storm, and Floods.
Using a panel data regression, our findings show that climate change has a negative effect on firm performance as measured by return on assets and return on equity. We suggest that the profitability is reduced due to the additional expenses caused by the ENEs, while firm value is not affected. In addition, evidence suggests that the effect of climate change is moderated by the type of industry. Thus, some industries are affected more than others by climate change. At last, the findings show a mitigating effect of internationalization on the main relation. Thus, we suggest that being more global does reduce the negative effect of climate change on firm performance. Additional proxies of firm performance and climate change are used to ensure the robustness of the results. The result of these checks remains the same.
These findings add evidence to the existing literature on the impact of climate change on firm performance. Noth and Rehbein (2019) show a positive relation between ENEs and firm performance, while only focusing on a single ENE. This paper includes multiple events to capture the effects of climate change on the assumption of the United Nations (2019), that climate change will intensify and initiate more ENEs. Hence, it is expected that firms will have less time and less capital to recover from an ENE. Furthermore, Strulik and Trimborn (2019) study the effects of ENEs on the macro-level performance, while this paper is focused on firm-level performance. According to the study, the annual output is not harmed by ENEs, while additional expenses are increasingly incurred. Hence, we assume that
profitability would decrease. To the best of our knowledge, no past studies have included the effect of internationalization into their studies to explain the effect of climate change on firm performance.
discussed, and section 4 presents the main results and provides some interpretation. In the last section, the findings will be concluded.
2. Literature review
2.1. Global warming and extreme natural events
Climate change is related to the increasing frequency and magnitude of ENEs. Global warming increases the average temperature of our planet. This is mainly caused by the emission of greenhouse gases, which trap Earth’s warmth. As a result of the increasing temperature, the cryosphere is melting, and more water is evaporated. While water vapor causes humidity, it also has the ability to trap warmth. Higher humidity increases the probability of intense tropical precipitation. In the U.S., the water vapor level increased 15% at the end of the 20th century. The increasing temperature increases water
evaporation and parching some parts of the Earth’s surface, that will result in low-pressure pockets. In turn, this pulls in wind and weather. Due to the increase in global temperature, some areas become drier, while other areas experience recurrent floods (Harvard Medical School, 2005). Hence, the frequency and the magnitude of ENEs are increasing.
2.2. Climate change and firm performance
2.2.1. Positive effects
Strulik and Trimborn (2019) show that the GDP and the economy’s output are not necessarily lower after a natural disaster. GDP is even higher than pre-disaster level, when the disaster mainly destroys durable consumption goods, e.g. cars and furniture. If mainly productive capital, e.g. raw materials, is destroyed, the GDP is lower than the pre-disaster level. Since an ENE does not selectively destroy goods, both mechanisms occur and will most likely cancel each other out. While GDP and output are not
harmed, it does decrease the aggregate wealth (Strulik and Trimborn, 2019). Another research (Noth and Rehbein, 2019) concludes that firm performance increases in the direct aftermath of a natural disaster. This could be caused by the sustainable recovery a firm made, which improves the business performance (Pielke, 2007). Although this positive conclusion, they do stress the importance of financial support, since it is required to overcome the damages made by the extreme weather. This is probably also one of the reasons why developing countries are affected more by a disaster than developed countries (Klomp and Valckx, 2013). In addition, the learning curve of a firm plays a minor role in the effect of subsequent disasters. For example, German firms that were hit by a flood in 2002 and 2013 had higher turnovers, lower leverage and higher cash reserves in the period after the disaster. Noth and Rehbein (2019) conclude that this recovering ability must be the result of experiencing an earlier flood in 2002.
Surprisingly, studies show some positive effects of a natural disaster on firm performance. However, in both situations there is a cost, namely the loss aggregate wealth or the use of financial support (Noth and Rehbein, 2019; Strulik and Trimborn, 2019).
2.2.2. Negative effects
socioeconomic exposure of different areas in the estimation of potential losses caused by ENEs. The predicted loss will grow by a factor 1.3 to 1.7 by 2025 and 1.8 to 3.9 by 2050. The calculations are based on the annualized loss of US$ 10-13 billion. Hence, this will lower the productivity and liquidity of firms, which increases the default rates and put the financial stability in danger (Carney, 2016; Dafermos et al., 2018). These forecasts would counteract the effect of financial support of Noth and Rehbein (2019), because their theory relies on the recovering ability of firms and financial support. On average, the
national consumption level requires six years to recover from any kind of rare disaster, e.g. financial crisis and war (Nakamura, Steinsson, Barro, and Ursúa, 2013). If ENEs happen more frequently, the time to recover will decrease and more importantly, the financial resources will be depleted. Thus, the most suitable strategy could be preventing rather than solving by leaving the country (Dai, Eden, and Beamish, 2017).
Furthermore, Gourio (2012) shows that an increase in disaster risk leads to lower expectations of investors, while increasing the overall risk of the company. In addition, an increase in disaster risk leads to a decline in employment, output, and stock prices. Hence, investors demand a higher risk premium to compensate for the additional disaster risk. This will decrease the aggregate wealth and increase the risk of financial instability (Dafermos, 2018; Strulik and Trimborn, 2019).
Besides the direct effects of climate change on firm performance, there could also be several
unexpected impacts that deteriorate firm performance. Climate change can for example cause reduction of air quality, increasing food shortage, or increasing epidemics. Therefore, indirect effect could cause problems in well-being of the citizens, nature, and infrastructure (Harvard Medical School, 2005). As a result, liability risks and transition risks could increase. The collective goal is towards a lower-carbon economy. As evidence gradually toughens, parties will be held liable for climate change due to their business models. Hence, there will be an increase in liability risk (Carney, 2016). This could also apply to a specific industry. Industries inhibit industry-specific characteristics that are accountable for the
provide more resilience to climate change. Furthermore, firms need to adapt to the changing regulations, which involve transition risks. Changes in policy and regulation could initiate a reassessment of the value of their assets as the environmental impact is taken more into account (Carney, 2016).
Pielke (2007) emphasizes the importance of adaptation to climate change, so that risks could be mitigated or reduced. To prevent and mitigate the risks associated with ENEs, the U.S government encourages its communities and business to create disaster resilience to withstand a disaster of any kind (National Academies, 2012). The U.S. government even provides support to firms to establish a Business Continuity Plan (hereafter BCP; Homeland Security, 2019). The BCP is a tool to provide confidence that the output of a firm can be delivered in the face of any kind of risk (Gibb and Buchanan, 2006). An effective BCP mitigates the risks and reduces the economic loss of a disaster. However, completely mitigating this risk is very unlikely.
Since literature is inconclusive about the effect of climate change on firm performance, the following hypothesis is formulated:
H1: Climate change does not have an effect on firm performance.
2.3 Internationalization and firm performance
the environment sustainable government and stakeholder management, since more and more people are becoming aware of global warming and its consequences. Firms and governments need to satisfy the needs of their customer or citizen. Neglecting their responsibility to take action and moving their assets to another location can have negative effects on the firm’s reputation (Lai, Chiu, Yang, and Pai, 2010).
Literature is inconclusive about the effect of internationalization on firm performance. Therefore, the following hypotheses are formulated:
H2: Internationalization does not moderate the relation between climate change and firm performance.
3. Methodology
3.1. Variables and data collection 3.1.1. Independent variables
Climate change will be measured using the independent variable: Damage ENE, measured in property damages caused by ENE in million US$. This paper includes the following ENE types: Tropical storm, Drought, Severe storm, and Floods. These ENEs are expected to increase in frequency and
disaster. However, this paper assumes that U.S. firms have an effective BCP in place. Hence, Damage ENE will be lagged by one year to partially account for the recovery time and measure the firm
performance in the aftermath of a disaster. Previous research has used the natural disaster data of EM-DAT (Klomp and Valckx, 2014). The “Emergency Events Database” of the Centre for Research on the Epidemiology of Disasters (CRED) contains detailed data about every reported ENE in the world, which are compiled by various sources, such as UN agencies and insurance companies. Data about tropical storms, severe storms, flooding, and droughts that occurred in the period of 1996-2016 are collected. Furthermore, firm data are collected from Compustat Global database for the period of 1996-2016. Only firms with a headquarter in the U.S. are included. Finally, financial firms and utility firms, with standard industrial code 6000-6999 and 4900-4999, are excluded. For financial firms, it is common to have high leverage and utility firms are often highly regulated. In addition, all firm-year observations consisting missing values are dropped. Hence, the sample consists of 12,917 firm-years (2,057 firms).
This study will measure Internationalization by using the ratio of foreign sales to total sales
(Sullivan, 1996). Additionally, we use a similar method as Park, Suh, and Yeung (2013) used to compare domestic firms to multinational companies. This will replace the previous Internationalization variable with a dummy variable. Thus, we establish multiple dummies for the degree of internationalization of a firm. The dummy variable will be created by using the amount of foreign sales over consolidated sales of a firm. Firms will be marked by either Domestic (< 20%), MNC20 (> 20%), or MNC50 (> 50%) (Park, Suh, Yeung, 2013). This makes it possible to capture the possible effect of internationalization of the different groups and compare them to each other. The ability to shift business operations to another foreign facility is expected to mitigate the effects of an ENE. Therefore, we measure the degree of
3.1.2. Dependent variables
Firm performance will be measured using the following dependent variables: Return on Assets (ROA), Return on Equity (ROE) and Tobin’s Q. As suggested in the current literature, climate change has an increasing effect on property, people and on the operational efficiency (Klomp and Valckx, 2014). ROA and ROE are widely used in other studies to identify firm performance (e.g. Berger and Ofek, 1994; Walls and Dyer, 1996). ROA and ROE indicate the profitability during a specific period in time. Hence, it evaluates the direct effect of climate change on the annual profitability of a firm. Similarly, Tobin’s Q is widely used as a performance measure, since a firm’s market value represents the (potential) results of the firm’s performance. It is calculated as natural log of the market value of the firm over the total book value of the assets (Varaiya and Kerin, 1987; Campbell and Minguez-Vera, 2008). Since Tobin’s Q also reflects the potential value of the company, it evaluates the effect on the firm’s long-term performance. Climate change could damage physical assets in place. Supposedly, this would indicate that a firm with a higher Tobin’s Q is less prone to climate change, since they require relatively fewer assets to create value. The data of U.S. firm’s performance are derived from the Compustat Global database on Wharton Research Data Services (WRDS) and all the values are measured in million US$ at the end of the fiscal year.
3.1.3. Control variables
Global database for the period of 1996-2016 and all the values are measured in million US$ at the end of the fiscal year.
3.2. Methods
This paper examines the effect of climate change, through the damages caused by natural disasters, on firm performance in the aftermath of the event. Furthermore, the study examines whether a more internationally diversified firms are differently affected than less internationally diversified firms or not. Therefore, a panel regression analysis is conducted on 2,057 firms, that have or had their headquarter located in the U.S., during the period of 1996-2016. Consequently, the following model is constructed:
Firm Performance i, t = ß0 + ß1Damage ENE i, t-1 + ß2FirmSize i, t + ß3Age i, t + ß4Leveragei, t
+ ß5Intangible Assets i, t + ß6Internationalization i, t-1 (1)
+ ß7Damage ENE * Internationalization i, t-1 + ŋ i, + µ t + ε i, t ,
where Firm Performance is the performance of firm i at time t; Damage ENE is the natural log of the annual amount of damages caused by extreme natural events for firm i at time t-1; Firm Size is the natural log of the book value of assets of firm i at time t; Age is the natural log of active years of firm i at time t; Leverage is the ratio of debt to total assets of firm i at time t; Intangible Assets is the natural log of the amount of intangible assets of firm i at time t; Internationalization is the ratio of foreign sales and total sales of firm i at time t-1; Damage ENE*Internationalization is the interaction term to test the moderating effect of Internationalization on Damage ENE of firm i at time t; fixed time effect is included in ŋ for firm i; fixed industry effects is included in µ at time t; ε is the error term.
If the previous model shows significant evidence of the effect of climate change on firm
their practices. This could affect the future of a firm’s or industry’s performance or even its existence (Carney, 2016). In addition, Andonoa and Ruíz-Pava (2016) address the importance of industry-related factors in the variance of performance of firms. Hence, we assume that this could also lead to better or less resilience to ENEs. For example, some industries are more depending on physical assets than others, which in turn could be more vulnerable to ENEs. Hence, industry dummy variables will be used to mark the firms to their industry. After that, an interaction term is added to see whether being a specific type of industry mitigate the effect of climate change on the performance or not. The following model is
constructed:
Firm Performance i, t = ß0 + ß1Damage ENE i, t-1 + ß2FirmSize i, t + ß3Age i, t + ß4Leverage i, t + ß5 Intangible Assets i, t + ß6Internationalization i, t-1 (2)
+ß7Industry Dummy i, t + ß8Damage ENE * Industry Dummy i,t-1
+ ŋ i + ε i,t, ,
where Industry Dummy is the industry specific dummy for firm i at time t, which equals one if the firm belongs to the industry, otherwise zero; Damage ENE * Industry Dummy is the interaction term for firm i at time t-1; fixed time effects are included in ŋ for firm i.
4. Results
4.1. Descriptive statistics
consists of the Manufacturing companies, whereas other industries, e.g. Agriculture and Non-classifiable, are accounting for the smallest part of the sample.
Table 2 presents the descriptive statistics of the key variables. In order to deal with outliers, all variables are winsorized at the 1% and 99% level. In addition, all rows consisting of missing values are dropped. As a result, an unbalanced panel of 12,917 firm-year observations (2,057 firms) is used in the regression over the period of 1996-2016.
Table 1: Industry distribution.
SIC code Industry Name Firms
Included
Firm-year Observations
0100-0999 Agriculture, Forestry and Fishing 6 28
1000-1499 Mining 91 527
1500-1799 Construction 26 170
1800-1999 Not used 0 0
2000-3999 Manufacturing 1,207 8,119
4000-4999 Transportation, Communication, Electric, Gas and
Sanitary service
75 453
5000-5199 Wholesale Trade 84 590
5200-5999 Retail Trade 81 470
6000-6799 Finance, Insurance and Real Estate 0 0
7000-8999 Services 475 2,514
9100-9729 Public Administration 0 0
9900-9999 Non-classifiable 12 46
Total 2,057 12,917
Note. The table reports the industry distribution of the sample based on the SIC code. All firms with SIC
between 4900-4999 and 6000-6999 are excluded from the sample. The sample contains 12,917 observations of 2,057 firms over the period of 1996-2016.
maximum is around US$ 70 billion. The mean of Internationalization is 0.28, indicating that the average U.S. firm in the sample has a foreign sales ratio of 28% to total sales. The average level of leverage is around 14%. The average Size of a firm in the sample is 5.73, which indicates an average asset book value of US$ 308 million. Minimum Size is -4.27, which indicates an asset book value of merely US$14,000, and a maximum Size of 10.91, which indicates an asset book value of over US$ 54 billion. Thus, the variation in Size is very large. At last, the average Age of the companies in the sample is 13 years, with a minimum of two and a maximum of 20 years.
Table 2: Descriptive statistics.
(1) (2) (3) (4) (5) Mean Standard deviation Minimum Maximum N ROA 0.02 0.16 -4.07 0.81 12,917 ROE 0.03 0.49 -16.34 9.08 12,917 Tobin’s Q 1.32 1.12 -10.37 9.55 12,917 Damage ENE 9.36 0.95 6.24 11.15 12,917 Internationalization 0.28 0.21 0.00 1.00 12,917 Size 5.73 1.91 -4.27 10.91 12,917 Age 2.57 0.34 0.69 3.00 12,917 Leverage 0.14 0.14 0.00 0.85 12,917 Intangible Assets 4.63 2.49 -6.21 10.14 12,917
Note. The table reports the main descriptive statistics for the key variables. The sample contains observations of
2,057 firms over the period 1996-2016. The sample is winsorized at the one percent level in both tails of the distribution. ROA is the return on assets; ROE is the return on equity; Tobin’s Q is the natural log of the market value over total asset book value; Damage ENE is the natural log of the annual amount of damage caused by extreme natural events in million US$; Internationalization is the ratio of foreign sales to total sales; Size is the natural log of book value of assets; Age is the natural log of firm age; Leverage is the debt divided by total assets; Intangible Assets is the natural log of intangible assets; N is the amount of observations.
4.2. Correlation matrix
coefficient of -0.04 with ROA, -0.03 with ROE, and -0.01 with Tobin’s Q. According to these results, Internationalization has no correlation with Damage ENE. All the other correlations with Damage ENE show a negative sign, indicating a negative correlation. Internationalization and Size have a positive correlation coefficient with the ROA and ROE. However, it shows negative correlations with Tobin’s Q. Age has a positive correlation coefficient with all three dependent variables. We would expect firms to improve their business and their value, when they obtain more experience, more assets, and grow their markets. Finally, the positive coefficient of Size with Intangible assets is relatively high with 0.65. To verify there is no multicollinearity, a Variation Inflation Factor test is performed. The results do not indicate multicollinearity in the sample.
Table 3: Pearson Correlation Matrix.
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (1) ROA 1.00 (2) ROE 0.73 1.00 (3) Tobin’s Q 0.11 0.08 1.00 (4) Damage ENE -0.04 -0.03 -0.01 1.00 (5) Internationalization 0.05 0.05 -0.06 0.00 1.00 (6) Age 0.14 0.12 0.06 -0.14 0.03 1.00 (7) Intangible Assets 0.25 0.21 0.37 -0.01 0.11 0.15 1.00 (8) Size 0.31 0.26 -0.13 -0.00 0.13 0.17 0.65 1.00 (9) Leverage -0.02 -0.01 0.18 -0.01 0.04 0.06 0.28 0.13 1.00
Note. The table reports the Pearson correlations of the key variables. ROA is the return on assets; ROE is the
4.3. Panel OLS regression results
Table 4 presents the effect of Total Damages, proxy for climate change, on the different measures of firm performance. Every model includes the control variables, fixed industry effects, and fixed year effects. We use the robust covariance estimator to account for the heteroskedasticity in the sample. Model 1 to 3 presents the estimates on ROA, ROE and Tobin’s Q, respectively. Notable is that Damage ENE has an estimate of -0.010 (t = 1.364), -0.046 (t = -1.875), and -0.057 (t = 1.050), which all suggest a negative effect on their dependent variable. The Damage ENE estimates are rather small in magnitude. However, the context of the variable should not be ignored, since the dependent variables are measured in millions of US$. On top of that, Damage ENE heavily fluctuates per year, with a minimum of over US$ 500 million and a maximum over US$ 70 billion. Nonetheless, only the estimate of model 2 is statistically significant at the ten percent level, suggesting that an increase of Damage ENE by 100% would, ceteris paribus, decrease ROE by 0.046 the following year. Furthermore, the estimates of Internationalization are positive in model 1 and 2 with 0.008 (t = 1.342) and 0.043 (t = 2.245), respectively. Only the latter is significant at the five percent level. In model 3, Internationalization is negative and significant at the one percent significance level with an estimate of -0.337 (t = -9.186). This would suggest that
negative effect of Damage ENE on ROE. Thus, we reject the null hypothesis of the first hypothesis, suggesting that climate change has a negative effect on firm performance.
In model 4 to 6, the interaction term is introduced to the regression, namely Damage
ENE*Internationalization. Again, Damage ENE enters the regressions negatively with -0.015 (t = 1.931), -0.057 (t = 2.178), and -0.054 (t = 0.977), respectively. The Damage ENE estimate of model 4 is
significant at the ten percent level and the estimate of model 5 is significant at the five percent significance level. Damage ENE is not statistically significant in model 6. Model 1 to 6 show little
evidence of the effect of Damage ENE on ROA and moderately significant evidence of the effect on ROE. However, there is no evidence to suggest that Damage ENE has an effect on Tobin’s Q. We suggest that Damage ENE only has an effect on the annual profitability, while there is no effect on the firm’s market value in the long-term. Thus, these results add additional evidence of the effect of climate change on firm performance.
After introducing the interaction term, Internationalization become negative for ROA and ROE, namely -0.176 (t = 3.018), and -0.350 (t = 1.493), respectively. Internationalization has an estimate of -0.227 (t = -0.642) in model 6. In addition, the estimate in model 4 becomes highly significant, while the estimates of model 5 and 6 are not significant. Furthermore, the interaction term enters the regression positively in model 4 and 5 and negatively in model 6 with 0.020 (t = 3.150), 0.042 (t = 1.661), and -0.012 (t = -0.313), respectively. Suggesting that being more internationalized mitigates the negative effect of Damage ENE on ROA and ROE, while it would strengthen the negative effect on Tobin’s Q. However, there is no evidence to suggest the latter. These results suggest that an increase of one unit in Internationalization would lead to, ceteris paribus, a decrease of the negative effect of Damage ENE by 0.020 of ROA and 0.042 of ROE. Hence, a mitigating effect on the relation. Furthermore, the estimates of the control variables are similar in sign, significance and magnitude after introducing the interaction term. Thus, we reject the null hypothesis second hypothesis, suggesting that Internationalization has a
Table 4: Climate change on firm performance.
Variable (1) (2) (3) (4) (5) (6)
ROA ROE Tobin’s Q ROA ROE Tobin’s Q
Damage ENE -0.010 -0.046 -0.057 -0.015 -0.057 -0.054 (1.364) (-1.875) * (1.050) (-1.931) * (-2.178) ** (0.977) Internationalization 0.008 0.043 -0.337 -0.176 -0.350 -0.227 (1.342) (2.245) ** (-9.186) *** (-3.018) *** (-1.493) (-0.642) Damage ENE* Internationalization 0.020 (3.150) *** 0.042 (1.661) * -0.012 (-0.313) Intercept -0.138 -0.222 1.918 -0.088 -0.116 1.888 (-2.058) ** (-0.976) (3.679) *** (-1.237) (-0.485) (3.566) *** Size 0.021 0.058 -0.352 0.021 0.058 -0.352 (18.821) *** (11.972) *** (-45.342) *** (18.839) *** (11.993) *** (-45.323) *** Age 0.045 0.113 0.161 0.045 0.113 0.161 (9.107) *** (8.120) *** (6.289) *** (9.088) *** (8.109) *** (6.292) *** Leverage -0.063 -0.142 0.560 -0.063 -0.142 0.560 (-6.223) *** (-2.175) ** (8.338) *** (-6.230) *** (-2.178) ** (8.339) *** Intangible Assets 0.005 0.013 0.336 0.005 0.013 0.336 (6.976) *** (5.740) *** (58.853) *** (7.008) *** (5.774) *** (58.843) ***
Year FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
R2 0.115 0.083 0.384 0.116 0.083 0.384
Cov. Estimator Robust Robust Robust Robust Robust Robust
Observations 12,917 12,917 12,917 12,917 12,917 12,917
Note. The table provides the estimated coefficients of the different dependent variables. ROA is the return on
assets; ROE is the return on equity; Tobin’s Q is the natural log of the market value over total asset book value; Damage ENE is the natural log of amount of damage caused by extreme natural events in million US$ lagged by one year; Internationalization is the ratio of foreign sales to total sales lagged by one year; Damage
4.5. Regression with industry dummies
The previous model presents significant evidence of the effects of climate change on firm
performance. Therefore, additional regressions are run on ROE and Tobin’s Q including industry specific dummies to see the effect of Damage ENE on a specific industry. The results are presented in Table 5.
As expected, Damage ENE enters every model with a negative sign. However, only the estimates on ROE, except for the regression including the Manufacturing dummy, are significant at the ten percent level. This suggests that Damage ENE has a negative effect on the ROE of most industries. As expected, the industry dummy is different in magnitude and sign than the other dummies. Industries have industry-specific characteristics, which play a role in their business model (Andonova and Ruíz-Pava, 2016). However, only the estimate of Agriculture dummy, 2.390 (t = 1.688), on Tobin’s Q is significant at the ten percent level.
In addition, the interaction term estimates are different in sign for some models, e.g. Services and Mining. For the Mining industry, the estimate of the interaction term suggests that the effect of Damage ENE on ROE would be strengthened with an estimate of -0.009 (t = -0.564). On the other hand, the effects of Damage ENE on the performance of the Service industry would be mitigated with an estimate of 0.017 (t = 1.643). We assume that this is caused by the differences in firm-specific characteristics, which will cause differences in disaster resilience (Andonova and Ruíz, 2016). For example, the Service industry is earning revenue from performing an intangible service, whereas the Mining industry needs to extract valuable minerals from the Earth. Hence, climate change could have a different effect across different industries (Carney, 2016). However, none of the estimates on ROE are statistically significant.
In model 1 to 3 with Tobin’s Q as independent variable, the estimates of the interaction term are significant. The estimates enter the regression negatively, indicating the negative moderating effect. Notable here is that Damage ENE is not significant. This suggests that the industry type negatively
there is only little evidence to suggest that Damage ENE has a different impact on the performance across different industries.
Table 5: Climate change and industry performance. Independent Variable Variable (1) (2) (3) (4) Ac Mn Cs Ma ROE Intercept -0.199 -0.200 -0.201 -0.275 (-0.877) (-0.883) (-0.885) (1.194) Damage ENE -0.044 -0.045 -0.044 -0.035 (-1.810) * (-1.858) * (-1.797) * (1.447) Industry dummy -0.030 -0.053 0.359 0.110 (-0.079) (-0.382) (1.529) (1.359) Damage ENE* Industry dummy 0.010 (0.233) -0.009 (-0.564) -0.039 (-1.546) -0.014 (1.578)
Control variables Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Cov. Estimator Robust Robust Robust Robust
Tobin’s Q Intercept 2.086 2.051 2.073 2.100 (3.948) *** (3.919) *** (3.918) *** (3.829) *** Damage ENE -0.059 -0.059 -0.058 -0.060 (-1.067) (-1.082) (-1.047) (-1.054) Industry dummy 2.390 0.536 0.429 -0.045 (1.688) * (1.337) (0.966) (-0.284) Damage ENE* Industry dummy -0.319 (-1.987) ** -0.087 (-2.062) ** -0.079 (-1.686) * 0.002 (0.115)
Control variables Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Table 5: Continued. Independent variable Variable (5) (6) (7) (8) (9) Tp Ws Rt Sv Nc ROE Intercept -0.199 -0.187 -0.195 -0.198 -0.196 (-0.878) (-0.825) (-0.859) (-863) (-0.866) Damage ENE -0.044 -0.045 -0.044 -0.047 -0.044 (-1.806) * (-1.859) * (-1.831) * (-1.917) * (-1.810) * Industry dummy 0.017 -0.107 -0.052 -0.120 -0.397 (0.082) (-0.825) (-0.261) (-1.195) (-0.831) Damage ENE* Industry dummy -0.002 (-0.106) 0.016 (1.150) 0.010 (0.480) 0.017 (1.643) 0.031 (0.698)
Control variables Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Cov. Estimator Robust Robust Robust Robust Robust
Tobin’s Q Intercept 2.055 2.038 2.081 1.960 2.086 (3.883) *** (3.856) *** (3.939) *** (3.689) *** (3.939) *** Damage ENE -0.057 -0.055 -0.061 -0.063 -0.058 (-1.032) (-0.992) (-1.096) (1.135) (-1.044) Industry Dummy -0.134 -0.023 0.125 -0.047 -2.180 (-0.341) (-0.075) (0.305) (-0.231) (-0.879) Damage ENE* Industry dummy -0.002 (-0.040) -0.027 (-0.840) 0.015 (0.352) 0.029 (1.360) 0.118 (0.464)
Control variables Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Cov. Estimator Robust Robust Robust Robust Robust
Note. The table provides the estimated coefficients of the regressions including the different industry dummies.
Every model includes a different industry dummy. ROE is the return on equity; Tobin’s Q is the natural log of the market value over total asset book value; Damage ENE is the natural log of the amount of damage caused by extreme natural events in million US$ lagged by one year; Ac is the Agricultural industry; Mn is the Mining industry; Cs is the Construction industry; Ma is the Manufacturing industry; Tp is the Transportation industry; Ws is the Wholesale industry; Rt is the Retail industry; Sv is the service industry; Nc are the Non-classifiable firms; Damage ENE* Industry dummy is the interaction term. All the models include firm-specific control variables (Size, Age, Leverage and intangible assets) and fixed year effects (Year FE). Each estimate is reported with robust t- statistics clustered at the firm level in the parentheses. The notation ***, **, and * denotes
4.6. Robustness checks 4.6.1. Performance measures
To check whether the results are robust or not, three performance measures are used as independent variable, namely ROA, ROE, and Tobin’s Q. As presented in Table 4, Damage ENE has the same sign in every model. Only the estimate in model 2 is significant at the five percent significance level, whereas the estimate of model 1 and 3 are not significant at all. Hence, there is significant evidence to suggest that climate change has an effect on firm performance. In model 4 to 6, the interaction term is added. The sign of the interaction term estimate in model 6 is contrary to model 4 and 5. ROA and ROE measure
profitability, whereas Tobin’s Q measures the firm market value. Since Tobin’s Q is depending on the market value, it could be that being very internationalized decreases the market value of the firm due to lower customer reputation (Lai et al., 2010). In addition, the regressions are run without winsorizing the data. The results remain the same.
4.6.2. Alternative measure for ENE
Additionally, a robustness check is performed by using another proxy of ENE. We substitute
Damage ENE with Injury Rate. This is the ratio of total injuries to the total of people affected by the ENE (Brooks and Adger, 2013). According to the literature, people will increasingly be affected by climate change. Due to the intensifying ENE, the chance of getting injured by an ENE will most likely increase over the years. Hence, the ratio of total injured people to the total of people affected by an ENE is taken (Brooks and Adger, 2013). Similar to the Damage ENE variable, Injury rate is lagged by one year. The results are presented in Table 6.
paribus, result in a decrease of ROA by 0.114 and by 0.118 of ROE. The Injury Rate estimate in model 3 suggests that, an increase by 1 would, ceteris paribus, decrease Tobin’s Q by 41.3%. However, none of these estimates are significant. Although, these results are not significant, the sign and direction of the estimates are similar to the regressions including Damage ENE. Hence, some robustness is added.
Table 6: Regression with alternative measure of ENE.
Variable (1) (2) (3) (4) (5) (6)
ROA ROE Tobin’s Q ROA ROE Tobin’s Q
Intercept -0.227 -0.641 1.401 -0.228 -0.643 1.395 (-14.018) *** (11.272) *** (18.203) *** (-13.978) *** (-11.230) *** (18.076) *** Injury Rate -0.114 -0.118 -0.413 -0.086 -0.070 -0.233 (-0.829) (-0.196) (-0.452) (-0.599) (-0.113) (-0.252) Internationalization 0.008 0.043 -0.337 0.012 0.050 -0.313 (1.318) (2.228) ** (-9.214) *** (1.573) (2.073) ** (-7.197) *** Injury Rate * Internationalization -0.119 (-0.960) -0.311 (-0.519) -0.777 (-0.969)
Control variables Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
Cov. Estimator Robust Robust Robust Robust Robust Robust
Note. The table presents the estimated coefficients of the regressions including the alternative proxy for climate
change. Injury rate is the ratio amount of injured people to total affected people caused by an extreme natural event; Internationalization is the ratio foreign sales to total sales lagged by one year; Injury rate *
Internationalization is the interaction term. All the models include firm-specific control variables (Size, Age, Leverage and Intangible assets), fixed year effects (Year FE) and fixed industry effects (Industry FE). Each estimate is reported with robust t- statistics clustered at the firm level in the parentheses. The notation ***, **, and * denotes statistical significance as the 1 %, 5 % and 10 % levels, respectively.
-0.119 (t = -0.960), -0.311 (t = -0.519), and -0.777 (t = -0.969). An increase of Internationalization by 1 would, ceteris paribus, result in strengthening (mitigating) the negative (positive) effect of Injury Rate on firm performance by -0.119 in model 4 and -0.311 in model 5. Thus, if we assume that Injury Rate has a negative estimate, these results suggest that being more internationalized would, ceteris paribus,
strengthen the negative effect of climate effect on firm performance. This is opposite to the results of the main regression in Table 4, where the interaction term estimate of model 4 and 5 have a positive sign. However, the negative estimates are not statistically significant. Hence, we still suggest that becoming more internationalized has a mitigating effect on the negative effect of climate change on firm
performance.
4.6.3. Internationalization dummies
Table 7: Climate change and internationalization.
Variable (1) (2) (3) (4) (5) (6)
ROA ROA ROE ROE Tobin’s Q Tobin’s Q
Intercept -0.113 -0.123 -0.151 -0.189 1.825 1.921 (-1.603) (-1.819) * (-0.639) (-0.829) (3.470) *** (3.684) *** Damage ENE -0.013 -0.011 -0.053 -0.049 -0.052 -0.057 (-1.638) (-1.542) (-2.067) ** (-1.999) ** (-0.948) (-1.046) MNE20 -0.037 -0.097 0.009 (-1.483) (-1.050) (0.060)
Damage ENE * MNE20 0.005 0.014 -0.009
(1.907) * (1.412) (-0.587)
MNE50 -0.095 -0.176 -0.481
(-3.480) *** (-1.598) (2.328) **
Damage ENE * MNE50 0.010 0.021 0.024
(3.474) *** (1.759) * (1.099)
Control variables Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
Cov. Estimator Robust Robust Robust Robust Robust Robust
Note. The table presents the estimated coefficients of the regressions on the different dependent variables
including different dummies representing the degree of internationalization. ROA is the return on assets; ROE is the return on equity; Tobin’s Q is the natural log of the market value over total book value; Damage ENE is the natural log of the annual amount of extreme natural event damages lagged by one year; MNE20 represents firms with a foreign sales ratio of at least 20 %; MNE50 represents firm with a ratio of at least 50 %. All the models include firm-specific control variables (Size, Age, Leverage and intangible assets), fixed year effects (Year FE) and fixed industry effects (Industry FE). Damage ENE*MNE dummy is the interaction term. Each estimate is reported with robust t- statistics clustered at the firm level in the parentheses. The notation ***, **, and * denotes statistical significance as the 1 %, 5 % and 10 % levels, respectively.
As expected, Damage ENE enters with a negative sign in every model. However, only the estimates of model 3 and 4 are significant at the five percent significance level. Model 1 and 2 presents the
ROA is reduced by -0.037. However, the estimate is not significant. The estimate of the interaction term is significant at the ten percent level, suggesting the positive moderating effect of MNE20 on the relation between Damage ENE and ROA. In model 2, MNE50 enters the regression with -0.095 (t = 3.480) and is highly significant. The estimate of the interaction term is 0.010 (t = 3.474) and is also highly significant. Hence, this suggests that if MNE50 is 1, this would lower ROA with -0.095, while mitigating the negative effect of Damage ENE on ROA with 0.010. These results are in line with the results of model 4 in Table 4. Furthermore, model 3 and 4 present the regressions on ROE. MNE20 enters the regression with a negative -0.097 (t = 1.050) and the interaction term with 0.014 (t = 1.412). This would suggest the same as model 1. However, none of these estimates are significant. MNE50 has an estimate of 0.176 (t = -1.598) and the estimate of the interaction term is 0.021 (t = 1.759). However, only the interaction term estimate is significant at the ten percent significance level. Hence, this suggests a mitigating effect of MNE50, 0.021(t = 1.759), on the negative effect of Damage ENE, -0.049 (t = -1.999), on ROA. In model 5 and 6, the regressions are run on Tobin’s Q. Notable is that MNE20 has a positive 0.009 (t = 0.060). This would suggest that Tobin’s Q increases with 0.90%, when MNE20 is one. Simultaneously, it has a negative estimate of -0.009 (t = -0.587) for the interaction term. It would suggest that if MNE20 is one, Tobin’s Q will be, ceteris paribus, 0.009 lower than for other firms. This looks similar to the results of model 6 in Table 4. However, none of these estimates are significant. Finally, MNE50 and Damage ENE*MNE50 are added. MNE50 enters the regression with an estimate of -0.481 (t = 2.328) and the interaction term with an estimate of 0.024 (t = 1.099). Like the estimate of Internationalization in model 6 in Table 4, MNE50 is negative. There is evidence at the five percent significance level, to suggest that having at least 50% foreign sales ratio, affects the Tobin’s Q negatively by 48.1%. The interaction term estimate is positive, which is in line with our previous findings. However, this estimate is not significant. Overall, the dummies show some differences in the effect of Damage ENE on the profitability measures between domestic and multinational firms. Thus, according to these results the degree of
little evidence to suggest that Damage ENE will affect multinational firm, with at least 20% foreign sales ratio, differently than Domestic firms. However, there is strong evidence to suggest that being a
multinational firm, with at least 50% foreign sales ratio, has a mitigating effect on the negative effect of climate change on firm performance. Thus, this adds robustness to the main findings: Internationalization has a mitigating effect on the main effect.
6. Conclusion and discussion
During the last decades, climate change became a popular topic of discussion. The frequency and intensity of ENEs increased dramatically and caused major damages. A few studies have tried to estimate the effects of ENEs on the firm performance. However, the evidence of these studies is rather
inconclusive. Several studies discuss one single ENE and the effects in the aftermath of that particular disaster. Besides that, none of these studies include the internationalization of firms in their studies. To the best of our knowledge, this paper is the first to study the effect of climate change, through extreme natural events, on firm performance and the role of internationalization in this relation.
Performing a panel data analysis using 12,917 firm-years observations of U.S. firms over the period of 1996-2016, the results can be summarized as follows. First, climate change has a negative effect on firm performance. In particular, the annual profitability of a firm. However, no evidence is found of the negative effect on the market value of firms. Second, there is some evidence to suggest the effect of climate change is different across different industries. Agricultural companies seem to be the biggest losers, whereas the Service industry is relatively the smallest losers of climate change. Third,
internationalization has a role of mitigating the negative effects of climate change on firm performance. These findings contribute to the existing literature on the impact of climate change on firm
However, studies (Klomp and Valckx, 2014; United Nations, 2019) expect the extreme natural events to increase in frequency and intensity. If firms have less time to recover and financial resources are
exhausted, it is likely that climate change will also have an effect on the market value of a firm. Second, it shows that several industries in particular are more prone to the consequences of climate change.
Industries that are more depending on physical assets, such as Agriculture and Mining, will be hit harder by a natural disaster than for example the Service industry. These findings provide additional evidence of industry- specific climate risks. Third, it shows that U.S. companies that do more business internationally are less affected by climate change. This could suggest that multinational firms are better able to shift their business activities whenever a disaster could occur than domestic firms.
From a managerial perspective, these findings provide evidence on additional risks of climate change on firm performance. In addition, the risk differs across different industries. As mentioned before, the additional risks lower investor expectations and increase risk premium. Hence, this could affect the financing of capital. Furthermore, this study suggests that being more internationally diversified could mitigate some of the effects of climate change. Thus, the inclusion of climate risks into the strategic decision-making should not be neglected.
organizational behavior is different across business environments every year, it is not included in this study. However, future research should explore the organizational behavior of firms, their reaction and their strategy on extreme natural evens. Third, the role of science and innovations is not included in this study. Climate change has many direct and indirect effects. However, some of these effects are not scientifically proven yet, and therefore, cannot be included. Hence, to capture all the relevant effects on firm performance, additional research is required exploring the effects of climate change. In addition, it seems like we always have found a solution to mitigate or solve the risks they have face at that moment. In addition, every year, new innovations are introduced, e.g. regulatory changes such as the Paris
References
Andonova, V. Ruíz-Pava, G., 2016. The role of industry factors and intangible assets in company performance in Colombia. Journal of Business Research 69, pp. 4377-4384.
Battiston, S., Mandel, A., Monasterolo, I., Schütze, F., Visentin, G., 2017. A climate stress-test of the financial system. Nature Climate Change 7, pp. 283-288.
Berger, P.G., Ofekm, E., 1995. Diversification’s effect on firm value. Journal of Financial Economics 37, pp. 39-65.
Brooks, N., Adger, N.W., 2013. Country level risk measures of climate-related natural disasters and implications for adaptation to climate change. Unpublished working paper. Tyndall Centre, Norwich. Carney, M., 2016. Resolving the climate paradox [Speech]. Retrieved from:
https://www.bankofengland.co.uk/speech/2016/resolving-the-climate-paradox.
Centre for Research on the Epidemiology of Disaster – CRED. (EM-DAT), 2019. Natural hazard U.S. [Data File].
Climate Policy Initiative, 2018. Global climate finance: an updated view 2018 [Report].
Dai, L., Eden, L., Beamish, P.W., 2017. Caught in the crossfire: dimension of vulnerability and foreign multinationals’ exit from war-afflicted countries. Strategic Management Journal 38, pp.1478- 1498. Desai, M.A., Foley, C.F., Forbes, K., 2008. Financial constraints and growth: multinational and local firm responses to currency depreciations. Review of Financial Studies 21(6), pp. 2857-2888.
Environmental Defense Fund (EDF), n.d.. Extreme weather gets a boost from climate change [Webpage]. Retrieved from: https://www.edf.org/climate/climate-change-and-extreme-weather.
Fauver, L., Hung, M., Li, X., Taboada, A.G., 2016. Board reforms and firm value: worldwide evidence. Journal of Financial Economics 125(1), pp. 120-142.
Gibb, F., Buchanan, S., 2006. A framework for business continuity management. International Journal of Information Management 26, pp. 128-141.
Harvard Medical School, 2005. Climate change futures: health, ecological and economic dimensions [Report].
The Department of Homeland Security, 2019. Business continuity plan [Webpage]. Retrieved from: https://www.ready.gov/business/implementation/continuity
Klomp, J., Valckx, K., 2014. Natural disasters and economic growth: a meta-analysis. Global Environmental Change 26, pp. 183-195.
Krapl, A., 2015. Corporate international diversification and risk. International Review of Financial Analysis 37, pp. 1-13.
Lai, C., Chiu, C. Yang, C., Pai, D., 2010. The effect of corporate social responsibility on brand
performance: the mediating effect of industrial brand equity and corporate reputation. Journal of Business Ethics 95 (3), pp. 457-469.
Millman, O., 2019, January 31. What is the polar vortex? – and how is it linked to climate change?. The Guardian. Retrieved from: https://www.theguardian.com/us-news/2019/jan/30/polar-vortex-2019-usa-what-is-it-temperatures-cold-weather-climate-change-explained
Nakamura, E., Steinsson, J., Barro, R., Ursúa, J., 2013. Crises and recoveries in an empirical model of consumption disasters. American Economic Journal: Macroeconomics 5(3), pp. 35-74.
National Academies., 2012. Disaster resilience: a national imperative. The National Academies Press, Washington, DC.
New Climate Economy, n.d.. The sustainable infrastructure imperative [Webpage]. Retrieved from: https://newclimateeconomy.report/2016/
Noth, F., Rehbein, O., 2019. Badly hurt? Natural disaster and direct firm effects. Finance Research Letter 28, pp. 254-258.
Preston, B.L., 2013. Local path dependence of U.S. socioeconomic exposure to climate extremes and the vulnerability commitment. Global Environmental Change 23, pp. 719-732.
Ritchie, H., Roser, M., 2019. Natural disasters. our world in data [Data file]. Retrieved from: https://ourworldindata.org/natural-disasters.
Strulik, H., Trimborn, T., 2019. Natural disasters and macroeconomic performance. Environmental and Resources Economics 72, pp. 1069-1098.
Sullivan, D., 1996. Measuring the degree of internationalization of a firm: a reply. Journal of International Business Studies 27 (1), pp. 179-192.
The Economist Intelligence Unit (EIU), 2015. The cost of inaction: recognizing the value at risk from climate change [Report]. The Economist.
United Nations, n.d.. The Paris agreement [Webpage]. Retrieved from: https://unfccc.int/process-and-meetings/the-paris-agreement/the-paris-agreement
United Nations, n.d.. Introduction to climate finance [Webpage]. Retrieved from: https://unfccc.int/topics/climate-finance/the-big-picture/introduction-to-climate-finance
Varaiya, N., Kerin, R.A., 1987. The relationship between growth, profitability, and firm value. Strategic Management Journal 8, pp.487-497.
Walls, M.R., Dyer, J.S., 1996. Risk propensity and firm performance: A study of the petroleum exploration industry. Management Science 42 (7), pp. 1004-1021.
Wharton Research Data Services, 2019. Fundamentals Annual [Data File]. Retrieved from: https://wrds-web.wharton.upenn.edu/wrds/ds/compd/funda/index.cfm?navId=83
Wilts, A., 2018, January 8. Natural disaster damage cost America $306 billion in 2017. The Independent. Retrieved from: https://www.independent.co.uk/news/world/americas/natural-disasters-us-damage-cost-money-2017-a8148771.html
