1. Empirical Model

We employ a recursive vector autoregression (Recursive VAR) model in order to estimate the impact of external uncertainty on Korean domestic real macroeconomic variables. The baseline model is defined as follows:

There are seven endogenous variables in the order of Russian geopolitical risk, US monetary policy uncertainty, the global trade volume, equipment investment, industrial production, prices, and interest rates. Specifically, A is a constant vector, and C is a coefficient matrix representing the contemporaneous relationships between the endogenous variables. Each A(i) is a coefficient matrix capturing the impact of past endogenous variables on current endogenous variables, with p representing the maximum lag. Finally, ε_t is the structural shock vector.

To identify structural shocks, we assume that matrix C is a lower triangular matrix with a diagonal of 1. Specifically, we assume that external shocks initially impact domestic variables. Hence, the external components, consisting of Russian geopolitical risk, US monetary policy uncertainty, and global trade, are placed at the top of the system. In addition, domestic financial variable, i.e., interest rates, is positioned after domestic real variables. This assumption is based on the idea that domestic macro variables immediately impact financial variables, but the impact of domestic financial variables on macro variables occurs with a lag.

Furthermore, given that Korea is a small open economy, we assume that external variables can affect domestic variables, but domestic variables do not affect external variables. This assumption applies not only to contemporaneous relationships but also implies that the movements of external variables in the entire system are independently determined, regardless of Korean domestic variables. Specifically, the matrices representing the impact of past endogenous variables on current endogenous variables can be constructed as shown below.

Here, A₁₁(i), A₂₁(i), and A₂₂(i) are the block matrices of each A(i). A₁₁(i) represents a coefficient matrix showing the dynamic relationship between external variables. A₂₁(i) is a coefficient matrix that captures the dynamic impact of external variables on domestic variables, and A₂₂(i) is a coefficient matrix representing the dynamic relationships among domestic variables. However, even if Korea is considered a small open economy, there may be concerns that its influence may not be insignificant in the global market. Therefore, we estimated the model without imposing the above assumptions and confirmed that the results are not sensitive to this assumption, as detailed in Figure A2 in the Appendix.

2. Data

The measures of geopolitical risk in Russia and uncertainty in US monetary policy are derived from the previous section. Therefore, the increase in each uncertainty in our analysis can be interpreted as the pure effect of uncertainty while controlling the economic situation of the US and Russia.

Investment is the equipment investment index of Statistics Korea, industrial production denotes all-industry production excluding agriculture and fisheries from Statistics Korea, and the domestic interest rate is the call rate of the Bank of Korea. Prices are sourced from the consumer price index of the Bank of Korea, and the world trade volume is the real trade volume obtained from the CPB NL World Trade Monitor as of June of 2022.2

In order to ensure the stationarity of the time-series data, all variables except for the uncertainty measures and interest rates are expressed as log differences from the previous year. The time period ranges from January of 2000 to December of 2021. Given that we use log differences compared to the previous year, the effective sample period starts in January of 2001. Dummy variables for the global financial crisis are controlled as exogenous variables. Lastly, due to the limited number of observations in the analysis, we constrain the number of endogenous variables by setting the lag to p=3. We confirm that the main results remain consistent even when the lag is set to p=6 (refer to the Appendix).

3. Baseline Analysis: Macro effects

Figure 2 shows the impulse responses of investment and industrial production to a one standard deviation increase of both the Russian geopolitical risk and US monetary policy uncertainty. The solid lines in each graph represent the responses, and the shaded areas indicate the 68% confidence intervals.

FIGURE 2.

IMPACTS OF UNCERTAINTY SHOCKS ON INVESTMENTS AND INDUSTRIAL PRODUCTION

Note: Shaded areas show the one standard deviation confidence interval of the impulse response.

The top two figures illustrate how investments respond to shocks from Russian geopolitical risk and US monetary policy uncertainty. Increased uncertainty leads to a decrease in corporate investment, corroborating existing literature. In particular, a one standard deviation rise in Russian geopolitical risk and US monetary policy uncertainty results in a maximum decrease in equipment investment by 1%p and 0.6%p, respectively, with a lag of 5-6 months. The effects of these uncertainties persist for up to a year.

The bottom two figures show the impulse responses of industrial production. A one standard deviation increases in Russia’s geopolitical risk and U.S. monetary policy uncertainty results in a maximum reduction in industrial production of 0.24%p and 0.27%p, respectively, with a lag. Similar to the investment responses, the impact of these shocks lasts up to one year. Therefore, despite some differences in magnitude, the impulse responses of investment and industrial production show similar patterns.

A one standard deviation increase in the structural shock of Russian geopolitical risk and US monetary policy uncertainty leads to a 36% and 33% rise in their respective uncertainties. In March of 2022, these shocks escalated by over 100% and 25%, respectively, compared to the average level. Given that the standard deviations of the log differences in investment and industrial production are 10%p and 3%p, respectively, the recent surge in external uncertainty could significantly impact the domestic real economy in Korea.

As shown by the results above, uncertainty can indirectly impact on the supply side by reducing corporate investment, deteriorating both resource allocation efficiency and aggregate productivity (Bloom, 2009). Additionally, uncertainty can lead to an economic downturn by reducing overall demand (Basu and Bundick, 2017; Leduc and Liu, 2016). Because uncertainty shocks adversely affect both total demand and total supply, policy implications may depend on which aspect is more significantly impacted. For example, if total demand decreases more, economic production will decrease and price levels will fall, requiring an expansionary monetary policy. Conversely, if total supply decreases more significantly, economic production will decline but price levels will rise, potentially necessitating a contractionary monetary policy. Therefore, we investigate the responses of additional variables, specifically consumer prices and policy interest rates from the basic model in order to examine which part of total demand and supply has been more significantly affected by the two uncertainty shocks.

Figure 3 shows the responses of consumer prices and domestic policy interest rates to the uncertainties of Russian geopolitical risks and US monetary policy. The solid line in each graph represents the impulse response function, and the shaded area indicates the 68% confidence interval.

FIGURE 3.

IMPACTS OF UNCERTAINTY SHOCKS ON INVESTMENTS AND INDUSTRIAL PRODUCTION

Note: Shaded areas show the one standard deviation confidence interval of the impulse response.

The top two figures show the responses of consumer prices. When the geopolitical risk in Russia and the uncertainty in US monetary policy increase by one standard deviation, the consumer price inflation rate decreases by 0.1%p and 0.05%p respectively. Additionally, each shock has a statistically significant negative impact for up to 30 months after it occurs.

The findings in the bottom two figures are consistent with the results above. Specifically, an increase in geopolitical risk in Russia and uncertainty in US monetary policy leads to a decrease in policy rates by 0.08%p and 0.1%p respectively. Similar to the responses of consumer prices, these rate cuts persist for up to 30 months after each shock occurs.

Our findings suggest that both types of uncertainty lead to a larger decrease in aggregate demand, thereby reducing consumer price inflation. Consequently, monetary policy responds by setting policy rates significantly lower for a considerable period. After controlling for the main driver of high inflation in recent periods, namely oil price fluctuations, the results remain consistent (see Figure A3 in the Appendix for details).

In fact, no significant events directly affecting the supply side, such as disruptions in Russia’s energy supply or major countries imposing trade sanctions, occurred until the end of the analysis period, which was set to 2021. Extending the analysis period to February of 2024 reveals that the statistical significance of consumer price inflation and interest rate responses disappears (see Figure A5 in the Appendix). Therefore, the observed results can be attributed to the increase uncertainty in a context where drastic decreases in total supply due to extreme events, such as war, are not considered.

We also confirm that our results remain robust when (i) extending our sample to February of 2024, (ii) controlling for a COVID-19 dummy variable, (iii) using an alternative measure of MPU from Husted et al. (2020) and (iv) applying an alternative ordering of external uncertainties, with MPU placed first, as shown in Figures A5-A7 in the Appendix.

B. Empirical Analysis: Sectoral Analysis

The previous analysis confirms that an increase in external uncertainty has a significant impact on the macroeconomy of South Korea. To understand the transmission channels of the uncertainties, we also analyze their effects on exports and domestic demand, as well as production outputs in the manufacturing and service sectors by extending the baseline regression model. Specifically, we divide industrial production into two categories: exports and domestic demand, and production outputs in the manufacturing and service sectors. Each version of the modified models includes a total of eight endogenous variables, as follows:

The export volume index of the Bank of Korea is used for our export measure, the domestic supply index of Statistics Korea is used for domestic demand, and the subindices of the total industrial production index of Statistics Korea are used for manufacturing and service sector production outputs.3

Figure 4 illustrates the impact of external uncertainties on exports and domestic demand. Following shocks from Russian geopolitical risk and US monetary policy uncertainty, exports decrease by up to 1%p and 0.9%p, respectively, with a time lag. Additionally, the impact of these uncertainties on exports remains significant for about a year. On the other hand, the domestic demand response is smaller, with corresponding decreases of 0.6%p and 0.5%p. The responses of domestic demand also persist for about a year. Given that the export sector is more exposed to the global economy, the interpretation of these differential effects is straightforward. Despite the responses of domestic demand being smaller, they are still statistically significant, indicating that an increase in external uncertainties affects the overall economy.

FIGURE 4.

IMPACTS OF UNCERTAINTY SHOCKS ON EXPORTS AND DOMESTIC DEMAND

Note: Shaded areas show the one standard deviation confidence interval of the impulse response.

Figure 5 shows the responses of manufacturing and service sector production. Following shocks from Russian geopolitical risk and US monetary policy uncertainty, manufacturing production decreases by 0.9%p and 0.5%p, respectively, with a time lag. Additionally, these impacts remain statistically significant for about a year. On the other hand, service sector production shows smaller responses, with decreases of 0.3%p in each case, and the negative impact on service sector production also persists for about a year.

FIGURE 5.

IMPACTS OF UNCERTAINTY SHOCKS ON MANUFACTURING AND SERVICE PRODUCTION

Note: Shaded areas show the one standard deviation confidence interval of the impulse response.

Similar to the results from the regression with exports and domestic demand, there is a significant decrease in manufacturing production. Although the decline in service industry production is smaller, it remains statistically significant over a considerable period. Therefore, we can conclude that an increase in external uncertainties affects not only industries heavily exposed to the global economy but also has a broad effect on the overall economy in Korea.

1. Impact of External Uncertainties on Global Trade

Before discussing the transmission channels through which Russian geopolitical risks and US monetary policy uncertainty affect the domestic economy by decreasing external demand, we confirm that external uncertainties have negative impacts on other countries’ economic activities, thereby reducing global demand. There are several possible channels through which external uncertainty impacts the global economy. First, it can lead to delays in business investments, which may slow economic activity worldwide. Second, rising uncertainty can slow global trade down through a stronger dollar, which is a key driver of international trade (Gopinath et al., 2020).4 The global trade volume is used as a proxy for global demand. Hence, without estimating a new regression model, the impact of external uncertainties on the global trade volume is readily available from our baseline regression results.

Figure 6 illustrates a significant decrease in world trade volume in response to shocks that affect Russian geopolitical risks and US monetary policy uncertainty. Specifically, we find that a one standard increase in Russian geopolitical risks and US monetary policy uncertainty leads to a maximum decrease in world trade volume of 0.6%p and 0.5%p, respectively, with a time lag. Furthermore, the response of the global trade volume persists for up to 30 months.

FIGURE 6.

IMPACTS OF UNCERTAINTY SHOCKS ON GLOBAL TRADE

Note: Shaded areas show the one standard deviation confidence interval of the impulse response.

2. Impact of Global Trade Shocks on the Korean Economy

Now, we explore how shocks to global trade affect the Korean economy. The decrease in external demand, reflected in global trade, will directly impact the Korean economy through exports, ultimately affecting industries as a whole. Specifically, when external demand contracts, Korean exporting firms will experience decreases in the level of demand for their products. This will lead to a contraction in the production activities of Korean firms that supply intermediate goods and services for exporters, negatively impacting the overall domestic economy. Additionally, a decrease in exports may result in reduced household spending or decreased investments by firms due to lower income and revenue levels. Hence, in the following analysis, we investigate how the Korean economy responds to global trade shocks.

In Figure 7, we observe that in response to a one standard deviation shock in the world trade volume, exports and manufacturing production, which are highly exposed externally, show corresponding maximum declines of 3.4%p and 1.7%p, with a delay. Investment and domestic consumption also decrease by up to 1.3%p, while service production and overall industrial production drop by up to 0.5%p and 0.7%p, respectively. These findings suggest that the impact of global trade shocks has effects beyond sectors directly exposed to the global economy, also propagating throughout the broader economy via various transmission channels. The responses of all real variables, except exports, are shown to last for about five to six months, indicating that the impacts dissipate more quickly than those from uncertainty shocks.

FIGURE 7.

IMPACTS OF GLOBAL TRADE SHOCKS ON THE KOREAN ECONOMY

Note: Shaded areas show the one standard deviation confidence interval of the impulse response.

The findings above provide evidence of the transmission path of uncertainties through decreased external demand. Subsequently, we analyze the extent to which external demand contributes to the transmission channel of uncertainty through a counterfactual analysis.

3. Transmission Channel of External Uncertainty Shocks – Counterfactual Analysis

In this section, we conduct a counterfactual analysis to understand the role of global demand during the transmission of external uncertainty shocks. Specifically, we consider a scenario in which uncertainties impact the Korean domestic economy in the absence of a global trade channel. We then compare the counterfactual impulse responses to those from the baseline model. The impulse response functions in the baseline model implicitly reflect all possible channels through which external uncertainties affect the Korean economy. Hence, comparing these different impulse responses allows us to evaluate the importance of the transmission channel through decreased external demand. This methodology was widely used in earlier studies (Akinci, Kalmeli-Özcam, and Queralto, 2022; Caldara and Iacoviello, 2022) to identify specific transmission channels of macro shocks.

The method of deriving the counterfactual impulse responses follows Caldara and Iacoviello (2022). The specific steps are as follows. First, we use the estimated baseline model. Next, we increase the structural shocks related to geopolitical risks in Russia or uncertainty in US monetary policy by one standard deviation in each case. To shut down the transmission channel through global trade for each type of uncertainty, we generate counterfactual structural shocks in global trade that prevent global trade from responding. Consequently, in each case of a structural shock related to geopolitical risks in Russia or uncertainty in US monetary policy, the global trade volume remains constant. By muting the response of global trade to each uncertainty shock, we can estimate the impact of external uncertainty without an external demand channel.

Figure 8 shows the analysis of Russia’s geopolitical risk. The black solid line represents the response of domestic real variables derived in the baseline model, while the red dashed line shows a counterfactual response with the world trade volume constant. As shown in all cases, the estimated counterfactual responses are smaller than the responses of the baseline model. In particular, for exports, manufacturing, and domestic consumption, the counterfactual response is not included in the confidence interval of the baseline responses, suggesting that the world demand channel would play an important role in transmitting Russia’s geopolitical risk. Specifically, for a one standard deviation increase in Russia’s geopolitical risk, the counterfactual responses of exports and manufacturing output, as well as domestic consumption, decreased by up to 0.5%p, 0.6%p, and 0.4%p, respectively, compared to the responses of the baseline model (-1.0%p, -0.9%p, -0.6%p). The difference between the two impulse responses decreases significantly and does not persist for more than a year, disappearing in the short term.

FIGURE 8.

BASELINE AND COUNTERFACTUAL RESPONSES TO RUSSIAN GEOPOLITICAL RISKS

Note: 1) Shaded areas show the one standard deviation confidence interval of the impulse response; 2) Red dashed lines depict the counterfactual impulse responses.

Figure 9 shows the results for US monetary policy uncertainty shocks. Similar to Russia's geopolitical risk, the counterfactual responses of all variables are estimated be smaller than those of the baseline model. The counterfactual responses of exports and manufacturing are not included in the confidence interval of the baseline responses. Specifically, if US monetary policy uncertainty increases by one standard deviation, the counterfactual responses of exports and manufacturing output decrease by up to 0.5%p and 0.2%p, respectively, compared to the decrease in each baseline model (-0.9%p, -0.5%p). Similar to Russia’s geopolitical risk, this difference is maximized in the short term, with no sign of a persistent discrepancy.

FIGURE 9.

BASELINE AND COUNTERFACTUAL RESPONSES TO US MONETARY POLICY UNCERTAINTY

Note: 1) Shaded areas show the one standard deviation confidence interval of the impulse response; 2) Red dashed lines depict the counterfactual impulse responses.

In sum, our analysis provides supportive evidence that the response of the world trade volume could play an important role in transmitting both Russian geopolitical risks and US monetary policy shocks, especially in sectors with higher external dependence, such as exports and manufacturing. Their impact is maximized for up to five to six months, and over time, the importance of this transmission channel gradually diminishes.

1. Baseline regression model

In order to investigate the impact of external uncertainty on exports, we estimate the following panel regression model. This model captures the dynamic responses to specific shocks in a manner similar to the local projection by Jordà (2005); it is also widely used in existing studies using micro data (Caldara and Iacoviello, 2022; Ottonello and Winberry, 2020). First, we estimate the baseline model for the average responses, after which we extend the model to analyze the heterogeneous responses of exports by country and product.

Here, y_{p,c,t + h} represents the value of the dependent variable after h ≥ 0 periods, and the dependent variable is the change in the log export volume from the previous year. Index p indicates individual products, c indicates the destination country, and t represents time. The variables GPU_t and MPU_t represent Russian geopolitical risk and the uncertainty of US monetary policy, respectively. X_c,t and Z_t refer to macro variables that may affect exports. Specifically, X_c,t refers to the industrial production of country c, representing the country- specific demand factor. Z_t includes Korea’s industrial production and the USD/KRW exchange rate. All additional control variables are the log differences from the previous year. Products and destination fixed effects are also controlled for, and two-way clustered standard errors, clustered by product × trading country and time, are used. As we find in the previous sections, the responses of macro variables to uncertainty shocks last up one year. Therefore, we examine the dynamic response up to one year after the uncertainty shocks.

2. Data

Data were sourced from the Korea Customs Service.5 Each product is classified according to the 10-digit level of the Harmonized System Code (HSK), and data for all export destinations of Korea are provided on a monthly basis. Because the export value and export weight (in kg) are available, this study uses the export weight as a proxy for quantity. Although the HSK 10-digit level provides a high level of disaggregation of products, changes in the product composition or quality within the same level may result in weight variations. Therefore, following previous studies (Amiti, Itskhoki, and Konings, 2022), products that have experienced an increase of three times or more or a decrease of a third in the annual unit value from the previous year or less are excluded from the analysis. Additionally, outliers belonging to the top 0.5% or bottom 0.5% of growth rates in export quantity from the same month of the previous year are excluded as well.

Industrial production in trading partner countries and in Korea, as well as the USD/KRW exchange rate (monthly average), are sourced from the International Financial Statistics from the International Monetary Fund (IMF).6 The data cover the period from January of 2000 to December of 2021. Because we use growth rates of each control variable, the actual period used in the analysis begins in January of 2001. The analysis also controls for the period of the global financial crisis by including a dummy variable for the corresponding period.

3. Baseline Results: Average Impact

Figure 10 shows the results of the average responses. The graph on the left shows the dynamic responses of the export volume to a one standard deviation shock in geopolitical risk. On impact, the export volume does not respond to Russian geopolitical risk. However, with a lag, the export volume gradually decreases, reaching a maximum decrease of up to 2.5%p. The graph on the right in Figure 10 depicts the responses of the export volume to a one standard deviation shock in US monetary policy uncertainty. Similar to the Russian geopolitical risk case, the initial impact of uncertainty is not significant. However, over time, its effect gradually increases, leading to a maximum decrease in exports of up to 2.5%p.

FIGURE 10.

IMPACT OF UNCERTAINTY SHOCKS ON EXPORT VOLUMES

Note: Shaded areas show the 90% confidence interval of the impulse response.

4. Additional Results: Heterogeneous Effects by Country

In this section, we explore the heterogeneous impacts of external uncertainties on exports. Specifically, we consider the heterogeneity of exports in two dimensions: export destination countries and export products.

If the overall economic environment of a specific country becomes more vulnerable to external uncertainties, Korean exports to that country are likely to decrease significantly more than those to other regions. The geopolitical risks in Russia and uncertainty in US monetary policy have clear regional origins, suggesting that their effects are relatively strong in Russia and the US, respectively. Therefore, first we investigate whether there is a significant relationship between regional export responses and the regions where external uncertainties originate. To do this, we include dummy variables for the regions where uncertainty arises, as well as their interaction terms with external uncertainties. Considering that geopolitical risks in Russia may directly impact the Commonwealth of Independent States (CIS) region, we add dummy variables for all countries in the CIS region and corresponding interaction terms with the geopolitical risks in Russia. Similarly, given that uncertainties in US monetary policy may directly affect the North American region, we create dummy variables and corresponding interaction terms with the North American region.

Table 1 shows the results. The interaction terms of geopolitical risks in Russia are estimated to be negative for all lags, and especially after five to six months, the impact becomes statistically significant. This indicates that if geopolitical risks in Russia increase, Korean exports to the CIS countries, which are closely related to Russia, decrease significantly more compared to those to other regions. However, uncertainty in US monetary policy shows somewhat different results. Specifically, the interaction terms are estimated to be positive for all lags and even statistically significant for several lags (h = 0,1,5,6,7).

TABLE 1

HETEROGENEOUS IMPACT OF UNCERTAINTY SHOCKS BY COUNTRY : ORIGIN

Note: 1) Regressions are weighted by export value; 2) Two-way clustered standard errors by productⅹtrading country and times in parenthesis.

The results are somewhat counterintuitive, as North America is the region where uncertainty is triggered. This suggests the presence of another important country-specific characteristic beyond country of origin that influences the impact of an uncertainty shock. One possible candidate is the country’s income level. Carrière-Swallow and Céspedes (2013) find that the impact of uncertainty triggered in the United States (VIX) varies depending on the country’s income level. In particular, they confirm that emerging countries are significantly more affected by US uncertainty compared to advanced countries. They attribute this result to the underdeveloped financial markets in emerging countries. Motivated by their findings, we additionally control for the country’s income level given that the North American region, including the United States, has a relatively high income level compared to those of other countries.

Table 2 illustrates that the effects of external uncertainty vary depending on a country’s income level. In this analysis, we additionally include interaction terms between a country’s income level and external uncertainty measures. Specifically, a variable representing a country’s income level is constructed, as follow: countries are classified into five categories, from the bottom 20% to the top 20% for each period, with variables assigned from 1 to the bottom 20% countries to 5 to the top 20% countries.

TABLE 2

HETEROGENEOUS IMPACT OF UNCERTAINTY SHOCKS BY COUNTRY: ORIGIN AND INCOME LEVEL

Note: 1) Regressions are weighted by export value; 2) Two-way clustered standard errors by product ⅹ trading country and times in parenthesis.

We find that the effects US monetary policy uncertainty tend to be weaker in exports to countries with higher incomes, with these differential effects being statistically significant for up to seven months. In addition, regarding geopolitical risks in Russia, exports to countries with higher incomes respond less, with this differential effect being statistically significant from the eighth month.

Moreover, the counterintuitive results from the previous analysis - the relatively large decrease in exports to the North American region - are no longer observed. In particular, the interactions between the North American region dummy variable and US monetary policy uncertainty are estimated to be negative, confirming that Korean exports to regions where uncertainty is triggered decreased more significantly. However, both types of geopolitical risk associated with Russia and US monetary policy uncertainty do not show persistently significant different impacts based on the region where the uncertainty originated. Specifically, the responses to Russian geopolitical risks are only statistically significant at h = 5, while those to US monetary policy uncertainty are only statistically significant at h = 3.

In summary, we find that both Russian geopolitical risk and US monetary policy uncertainty have varying impacts on Korean exports to destination country depending on the country’s income level: low-income countries especially experiencing a larger negative impact of these uncertainties. We also find that Korean exports to regions where uncertainty originates tend to decrease relatively more compared to exports to other regions.

5. Additional Results: Heterogeneous Effects by Product

External uncertainty shocks can also have heterogeneous effects on exports according to the product type. From the analysis in the previous section, we observe that an increase in external uncertainty reduces Korea’s investments significantly. Assuming that our trading partner countries experience similar conditions, their real investments are also likely to be significantly affected. This would lead to a larger decrease in exports of items used for investments. Therefore, in this section, we examine whether the responses to external uncertainty shocks differ among consumer goods, primary products, intermediate goods, and capital goods. Specifically, we extend our model further by introducing a dummy variable for each export category, and its interaction terms with external uncertainty variables.

Figure 11 shows the effect of geopolitical risk in Russia. We find that most export items are affected by an increase in geopolitical risk in Russia. In particular, the growth rate of capital goods exports decreases by up to 7.39 percentage points compared to other goods, showing a relatively large decrease.

FIGURE 11.

DIFFERENTIAL IMPACTS OF RUSSIAN GEOPOLITICAL RISKS BY PRODUCT

Note: Shaded areas show the 90% confidence interval of the impulse response.

Figure 12 shows the export responses to an increase in uncertainty in US monetary policy. As can be seen in each graph, the growth rate of capital goods exports decreases by up to 7.07 percentage points, showing a relatively large decrease compared to other goods.

FIGURE 12.

DIFFERENTIAL IMPACTS OF US MONETARY POLICY UNCERTAINTY BY PRODUCT

Note: Shaded areas show the 90% confidence interval of the impulse response.

Although capital goods exports respond more strongly to both types of uncertainty, whether the difference is statistically significant varies depending on the type of uncertainty. As shown in Figure 13, it is somewhat difficult to distinguish statistically the decrease in capital goods exports due to geopolitical risks in Russia from the decrease in exports of other items. However, regarding the uncertainty in US monetary policy, a statistically significant difference can be observed for a considerable period in the initial stage.

FIGURE 13.

DIFFERENTIAL IMPACTS OF UNCERTAINTY SHOCKS BY PRODUCT

Note: Shaded areas show the 90% confidence interval of the impulse response.

In sum, although a decrease in exports due to increased external uncertainty shows heterogeneous patterns across different items, the difference is not statistically significant with regard to geopolitical risks in Russia. In contrast, it is evident in the case of US monetary policy uncertainty shocks. Specifically, the significant decrease in capital goods exports due to uncertainty in US monetary policy aligns earlier findings in the literature (Bloom, 2009), which suggests that uncertainty propagates to the real economy through reduced investments.

1. Baseline Regression Model

We estimate a panel regression model similar to the previous analysis. First, we estimate the baseline model, which is defined as follows:

The dependent variable y_i,t+h represents the h-period-ahead value of y_i,t, and y_i,t denotes the change in log real tangible assets. Here, the index i refers to individual firms, and t represents time. GPU_t and MPU_t represent the geopolitical risk in Russia and uncertainty in US monetary policy, respectively. X_i,_t represents the firm-level characteristics that could affect investment decisions, as noted in the literature (Gulen and Ion, 2016; Husted, Rogers, and Sun, 2020; Caldara and Iacoviello, 2022). It includes the ratio of cash flow to total assets, log growth in real sales, Tobin’s Q, and the ratio of exports to total sales. The key macro variables, denoted here by Z_t, are controlled to identify the impact of external uncertainty accurately, including the real-world GDP, the USD/KRW exchange rate (quarterly average), Korean real GDP, Korean CPI, the Korean policy interest rate, and the Korean unemployment rate. All macro variables, except for policy interest rates and unemployment rates, use the log growth rate from the previous year. Furthermore, industry (KSIC 3-digit level) and quarterly fixed effects are included to control for seasonality and industry-specific characteristics. Firm-level fixed effects are included to control for persistent company characteristics. Additionally, two-way cluster standard errors clustered by company and time are used. Because the impact of uncertainty on macro real variables typically lasts up to one year, we examine the dynamic responses up to one year after the uncertainty shocks.

2. Data

We use firm-level financial statements data provided by KIS-VALUE. As discussed earlier, the impact of external uncertainty typically materializes within a few months and generally dissipates within one year. Therefore, to identify the impact of uncertainty rigorously, it is appropriate to use data with a frequency shorter than that of annual data. In this study, our analysis focuses on listed firms that provide quarterly balance sheet information.

Given that manufacturing production is significantly affected by uncertainty, we limit our analysis to firms in the manufacturing industry. Following existing literature (Ottonello and Winberry, 2020), we perform data cleaning and remove outliers, after which we deflate firm-level variables using GDP deflators or investment deflators. Detailed explanations and sample statistics of the firm-level variables used here are included in the Appendix. In line with existing studies (Jeenas, 2019; Ottonello and Winberry, 2020), we use the change in real tangible assets as a variable representing investment decisions by firms.

The Russian geopolitical risks and US monetary policy uncertainty are identical to those used in the previous analysis. World GDP is downloaded from Global Insight7 data and Korean macro variables are based on data from Statistics Korea or the Bank of Korea. The sample period is from the first quarter of 2000 to the fourth quarter of 2021. Because we use the growth rate for most variables, the actual sample period begins in the first quarter of 2001. As before, dummy variables for the global financial crisis are also controlled.

3. Baseline Result: Average Effect

Figure 14 shows how a firm’s investment decision changes when external uncertainty increases. The graph on the left depicts the impact of a one standard deviation increase in Russian geopolitical risk. Specifically, the growth rate of real tangible assets decreases by up to 1.3%p immediately and then gradually weakens over time.

FIGURE 14.

IMPACT OF UNCERTAINTY SHOCKS ON INVESTMENT DECISIONS

Note: Shaded areas show the 90% confidence interval of the impulse response.

The graph on the right shows the impact of US monetary policy uncertainty. When US monetary policy uncertainty increases by one standard deviation, the growth rate of real tangible asset decreases by up to 1.0%p with a lag, and the impact disappears within a year. Unlike Russian geopolitical risk, the immediate effect of US monetary policy uncertainty is not statistically significant. However, a negative impact gradually expands from the second quarter onwards.

4. Additional Results: Heterogeneous Effects by Firm

In this section, we investigate whether the impact of uncertainty varies depending on the firm-level characteristics. Existing literature suggests that an increase in uncertainty affects a firm’s investments through two main mechanisms. The first is the real option channel, as proposed by Bloom (2009), mainly due to the irreversibility of investments. Investment decisions typically involve a certain amount of fixed costs. Accordingly, as uncertainty in the business environment increases, firms tend to delay their decisions. The second channel is the financial friction mechanism proposed by Gilchrist, Sim, and Zakrajšek (2014). Increased uncertainty in the business environment can lead to higher bankruptcy risks and raise a firm’s financial costs. Higher financing costs cause firms to reconsider their decisions on investments, thereby negatively impacting investments. Therefore, in this section, we explore as well the transmission channels of uncertainty by specifically controlling for firm-level variables related to the two main channels above.

To capture the nature of irreversibility, we employ variables suggested in the literature (Gulen and Ion, 2016; Husted, Rogers, and Sun, 2020; Kim et al., 2022). First, we use the ratio of fixed assets to total assets, indicating that firms with a higher ratio tend to make large investments with high fixed costs. Based on previous studies, firms are classified into five categories according to their fixed assets ratio at each time period, assigning values ranging from 1 to the bottom 20% to 5 to the top 20%. The variable, labelled irre1, can be interpreted as increasing capital irreversibility as the value rises. The second variable is based on the fact that capital subject to rapid depreciation or leased from other firms may have relatively low sunk and fixed costs. Specific variables reflecting these factors are generated according to the literature (Kim et al., 2022). After calculating depreciation and lease costs relative to total fixed assets, firms are divided into three groups at each time point. A value of 3 is assigned to firms with low depreciation and low lease costs, 2 to firms with either high depreciation and low lease costs or low depreciation and high lease costs, and 1 to firms with high depreciation and high lease costs. This variable, defined as irre2, is associated with increasing capital irreversibility as its value increases. Table 3 reports the coefficient estimates of interaction terms between the degree of these types of irreversibility and uncertainty shocks, illustrating the differential impact of uncertainty shocks across different firms.

TABLE 3

HETEROGENEOUS IMPACTS OF UNCERTAINTY SHOCKS BY FIRM : REAL OPTION CHANNEL

Note: Two-way clustered standard errors by firm and time in parenthesis.

Specifically, firms with high investment irreversibility (irre1) are strongly affected by uncertainty. Geopolitical risks in Russia have heterogeneous effects until the third quarter after a shock, while uncertainty in US monetary policy continues to have differential impacts until the fourth quarter. Similarly, firms with high investment irreversibility (irre2), based on depreciation and lease costs, also exhibit a strong response to uncertainty. The heterogeneous effects are statistically significant up to the third quarter after a shock for geopolitical risks in Russia and up to the second quarter after a shock for uncertainty in US monetary policy.

We also investigate the financial channel of uncertainty shocks. We employ firm-level credit ratings and leverage, as presented in earlier studies (Alfaro, Bloom, and Lin, 2018; Ottonello and Winberry, 2020), to capture the degree of financial constraints. We compute the interaction terms between the degree of financial friction and uncertainties to identify heterogeneous effects. Credit ratings, provided in KIS-VALUE, are divided into a total of ten grades. Consistent with the previous analysis, these grades are divided into five categories. A higher value of the constructed variable indicates a low rating and a higher level of financial constraint. Furthermore, the leverage of each firm is defined as total debt divided by total assets. Based on the debt ratio at each time point, companies are divided into five groups, with values ranging from 1 assigned to the bottom 20% to 5 to the top 20%. Therefore, an increase in this variable indicates a high debt burden level, which can be interpreted as a high level of financial constraints.

Table 4 shows the heterogeneous effect of uncertainty based on financial friction. Unlike investment irreversibility, heterogeneous effects of Russian geopolitical risk are not observed. However, heterogeneous effects based on credit ratings are statistically significant from the second quarter to the third quarter after the shocks to US monetary policy uncertainty. Additionally, heterogeneous effects based on debt burden are statistically significant from the immediate aftermath of shocks to the first quarter.

TABLE 4

HETEROGENEOUS IMPACTS OF UNCERTAINTY SHOCKS BY FIRM : FINANCIAL CHANNEL

Note: Two-way clustered standard errors by firm and time in parenthesis.

In sum, we find that both the geopolitical risk of Russia and the uncertainty of US monetary policy reduce investment. This effect is particularly pronounced for firms with high irreversibility of their investments. Investment irreversibility, primarily caused by fixed costs, is considered as a real friction in the literature. Therefore, both types of uncertainty lead to a contraction in the domestic real economy due to real friction. However, the geopolitical risk of Russia is largely unrelated to financial friction factors, whereas the uncertainty of US monetary policy impacts the real economy through financial friction channels. The uncertainty of US monetary policy may have a greater impact on the financial markets compared to other uncertainties, making this result somewhat intuitive.