Using a sample of comparably sized public listed and private firms from nine European countries, we show that public firms reduce their investments by about 50% more than private firms in response to an increase in policy-related uncertainty. We find suggestive evidence that this can be explained by public firms’ management being typically subject to greater shareholder scrutiny than private firms’ management. Furthermore, only public firms invest more efficiently when confronted with uncertainty. Thus, private firms may benefit from emulating the decision-making processes of public firms in uncertain times.
Hinweise
We are grateful for the comments and insights of Serkan Akguc, Philip Mensing, and the participants of the 7th Paris Financial Management Conference.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
1 Introduction
Government agencies regularly alter and pass new regulations that companies must comply with. This leaves the entire business sector in a continual state of uncertainty as to if, when, and how new economic policy proposals will be signed into law. Uncertainty alone can lead to serious economic repercussions (Bloom 2009) by, among other things, depressing corporate investment activities (e.g., Kang et al. 2014; Baker et al. 2016). Real options theory attempts to explain such behavior by the desire of corporate managers to avoid costly mistakes. By postponing part of their investment projects, perhaps even indefinitely, managers try to wait until the uncertainty is removed and more informed decisions can be made (Bernanke 1983; McDonald and Siegel 1986; Ingersoll and Ross 1992; Dixit and Pindyck 1994).
In light of the sharp rise of policy-related uncertainty in Europe—heavily fueled by Britain’s withdrawal from the European Union—we seek to advance this debate by investigating cross-sectional differences in the investment sensitivity of public and private European firms to Baker et al. (2016) Economic Policy Uncertainty (EPU) index. Despite private firms typically accounting for the largest share of a country’s GDP (see Giannetti 2003), most of the empirical literature on uncertainty focuses exclusively on publicly listed firms (e.g., Gulen and Ion 2016; Kim and Kung 2017). However, private companies are fundamentally different from their public counterparts. They must comply with different legal requirements and operate under different ownership structures. Such fundamental differences can lead to diverging investment sensitivities to growth opportunities (Mortal and Reisel 2013; Asker et al. 2015; Gilje and Taillard 2016) and may thus also lead to contrasting investment decisions under uncertainty. Therefore, we are convinced that including private firms in our analysis and exploring possible differences between the two types of firms represents a more comprehensive and appropriate approach to gauge the actual impact of policy uncertainty on corporate investment.
Anzeige
The goal of this paper is threefold. First, we analyze whether EPU can depress investment decisions of private firms. Second, we examine whether there are significant differences in the investment sensitivity of public and private firms to EPU. Third, do public and private firms invest more efficiently in times of policy-related uncertainty?
Our firm-level analysis is based on a dataset containing financial information on a large set of public and private European firms from nine countries between 2009 and 2017. Firm-level information comes from the Amadeus Database by Bureau van Dijk.
To preview our results, we find a negative contemporaneous relation between private firms’ fixed asset investment and uncertainty surrounding future economic policy decisions. This result withstands a range of sensitivity tests, including specifications aimed at eliminating potential endogeneity concerns and alternative sample selection criteria. It shows that the theoretical prediction of real options theory, which aims to capture the value of managers’ temporal flexibility—independent of their work domain—seems to apply not only to public but also to private firms. Moreover, we demonstrate that the impact of EPU on investment is not felt equally by all firms. Public firms reduce their investments by about 50% more than private firms in response to an increase in policy-related uncertainty. We find suggestive evidence that this cross-sectional heterogeneity can be attributed to public firms’ management being typically subject to greater scrutiny and myopic pressure from their shareholders than private firms’ management (e.g., Asker et al. 2015; Gao et al. 2017). Specifically, we show that increases in shareholder power, transparency requirements, and liability standards are associated with fewer investments under uncertainty. This indicates that managers under higher shareholder scrutiny are likely to anticipate more severe negative consequences in the event of poor investment decisions, which fosters a larger degree of uncertainty averse decisions.
There are at least three possible explanations for why private European firms should face a lower baseline level of shareholder scrutiny in comparison to their public counterparts.
Anzeige
First, private firms are often managed by the major shareholder (founder) (see Asker et al. 2015), making it impossible to veto or fire the CEO. In contrast, CEOs of public firms act as mere agents of shareholders, leaving them principally vulnerable to being fired. In fact, Fisman et al. (2014) note that public shareholders are prone to call for the CEO’s dismissal when short-term performance deteriorates, even if such events are entirely outside the CEO’s control. Unsurprisingly, public firms display higher CEO turnover rates and have greater CEO and top management turnover-performance sensitivities than private firms (Lel et al. 2014; Gao et al. 2017).
Second, unlike private firms, public firms are obligated to hold an annual ordinary shareholders’ meeting at which management is subject to open examination by their shareholders.
Third, while European law obliges private firms to report annual financial statements, public listed firms face even stricter disclosure requirements based on the Transparency Directive 2004/109/EC and country-specific stock exchange rules. The Transparency Directive, which was modified in 2013 by Directive 2013/50/EU, requires, among other things, the publication of half-yearly financial reports.1 As a result, public firm managers should increasingly be aware of the ease and frequency with which shareholders can identify poor investment decisions.
In light of these differences, we expect public firm managers to cut their investments more acutely than private firm managers in times of uncertainty, as they should be even more concerned about having to account for their investment mistakes or getting laid off because of them.
As to whether public and private firms invest more efficiently in periods of uncertainty, our findings suggest that only public firms adapt to these changing circumstances. We find that in response to increases in policy-related uncertainty, public firms invest relatively more in high yielding projects. In comparison, the consolidated investments of private firms do not predict a significant increase in relative profitability when faced with uncertainty. Although insightful, these findings do not necessarily permit the conclusion that being publicly listed is the superior form of organization to adequately cope with spells of uncertainty. This is because private firms, unlike public ones, could (theoretically) always invest in the most efficient manner and therefore simply cannot invest even more efficiently in uncertain times.
From a time-series point of view, we observe that EPU negatively affects the investment activities of both public and private firms for an approximate duration of two years. After this period the negative impact begins to subside and will even turn positive for public firms. It is conceivable that this cross-sectional variation in recovery responses several years after the initial shock might be connected to the idiosyncratic increase in investment profitability rates in uncertain times. This could encourage public firms to pursue new investments at a higher rate than before. As far as private firms are concerned, their inability to identify the projects that would justify immediate implementation the most potentially leads them to refrain from offsetting the initial investment decline in later periods.
Our paper contributes to the general literature on real options theory (e.g., Schwartz and Trigeorgis 2004; Grenadier and Malenko 2010). We show that the organizational form of a firm—public listed or private—is of critical importance in assessing the impact of policy-related uncertainty on corporate investment decisions. Having identified differences in the degree of shareholder scrutiny as a plausible explanation for this finding, this study contributes to the works of Baker et al. (2016), Gulen and Ion (2016), and Kim and Kung (2017) in demonstrating that the link between EPU and investment is not just a function of government dependencies or the degree of investment irreversibility. Thus, this study underscores the role of investor protection laws (La Porta et al. 1998, 2006; Djankov et al. 2008) in periods of uncertainty.
Assuming private firms do not always invest in the most efficient way possible, our results support the policy implication that if they were to adopt the decision-making processes of public companies in uncertain times, this could help them prioritize those projects that most warrant immediate implementation. Thus, when confronted with uncertainty, private firm decision-makers should be more transparent about their planned investments and actively seek and incorporate feedback from their peers.
The remainder of this paper is organized as follows. Section 2 outlines the data and sampling methods used in the study. Section 3 details our empirical results, starting with the analysis of the relationship between policy-related uncertainty and private firms’ investments. This is followed by the examination of cross-sectional differences between the investment sensitivities of public and private firms to uncertainty and the channel that serves to explain their heterogeneous responses. Section 4 tests whether public and private firms invest more efficiently in periods of uncertainty. Section 5 analyzes the link between policy-related uncertainty and investment over time. Section 6 contains robustness tests. Section 7 concludes.
2 Data
Our dataset comprises firms from nine European countries covering the period 2009 to 2017. We obtain firm-level fundamentals from the Amadeus Database by Bureau van Dijk. Amadeus provides accounting data for a set of public and private European firms. These data are available because European law obligates private firms to disclose annual financial reports. Public firms face even stricter disclosure requirements.
The focus of this study is to analyze the effect of economic policy uncertainty on corporate investment. However, the Amadeus Database does not report any investment measure such as capital expenditures. For this reason, we follow Asker et al. (2015) and define Investment as the annual change in total fixed assets scaled by beginning-of-year total assets. To capture the impact of uncertainty related to future economic policies, the firm-level data are complemented by Baker et al.’s (2016) EPU index.
2.1 Economic policy uncertainty
As of January 2020, the website www.policyuncertainty.com provides policy uncertainty measures for ten European countries based on the methodology of Baker et al. (2016). Separate monthly policy uncertainty indexes are created for France, Germany, Italy, Spain, and the United Kingdom by Baker et al. (2016). Additionally, the same index has been calculated for Greece by Fountas et al. (2018), for Ireland by Zalla (2017), for the Netherlands by Kroese et al. (2015), and for Sweden by Armelius et al. (2017). A policy uncertainty index is also available for Russia, but following Giannetti (2003) and Mortal and Reisel (2013), we exclude firms from Eastern European countries from our sample due to insufficient data availability in the Amadeus Database. Thus, our empirical analysis is limited to nine European countries.
To measure economic policy uncertainty, newspaper articles are: (1) counted that include keywords regarding: policy, economic, and uncertainty; and (2) set in proportion to the total count of articles publicized by the corresponding newspaper outlet in the same month. The list of keywords varies from country to country depending on the native language and regional peculiarities. The index draws on leading newspapers in individual countries, e.g. Le Monde in France or Financial Times in the U.K. To generate an index at the country-level, “[...] each monthly newspaper-level series [is standardized] to unit standard deviation [...] and [averaged] across the papers by month” (Baker et al. 2016, p. 1599) before being normalized to an average value of 100 prior to 2011.
Fig. 1
Economic Policy Uncertainty in the U.K
×
Figure 1 shows the monthly time-series of the economic policy uncertainty index for the U.K. As expected, Fig. 1 illustrates that the highest values are achieved around the Brexit referendum. This observation supports the assumption that the index is a reliable indicator of economic policy-related uncertainty. Regardless of this meaningful observation, Baker et al. (2016) also conduct several formal tests to ensure that their measure adequately captures the uncertainty pertaining to future economic policies. Besides, Fig. 1 shows that the time-series does not only indicate rare events—such as elections or referendums—but also exhibits substantial fluctuations over the entire sample period.
2.2 Control variables
On firm-level, we control for growth opportunities, cash flow, and size in all regressions. The investment literature typically uses Tobin’s q or sales growth as a measure of growth opportunities. Tobin’s q is commonly defined as the ratio between the market value of the company and the book value of its assets. However, the market values of private firms cannot be determined as they are not listed on a stock exchange. Therefore, similar to Mortal and Reisel (2013) and Gilje and Taillard (2016), we utilize sales growth as an alternative measure of growth opportunities because it can be calculated for both types of firms. Sales growth is calculated as the change in the firm’s turnover scaled by the turnover of the previous year.
Cash flow is reported in the Amadeus Database for all firms. For our analysis, the cash flow variable is scaled by beginning-of-year total assets. We measure the size of a company by the natural logarithm of its total assets. To account for the macroeconomic environment in a country, we include the annual GDP growth and inflation rate as provided by the World Bank as additional control variables.
Although we rely on the classic corporate finance literature to select firm-level control variables (e.g., Mortal and Reisel 2013; Foucault and Fresard 2014), we recognize that our list of controls is not exhaustive. For example, prior empirical studies find that a firm’s payout policy influences its investment behavior (Almeida et al. 2016; Wang et al. 2021) while being significantly correlated with Baker et al.’s (2016) EPU index (Smietanka et al. 2018; Pirgaip and Dinçergök 2019; Attig et al. 2021; Anolick et al. 2021). Unfortunately, information about private companies’ payout policies is not available. As a consequence, we must be conservative in the causal interpretation of our results.
2.3 Sample construction
Our sample covers the period 2009–2017. We classify firms into public listed and private firms. The Amadeus Database includes an identifier that indicates the listing status of every company. However, this identifier only provides information on the contemporaneous listing status and does not contain historical information. Therefore, we supplement this variable by extracting information on initial public offering (IPO) and delisting dates from the Osiris dataset by Bureau van Dijk. This allows us to elicit the annual listing and delisting status of every company in our sample. Consequently, a firm-year is indicated as ‘listed’ if the firm is quoted on a stock exchange in a given year and was reported to be unlisted prior to its IPO. Likewise, a firm-year is classified as ‘unlisted’ if the firm is not quoted on a stock exchange in a given year and was reported to be listed prior to its delisting. In line with Mortal and Reisel (2013), we classify private firms by their legal form as indicated in the dataset and exclude from our sample cooperatives, foreign companies, foundations, government enterprises, sole proprietorships, and unlimited partnerships.
In a manner similar to Baker et al. (2003) and McLean and Zhao (2014), we omit firm-years with total assets less than 10 million euros to reduce undue effects of small businesses. In addition, we exclude financial and (regulsated) utility firms from our sample—that is, firms with a one-digit SIC code of six and a two-digit SIC code of 49. All firm-level variables are winsorized at the 1% and 99% level in order to mitigate the impact of outliers.
In a last step, we drop all firm-year observations with missing data for Investment, Sales growth, Cash flow, or Size. These filters result in 291,415 private firm-year observations. The number of private firm-year observations is substantially larger than the number of public listed firm-year observations. They only amount to 4.5% of the private firm-year observations. Table 1 shows that the full sample covers 56,802 unique private firms and only 2,301 unique public listed firms.
Table 1
Descriptive statistics
Full sample
Matched sample
Public
Private
Difference
Public
Private
Difference
Total asset
Mean
2898.701
129.257
2769.444
770.740
480.307
290.433
SD
17,046.884
1235.407
8090.304
5265.316
Investment
Mean
0.035
0.018
0.017
0.039
0.017
0.023
SD
0.151
0.116
0.160
0.120
Sales growth
Mean
0.164
0.140
0.024
0.193
0.110
0.083
SD
0.775
0.674
0.834
0.637
Cash flow
Mean
0.066
0.084
− 0.018
0.069
0.085
− 0.016
SD
0.121
0.108
0.127
0.113
Size
Mean
18.957
17.479
1.478
18.617
17.897
0.720
SD
1.659
1.076
1.434
1.361
No. of obs.
13,128
291,415
9456
9456
No. of firms
2301
56,802
1786
2238
Note: This table contains summary statistics for the full and matched samples of public listed and private firms between 2009 to 2017. A description of the construction of the samples and details on the matching procedure are given in Sect. 2. Samples are divided into calendar years; financial years ending between 1 January and 31 May are counted towards ending in the preceding calendar year. The table presents means and standard deviations of the main firm-level variables used in our analysis as well as pairwise differences in means. All firm-level data are obtained from the Amadeus Database by Bureau van Dijk. Total asset is expressed in millions of euros. Investment is the annual change in total fixed assets scaled by beginning-of-year total assets, Sales growth is the change in the firm’s turnover scaled by the turnover of the previous year, Cash flow is reported in the Amadeus Database and it is scaled by beginning-of-year total assets, and Size is the natural logarithm of total assets. All firm-level variables are winsorized at the 1% and 99% level to reduce the impact of outliers
There are large differences between the public listed and private firm samples, especially with regard to their total assets. The average public firm has about 22 times more assets than the average private firm. To improve the comparability of these samples, we follow (Gilje and Taillard 2016) and try to find for every public company the private company that is most similar in terms of total assets in the same industry, country, and year. In those cases where there is no suitable match in the same industry and country with an overall similar asset amount for a public firm in the year of entering the sample, we attempt to find a match in the following year. Matched private and public listed firms stay in the sample for the entire duration of the public firm’s existence. In the event that a matched private firm stops existing, we try to identify a new match for the public firm during the exact same year, thus ensuring that the public firm will continue to belong to the matched sample. Similar to Asker et al. (2015) and Gilje and Taillard (2016), our matching procedure is based on replacements, allowing us to take within-firm effects into account in our analyses. To the degree that large private firms behave more like public firms, this matching procedure has the potential to bias the estimates towards detecting no differences between them.
The determined matched sample includes 18,912 firm-year observations. It comprises 2238 unique private firms and 1786 unique public listed firms. The difference in total assets reduces dramatically. In the full sample, the average public listed firm has about 22 times more assets than the average private firm. In the matched sample, the average public firm has only 1.6 times more assets than the average private firm.
For our matching procedure, we apply a caliper of 0.05 on the maximum total asset difference and use three-digit SIC industries. In Table 11 in the Appendix, we show that our results are robust to using other calipers (\(\pm \, 0.04\)), two-digit SIC, or four-digit NAICS industries.
Table 2 presents country-specific summary statistics for our main macroeconomic variables. As expected—due to the Brexit referendum—the country with the highest average EPU index in our sampling period is the U.K. The Netherlands and Sweden have the lowest mean EPU scores and are therefore the countries least affected by economic policy uncertainty in our sample. The mean values of GDP growth and Inflation are relatively similar across all nine European countries. The average GDP growth ranges between 0.184% and 2.696%. There are only two outliers: Ireland (6.657%) and Greece (− 2.942%). Both countries were heavily affected by the euro crisis in the late 2000s, but starting in 2012 the Irish economy began to recover significantly. The average Inflation varies from 0.307% in Spain to 1.666% in the U.K. Only Greece has a slightly negative average Inflation of − 0.384%.
Table 2
Descriptive statistics by country
Mean
SD
France
EPU
253.344
45.882
GDP growth
1.306
0.722
Inflation
0.814
0.330
Germany
EPU
168.627
33.495
GDP growth
2.130
1.292
Inflation
1.496
0.426
Greece
EPU
149.986
39.246
GDP growth
− 2.942
3.982
Inflation
− 0.384
1.163
Ireland
EPU
143.945
24.411
GDP growth
6.657
8.019
Inflation
0.676
3.204
Italy
EPU
124.488
25.607
GDP growth
0.184
1.625
Inflation
0.987
0.403
Netherland
EPU
100.355
30.482
GDP growth
1.434
1.357
Inflation
0.800
0.472
Spain
EPU
128.349
24.758
GDP growth
0.695
2.471
Inflation
0.307
0.434
Sweden
EPU
101.071
6.436
GDP growth
2.696
1.933
Inflation
1.505
0.498
United Kingdom
EPU
298.994
135.716
GDP growth
2.029
0.472
Inflation
1.666
0.546
Note: This table presents summary statistics for the macroeconomic variables used in the main part of our analysis. Economic Policy Uncertainty (EPU) is the yearly average of a country’s news-based policy uncertainty index, as developed by Baker et al. (2016). GDP growth and Inflation are retrieved from the World Bank’s World Development Indicators. GDP growth is the annual percentage change in GDP and Inflation is the annual GDP deflator
3 The relationship between policy uncertainty and public and private firms’ investment decisions
We begin our empirical analysis by estimating whether private firms’ fixed asset investments are sensitive to changes in economic policy uncertainty. The existing empirical literature finds a negative link between policy-related uncertainty and the investment decisions of public firms (e.g. Kang et al. 2014; Gulen and Ion 2016; Baker et al. 2016). However, because private firms differ fundamentally from their public counterparts in terms of their legal requirements or ownership structures, we cannot draw reliable conclusions about private firms’ investment sensitivity to EPU based purely on public firm results. Mortal and Reisel (2013), Asker et al. (2015), and Gilje and Taillard (2016), for example, show that public and private firms adjust their investments very differently to changing economic circumstances; the same may thus also apply in times of economic policy uncertainty. Therefore, we run the following regression model for our sample of only private European firms:
where i indexes firms, t indexes years, and j denotes countries. The dependent variable, Investment, is the annual change in total fixed assets scaled by beginning-of-year total assets. The variables \(\alpha _{i}\) and \(\alpha _{t}\) are firm and year fixed effects, respectively. Our independent variable of interest is log(EPU). It is the natural logarithm of Baker et al. (2016) economic policy uncertainty measure. SG stands for sales growth and it captures the change in the firm’s turnover scaled by the turnover of the previous year. CF is cash flow, which is scaled by beginning-of-year total assets. Firm Size is captured by the natural logarithm of total assets. GDP growth is the annual percentage change in GDP and Inflation is measured by the annual GDP deflator. In all regressions reported, standard errors are double clustered at the firm- and year-level.
In line with previous literature, regression 1 in Table 3 shows that the variable of interest (EPU) has a negative relationship with public firm-level investment. Similarly, regression 2 reveals that an increase in policy-related uncertainty is associated with a significant decrease in private firms’ investment decisions. This demonstrates that the theoretical prediction of real options theory applies not only to public but also to private firms. Specifically, the coefficient on the natural logarithm of EPU is \(-\,0.031\) (t-statistic = \(-\,3.052\)) and Investment in our private firm sample has a mean value of 0.018. This implies that, ceteris paribus, a 10% increase in the level of policy uncertainty is associated with a contemporaneous decline in private firm Investment equivalent to \(0.031/(10\times 0.018)\) = 17.222% of the sample mean. Hence, this finding is both statistically significant as well as economically relevant. With regard to the additional control variables, which remain qualitatively similar in all specifications, their relations with Investment do not exhibit salient discrepancy to the extant empirical finance literature.
Table 3
Baseline investment regressions
Dependent variable: Investment
Private firm sample
Public firm sample
Full sample
Matched sample
(1)
(2)
(3)
(4)
(5)
(6)
Log(EPU)
− 0.031***
− 0.051***
− 0.032***
− 0.031***
− 0.051***
− 0.033***
(− 3.052)
(− 3.209)
(− 3.126)
(− 3.027)
(− 4.020)
(− 2.666)
Log(EPU) \(\times\) Public listed
− 0.018**
− 0.034***
(− 2.024)
(− 4.595)
Public listed
0.100**
0.281***
(2.102)
(4.449)
Sales growth
0.014***
0.034***
0.016***
0.016***
0.026***
0.026***
(18.591)
(8.139)
(19.393)
(19.403)
(9.110)
(9.027)
Cash flow
0.135***
0.312***
0.144***
0.144***
0.200***
0.200***
(21.636)
(8.954)
(22.819)
(22.830)
(9.695)
(9.672)
Size
0.084***
0.111***
0.086***
0.086***
0.104***
0.105***
(7.698)
(6.487)
(7.701)
(7.694)
(6.754)
(6.845)
GDP growth
0.000
0.001
0.000
0.000
− 0.001
− 0.001
(0.095)
(0.940)
(0.098)
(0.125)
(− 0.663)
(− 0.689)
Inflation
− 0.002
− 0.004
− 0.002
− 0.002
0.000
0.000
(− 0.905)
(− 1.596)
(− 0.959)
(− 0.975)
(0.051)
(0.084)
Firm fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Adjusted R\(^{2}\)
0.076
0.15
0.079
0.079
0.113
0.115
Observations
291,415
13,128
304,543
304,543
18,912
18,912
Note: This table reports firm-level regression estimates of Eq. (1). The dependent variable is Investment. The independent variables of interest are Log(EPU) and Public listed. Log(EPU) is the natural logarithm of a country’s economic policy uncertainty index. Public listed is an indicator variable that takes the value of 1 for each year the firm is identified as being listed on a stock exchange and 0 otherwise. See Tables 1 and 2 for further variable definitions. Regressions 1 and 2 show the results for the two samples of private or public firms, respectively. Regressions 3 and 4 include results for the full sample of private and public firms. Regressions 5 and 6 show results for the matched sample. All regressions include year dummies as well as firm fixed effects. Standard errors are clustered on both firm and year. Robust t-statistics are reported in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively
In regression 3, we estimate the same equation for our full sample of public listed and private firms, consisting of 304,543 firm-year observations. The coefficient on log(EPU) loads negatively at the 1% significance level. It indicates that, on average, a rise in EPU is associated with a reduction in public and private firms’ investment.
To investigate potential cross-sectional differences in investment sensitivities between public listed and private firms, we introduce an additional interaction term to Eq. (1). It interacts the natural logarithm of EPU with an indicator variable which is equal to unity if the firm is listed on a stock exchange, and zero if it is a private firm. The results are reported in regression 4. They show that the relationship between policy uncertainty and corporate investment is a function of the organizational form of the firm. The negative and significant interaction term of \(-\,0.018\) (t-statistic = \(-\,2.024\)) demonstrates that investments of public listed firms are more sensitive to economic policy uncertainty than investments of private firms. To quantify the total impact of EPU on public firm investment, we need to aggregate the coefficients on log(EPU) and on the log(EPU)-Public listed interaction term. The log(EPU) coefficient in this regression is \(-\,0.031\). Thus, a 10% increase in EPU is associated with a decrease in public firm Investment of about \((-0.031-0.018 \times 1)/10 = -\,0.005\). Investment in our full sample has a mean value of 0.019,2 so this represents a reduction equivalent to 26.316% of the sample mean. As for private firms, Investment only declines by 16.316% relative to the sample mean, which corresponds to approximately three fifth of the decrease of public firms.
Regressions 5 and 6 in Table 3 are based on our matched sample. Their results are comparable to those observed in the full sample. Focusing on the non-uniformity in the cross-section, regression 6 provides additional evidence that the relationship between EPU and corporate investment is a function of the type of firm—that is, being public listed or private. The coefficient on the interaction term is \(-\,0.034\) (t-statistic = \(-\,4.595\)). Investment in our matched sample has a mean of 0.028. These results imply that when EPU assumes a level 10% higher as what is used to be, public firm Investment is 23.929% lower (relative to the sample mean) than before the increase. This compares to a reduction in Investment by 11.786% of the sample mean if the firm were not listed on a stock exchange. Hence, public firms cut their investments by about 50% more than private firms in response to an increase in policy-related uncertainty.
In the following subsection, we provide empirical evidence that this cross-sectional variation is likely to be related to public firms’ management being typically subject to greater shareholder scrutiny than private firms’ management.
3.1 Shareholder scrutiny
Real options theory predicts that when confronted with uncertainty, managers tend to respond by reducing their contemporaneous investment activity in anticipation that they would otherwise make costly investment mistakes (e.g., Bernanke 1983; Dixit and Pindyck 1994). We argue that this wait-and-see behavior should intuitively be reinforced when managers are aware that they are under a higher level of scrutiny and will face more severe consequences if such mistakes were indeed to occur.
Prior literature suggests that the management of public companies typically faces greater shareholder scrutiny and myopic pressure than the management of private companies (e.g., Asker et al. 2015; Gao et al. 2017). Consequently, this should help explain the heterogeneous investment-EPU sensitivities observed in Table 3.
We submit that there are at least three reasons why public European firms should be exposed to higher levels of scrutiny from their shareholders compared to their private counterparts.
First, most private firms are managed by the owner (major shareholder) (see Asker et al. 2015). This frequently makes it impossible to veto their decisions, let alone fire them. In contrast, the CEOs of public firms are merely agents of their shareholders, which theoretically puts them in constant danger of being dismissed. For instance, Fisman et al. (2014) find that public shareholders are quick to demand the CEO’s removal when short-term performance deteriorates, even when such events are completely outside the CEO’s control. Not surprisingly, public firms display higher CEO turnover rates and exhibit greater CEO and top management turnover-performance sensitivities than private firms (Lel et al. 2014; Gao et al. 2017).
Second, listed firms are required by law to hold an annual ordinary shareholders’ meeting at which management is in principle subject to open scrutiny by their shareholders. Conversely, private firms are not legally compelled to hold such a meeting and thus do not have to publicly answer to their shareholders.
Lastly, while European law obliges private firms to report annual financial statements, public listed firms face even stricter disclosure requirements. According to the Transparency Directive 2004/109/EC—amended by Directive 2013/50/EU in 2013—public firms are required to publish half-yearly financial reports. Accordingly, public firm managers should be increasingly conscious of the ease and frequency with which shareholders can pinpoint poor investment decisions.
Given these differences in ownership structures and legal requirements, we expect managers of public firms to become even more apprehensive about making investments amidst the uncertainty, as they should be even more concerned than their private firm counterparts about having to account for their investment mistakes or getting laid off because of them.
In order to (indirectly) test this hypothesis, we rely on cross-country differences in public firms’ minority shareholder rights (Djankov et al. 2008; La Porta et al. 1998), disclosure requirements, and liability standards (La Porta et al. 2006). If there is any truth to our argument, we would expect publicly traded firms from countries with stronger shareholder rights and higher disclosure and liability standards to reduce their investments more acutely to an increase in EPU. As public and private firms are confronted with similar differences in legal requirements as the countries in our sample, the underlying intuition of such public firm results should principally be transferable to the public versus private firm domain as well. We test this prediction with the following regression model:
where Inv. Protection stands for four mean-centered investor protection measures: Anti-Self, Anti-Dir, Disclosure, and Liability. We define each of these country-level legal variables in detail in the Appendix. According to our hypothesis, we expect \(\beta _{2}\) to be negative and statistically significant.
The estimation results of Eq. (2) are reported in Table 4. In regressions 1 and 2, we alternatively use two minority shareholder rights indexes (Anti-Self and Anti-Dir) to directly quantify shareholders’ theoretical control over the decision-making processes within a listed firm. Intuitively, the greater the control of (minority) shareholders over corporate decision-making processes, the lower the ability of CEOs and top management to act independently and they should therefore be more cautious of their own shareholders. The negative and significant coefficients on the interaction term indicate that the better the shareholder protection in a country, the greater the investment reduction of public firms to policy-related uncertainty. Specifically, in regression 1, the coefficient on the interaction term between the natural logarithm of EPU and the mean-centered Anti-Self variable is \(-\,0.115\) (t-statistic = \(-\,5.789\)). Investment in our public firm sample has a mean of 0.035. Thus, a 10% increase in EPU is associated with a decline in Investment equivalent to 8.000% of the sample mean for a public firm that resides in a country with an average level of shareholder protection (e.g., Italy). This compares to a reduction in Investment by 19.500% relative to the sample mean if the firm were located in the U.K., the country with the highest level of Anti-Self of 0.35 in our sample. This is nearly two and a half times as strong a reaction as that of an average Italian firm.
Table 4
Investment regression with investor protection variables
Dependent variable: Investment
Anti-self
Anti-dir
Disclosure
Liability
(1)
(2)
(3)
(4)
Log(EPU)
− 0.028***
− 0.036***
− 0.035***
− 0.034**
(− 2.971)
(− 3.396)
(− 3.431)
(− 2.207)
Log(EPU) \(\times\) Inv. protection
− 0.115***
− 0.020***
− 0.186***
− 0.109***
(− 5.789)
(− 4.942)
(− 6.173)
(− 3.633)
Sales growth
0.033***
0.033***
0.033***
0.034***
(7.947)
(7.979)
(7.883)
(8.051)
Cash flow
0.309***
0.310***
0.309***
0.311***
(8.958)
(8.957)
(8.932)
(8.977)
Size
0.112***
0.112***
0.113***
0.111***
(6.689)
(6.642)
(6.759)
(6.485)
GDP growth
− 0.000
0.000
− 0.000
0.001
(− 0.255)
(0.171)
(− 0.206)
(0.547)
Inflation
− 0.003
− 0.003
− 0.003**
− 0.003
(− 1.587)
(− 1.641)
(− 1.967)
(− 1.086)
Firm fixed effects
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Adjusted R\(^{2}\)
0.154
0.153
0.153
0.151
Observations
13,128
13,128
13,128
13,128
Note: This table reports firm-level regression estimates of Eq. (2) for the full sample of only publicly listed firms. The dependent variable is Investment. The independent variables of interest are Log(EPU) and Inv. Protection. Log(EPU) is the natural logarithm of a country’s economic policy uncertainty index. Inv. Protection is a mean-centered placeholder variable that either stands for a measure of anti-self-dealing, anti-director rights index, disclosure requirements, or liability standards. Detailed variable definitions are provided in the Appendix. See Table 1 for firm-level variable definitions. All regressions include year dummies as well as firm fixed effects. Standard errors are clustered on both firm and year. Robust t-statistics are reported in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively
Regressions 3 and 4 interact our measures of disclosure requirements (Disclosure) and liability standards (Liability) with the natural logarithm of EPU, respectively. While stricter disclosure requirements increase the company’s transparency towards its shareholders, stronger liability standards allow investors to place managers under greater (legal) scrutiny if they provide misleading information. In both regressions, the coefficients on the interaction term load negatively at the 1% level. They indicate that the stronger the disclosure and liability standards, the greater the reduction in public firms’ investment in times of heightened economic policy uncertainty.
These results suggest that different levels of shareholder scrutiny—induced by varying minority shareholder rights and disclosure and liability standards—can explain cross-countries differences in public firms’ investment sensitivities to EPU. Since public and private European firms differ to a similar extent with regard to the legal requirements they need to adhere to, these findings provide some (indirect) explanation for the non-uniform reaction between the two types of companies as well.
4 Policy uncertainty and future operating performance
This section explores whether public and private firms invest more efficiently when exposed to increasing policy-related uncertainty. Theoretically, return on investment should be comparatively high when EPU is high as both public and private firms significantly reduce their investments due to their increasing fear of making costly mistakes (see Table 3). Consequently, both firms should primarily invest in their most valuable projects, i.e. those projects that justify immediate implementation rather than postponement. This line of reasoning is related to McLean and Zhao’s (2014) findings. They demonstrate that when firms’ financing costs are high, some of their projects will have to be abandoned. The ensuing consolidation of investments is then carried out in such a manner that the firms primarily realize their most profitable projects. Therefore, we run the following regression for both public and private firms:
The dependent variable is the average annual return-on-assets (ROA) measured over a 3-year period including the year in which the investment is undertaken. Annual ROA is captured by the variable used in the Amadeus Database for the return on total assets. It is defined as profit before taxation scaled by total assets. \({\textbf {X}}'\) denotes our typical set of control variables. According to our hypothesis, we expect the interaction term between investment and the natural logarithm of EPU to be positive and significant for both types of firms.
Table 5 presents the results of the operating performance regressions. We divide our matched sample into public and private firms. For expositional clarity, we only present the coefficients of interest (\(\beta _{1}\) to \(\beta _{3}\)). Regression 1 examines public firms. The coefficient on the interaction term between investment and the natural logarithm of EPU is 0.050 (t-statistic = 2.593). It shows that the greater the threat of policy-related uncertainty, the more efficiently public firms invest. In detail, a one unit increase in Investment portends a 0.011 rise in public firms’ ROA whenever they experience the average level of uncertainty of 163.240 in our sample. If EPU were to reach its maximum of 542.766 in our sample, public firms’ average ROA would increase by 0.071, which is more than six times as much as before. This result demonstrates that public firms invest relatively more in their most valuable projects when they have to deal with mounting economic policy uncertainty. Or, put differently, the quality of realized investments improves as uncertainty increases.
Table 5
Investment, EPU, and Ex post efficiency
Dependent variable: Average ROA
Public firms
Private firms
(1)
(2)
Investment
− 0.244**
− 0.051
(− 2.329)
(− 0.577)
Log(EPU) \(\times\) Investment
0.050***
0.007
(2.593)
(0.397)
Log(EPU)
0.016**
0.004
(2.179)
(0.576)
Control variables
Yes
Yes
Firm fixed effects
Yes
Yes
Year dummies
Yes
Yes
Adjusted R\(^{2}\)
0.200
0.243
Observations
6431
6361
Note: This table reports firm-level regression estimates of Eq. (3). The dependent variable is the Average ROA over a 3-year period (t, t + 1, and t + 2). The independent variables of interest are Investment and Log(EPU). Investment is defined as the annual change in total fixed assets scaled by beginning-of-year total assets. Log(EPU) is the natural logarithm of a country’s economic policy uncertainty index. The set of control variables includes Sales growth, Cash flow, Size, GDP growth, and Inflation. See Tables 1 and 2 for definitions of these variables. Regressions 1 and 2 present results for the public listed and private firms of the matched sample, respectively. All regressions include year dummies as well as firm fixed effects. Standard errors are clustered on both firm and year. Robust t-statistics are reported in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively
In regression 2—in contrast to public firms—the coefficient on the Investment-log(EPU) interaction term is insignificant. This estimate implies that a greater level of policy uncertainty is not associated with an increase in private firms’ investment profitability.
In summary, the estimates show that when faced with uncertainty, only public firms invest in a more profitable manner than in tranquil periods. Since both types of firms significantly reduce their investments in uncertain times, this suggests that public firms are more capable of identifying their most valuable projects—that is, those projects that warrant immediate execution rather than deferral.3
We suspect that these cross-sectional differences are again related to the fact that managers of public firms have to be more concerned about potential costly investment mistakes and the associated consequences than their private firm counterparts. This is because if they decided to pursue uncertain investments and they later prove to be underperforming, they would be more likely to lose their jobs than private firm decision-makers (Lel et al. 2014; Gao et al. 2017). Embracing the decision-making processes of public firms in uncertain times could therefore aid private firms in detecting those investment opportunities that most justify immediate implementation.
Consequently, in the face of uncertainty, private company decision-makers should increasingly incorporate feedback from their peers into their investment strategy. Having to present and justify investment ideas to others may not put their jobs at risk, but at least their reputation. This should increase the propensity to invest in projects with the highest degree of certainty for a given rate of expected profitability, i.e. easy-to-justify investments. This idea is related to the literature that finds that when investors have to engage and explain their investment decisions to human brokers instead of anonymously trading online (Barber and Odean 2001, 2002; Konana and Balasubramanian 2005) or when they have to justify their decisions to other members of an investment club (Barber et al. 2003), they tend to prefer investing in projects with lower levels of uncertainty. Konana and Balasubramanian (2005) find that online investors are partially afraid of their interactions with traditional brokers as they do not want to be judged by them of lacking in knowledge about the economics of the stock market.
We conclude this section with a small caveat regarding the interpretation of our results. On first sight, it might be tempting to deduce that public firms operate more profitably than private ones in uncertain times. While this might be the case, our results do not necessarily imply that. This is because private firms might have higher initial profitability rates that simply do not change during periods of uncertainty, but are still greater than the increased profitability rates of public firms in uncertain times.
5 The dynamic relationship between policy uncertainty and investment decisions
This section examines the evolving relation over time between policy uncertainty and fixed asset investment for both public and private firms. The most commonly used method for studying such dynamic relationships is the estimation of vector autoregressions (VARs) with the subsequent computation of impulse response functions (IRFs). IRFs would allow us to assess and visualize the endogenous investment reaction to an initial EPU shock across subsequent points in time. Consequently, we start by estimating VARs based on our standard set of variables: log(EPU), Cash flow, Sales growth, Size, GDP growth, Inflation, and Investment. The VARs are run on annual data from 2009 to 2017, using a lag structure of one. To calculate the IRFs, we apply a Cholesky decomposition using the same variable order outlined above.4
Fig. 2
Effect of a policy uncertainty shock on corporate investment
×
Both IRFs are displayed in Fig. 2. As shown in the top panel, we observe that a one unit shock to EPU has a negative and statistically significant impact on public firms’ fixed asset investment for an approximate duration of two years, reaching its trough 1 year after the shock. This finding is in accordance with real options theory. As (policy-related) uncertainty increases, the urge of firm managers to inform themselves (about government commitments) becomes more valuable than investing immediately in uncertain projects. Thus, the option value of an investment delay exceeds the value of immediate investment. This causes companies to postpone part of their investments. However, when focusing on later periods, we discover a rebound effect that starts towards the end of the third year and lasts almost until the end of the fourth year after the initial shock. This observation is in line with the notion that although uncertainty may trigger investment delays, once it is cleared up, the levels of investment will rise to meet pent-up demand. These results are similar to those of Gulen and Ion (2016), who find that in times of policy-related uncertainty spells, public U.S. firms experience an initial decline in investment for about two and a half years, which is attenuated by a rebound effect in later periods.
The bottom panel of Fig. 2 indicates that, comparable to public firms, a one unit shock to EPU has a negative and significant impact on private firms’ fixed asset investment for about two years. Again, the IRF reaches its trough 1 year after the shock. However, contrary to public firms, we fail to detect a statistically significant rebound in investments at any time in the future. Thus, consistent with real options theory, both public and private firms postpone part of their investments, whereas private firms delay them indefinitely.
It is possible that this cross-sectional variation in recovery responses several years after the initial shock might be connected to the idiosyncratic increase in investment profitability rates in uncertain times. This could encourage public firms to pursue new investments at a higher rate than before. As far as private firms are concerned, their inability to identify the projects that would justify immediate implementation the most potentially leads them to refrain from offsetting the initial decline in later periods. Consequently, no significant rebound in their investments can be observed.
6 Robustness
6.1 Omitted variable bias
In this subsection, we adopt multiple specifications to assess the robustness of our baseline finding of a negative relation between EPU and private firms’ investment decisions. One potential point of concern with our analysis thus far is that the EPU index may track some macroeconomic information pertaining to investment opportunities, despite the efforts of Baker et al. (2016) to ensure that it does not merely reflect macroeconomic conditions. Several authors point out that economic uncertainty is counter-cyclical (e.g., Bloom 2009; Baker et al. 2016). As policy-makers often feel the urge to implement policy changes to combat prevailing poor economic conditions, it appears reasonable to assume that EPU is also counter-cyclical and thus negatively correlates with investment opportunities. Consequently, if our current control variables in the form of GDP growth and Inflation do not fully reflect firms’ overall investment chances, our estimates might suffer from a potential omitted variable bias.
In order to alleviate such concerns, we estimate Eq. (1) for our full sample of 291,415 private firm-years with the addition of two country-level economic indicators that are likely to impact corporate investment decisions. These controls consist of annual GDP per capita (GDP per capita) and annual exports plus imports scaled by GDP (Trade).
The results are reported in Table 6. Regression 1 shows that despite the inclusion of GDP per capita and Trade, the magnitude and significance of the EPU coefficient remain virtually unchanged. Specifically, the coefficient on log(EPU) is \(-\,0.030\) (t-statistic = \(-\,2.737\)). Thus, a 10% increase in EPU stands in association to reduce private firms’ Investment by about 16.667% of the sample mean. Compared to the initial decline of 17.222% reported in regression 1 in Table 3, this provides us with substantiated evidence that our results are not compromised by omitted investment opportunity measures. Similarly, in regression 2, we use the first principal component from a different set of control variables developed to capture investment opportunities. This set of variables follows (Greenland et al. 2019) and comprises the forecasted real GDP growth (Predicted RGDP growth), the Composite Leading Indicator (CLI), the Business Confidence Indicator (BCI), and the Consumer Confidence Indicator (CCI). All variables are obtained from the OECD database. Nevertheless, despite the inclusion of the Economic condition, first PC variable, the coefficient on log(EPU) remains negative and significant.5
Table 6
Full private firm sample: further macroeconomic control variables
Dependent variable: Investment
(1)
(2)
(3)
(4)
Log(EPU)
− 0.030***
− 0.031***
− 0.029***
− 0.029***
(− 2.737)
(− 2.991)
(− 3.128)
(− 2.859)
GDP growth
0.000
0.000
0.000
0.000
(0.336)
(0.005)
(0.239)
(0.425)
Inflation
− 0.002
− 0.001
− 0.002
− 0.002
(− 0.829)
(− 0.778)
(− 0.888)
(− 0.765)
GDP per capita
− 0.000**
− 0.000*
(− 1.993)
(− 1.663)
Trade
0.000
0.000
(0.620)
(0.152)
Economic condition, first PC
0.000
0.000
(0.345)
(0.176)
Return volatility
− 0.520
− 0.468
(− 1.494)
(− 1.512)
CS std. dev. of sales growth
− 0.040***
− 0.038***
(− 2.749)
(− 2.655)
Firm-level control variables
Yes
Yes
Yes
Yes
Firm fixed effects
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Adjusted R\(^{2}\)
0.076
0.076
0.077
0.077
Observations
291,415
291,415
291,415
291,415
Note: This table reports firm-level regression estimates of Eq. (1) for our full private firm sample with additional macroeconomic control variables. The dependent variable is Investment. The independent variable of interest is Log(EPU). It is the natural logarithm of a country’s economic policy uncertainty index. GDP per capita is expressed in constant 2011 international dollars. Trade is exports plus imports scaled by annual GDP. GDP per capita and Trade are retrieved from the World Bank’s World Development Indicators. Economic condition, first PC is the first principle competent of four different OECD measures for economic prospects. Theses OECD measures are described in detail in the Appendix. Return volatility is the time-series volatility of realized monthly returns on the main stock exchange in each country of the previous year and CS std. dev. of sales growth is the cross-sectional standard deviation of sales growth of all firms in our sample in each country of the previous year. The set of firm-level control variables includes Sales growth, Cash flow, and Size. See Tables 1 and 2 for definitions of the other variables. All regressions include year dummies as well as firm fixed effects. Standard errors are clustered on both firm and year. Robust t-statistics are reported in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively
A second potential concern to which the previous literature has already alerted is that the EPU index may simply reflect macroeconomic uncertainty, which is not necessarily tied to future policy decisions (see Gulen and Ion 2016; Greenland et al. 2019). Since events such as recessions, wars, and financial crises do not only increase policy-related uncertainty but also contribute to intensifying overall macroeconomic uncertainty, firms facing policy uncertainty may also experience uncertainty about other elements of their operations (e.g., supply chain issues).
To address this matter, we include two alternative measures of macroeconomic uncertainty. First, we use one of the most frequently applied proxies of general macroeconomic uncertainty as perceived by the stock market (e.g., Bloom 2009; Taglioni and Zavacka 2013). We therefore calculate the time-series volatility of realized monthly returns on the respective main stock exchange in each of our nine European countries (Return volatility). The monthly return data are from the Thomson Reuters Eikon database. Second, in line with Bloom (2009) we also utilize the cross-sectional standard deviation of sales growth for all firms in each country included in our sample (CS std. dev. of sales growth) to measure macroeconomic uncertainty. Regression 3 in Table 6 shows that while the coefficients on Return volatility and CS std. dev. of sales growth have a negative relationship with private firm investment, only CS std. dev. of sales growth is statistically distinguishable from zero. Crucially, these controls have little impact on our point estimate of interest, which remains qualitatively unaffected. Lastly, regression 4 demonstrates that economic policy uncertainty affects private firm investment in a way that is consistent with our model, even when investment opportunities and general macroeconomic uncertainty are included in a single specification.
While the estimates of Table 6 are based on our full sample of 291,415 private firm-years, Table 7 presents estimates based on the private firm observations used in our matched sample. Regarding the primary covariate, the results show no salient discrepancies from the previous table. Hence, Tables 6 and 7 are in line with our baseline finding. The negative relationship between economic policy uncertainty and investment decisions of private firms does not systematically depend on the omission of investment opportunity and macroeconomic uncertainty measures.
Table 7
Matched private firm sample: further macroeconomic control variables
Dependent variable: Investment
(1)
(2)
(3)
(4)
Log(EPU)
− 0.032***
− 0.033***
− 0.029***
− 0.028***
(− 3.341)
(− 3.594)
(− 3.106)
(− 2.996)
GDP growth
0.001
0.002*
0.002**
0.001
(1.114)
(1.793)
(2.287)
(1.103)
Inflation
− 0.003
− 0.002
− 0.002
− 0.003
(− 1.281)
(− 0.866)
(− 0.979)
(− 1.270)
GDP per capita
0.000
0.000*
(1.206)
(1.702)
Trade
0.001*
0.001
(1.691)
(0.910)
Economic condition, first PC
− 0.000
0.001
(− 0.250)
(0.575)
Return volatility
− 0.738
− 0.779
(− 1.286)
(− 1.439)
CS std. dev. of sales growth
− 0.080***
− 0.081***
(− 5.172)
(− 4.966)
Firm-level control variables
Yes
Yes
Yes
Yes
Firm fixed effects
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Adjusted R\(^{2}\)
0.061
0.061
0.063
0.063
Observations
9456
9456
9456
9456
Note: This table reports firm-level regression estimates of Eq. (1) for our matched private firm sample with additional macroeconomic control variables. The dependent variable is Investment. The independent variable of interest is Log(EPU). It is the natural logarithm of a country’s economic policy uncertainty index. GDP per capita is expressed in constant 2011 international dollars. Trade is exports plus imports scaled by annual GDP. GDP per capita and Trade are retrieved from the World Bank’s World Development Indicators. Economic condition, first PC is the first principle competent of four different OECD measures for economic prospects. Theses OECD measures are described in detail in the Appendix. Return volatility is the time-series volatility in monthly returns on the main stock exchange in each country of the previous year and CS std. dev. of sales growth is the cross-sectional standard deviation of sales growth of all firms in our sample in each country of the previous year. The set of firm-level control variables includes Sales growth, Cash flow, and Size. See Tables 1 and 2 for definitions of the other variables. All regressions include year dummies as well as firm fixed effects. Standard errors are clustered on both firm and year. Robust t-statistics are reported in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively
6.2 Instrumental variable analysis
The existing literature on corporate investment has long acknowledged the difficult task to discriminate between uncertainty and bad investment opportunities. Another important objective is to ensure that EPU reflects only uncertainty related to future economic policies and does not capture the uncertainty surrounding future macroeconomic developments in general. As demonstrated above, we deal with this challenge by controlling for these interfering factors using various measures of investment opportunities and overall macroeconomic uncertainty. However, the effectiveness of this strategy hinges critically on the precision and appropriateness of the measures employed. We thus use an instrumental variable (IV) approach to mitigate the risk that the measures may not completely eliminate potential endogeneity problems present in our previous analysis.
We propose one such variable, namely the relative strength of right-wing parties in government (Gov. right). It is based on the “gov_right2” index taken from the Comparative Political Data Set 1960-2017 (CPDS).6 It measures the percentage of parliamentary seats held by right-wing parties in government. Intuitively, this index ranges from 0 to 100, taking a value of 0 if no right-wing party is represented in parliament and 100 if all parliamentary seats are held by right-wing parties.
We expect to find a positive link between Gov. right and policy-related uncertainty. The rationale for this positive relationship is straightforward: In recent decades, the rising popularity of right-wing populist parties has become a central issue in the European political environment (Trumm 2018). In some countries, they were already part of the government or strongly associated with it, such as the Lega Nord (LN) in Italy or the Partij voor de Vrijheid (PVV) in the Netherlands. With their anti-establishment and pro-national stance, these parties have created much instability in the political system and in the business world (de Sousa et al. 2020). Even if these parties are not directly part of the government, they influence the positions of the established parties (Muis and Immerzeel 2017). The established parties feel they have to adapt some of the positions of the right-wing populist parties in order to win back their voters. For example, one of the greatest successes of European right-wing populist parties was in 2015, when UKIP forced the ruling Tories to hold the Brexit referendum in 2016, leading to dramatic economic policy uncertainty across the European continent (Bale 2018).
Thus, we expect that a higher percentage of right-wing parties in government increases the impact of populism on the political system, which should lead to greater policy-related uncertainty. Therefore, our measure should meet the relevance condition as an instrument. In contrast, it is not apparent how the composition of the government affects business investments, except through its influence on economic policy uncertainty. Correspondingly, we are certain that our proposed instrumental variable should also meet the exclusion restriction.
In Table 8, we re-run our baseline regression for the full and matched sample of private firms using the index as outlined above as an instrument for economic policy uncertainty. The results of the first-stage regression show that the Gov. right variable is a significant and strong determinant of the natural logarithm of EPU. In particular, the coefficient on log(1 + Gov. right) is 0.111 (t-statistic = 3.765) for the full private firm sample. The F-statistic is 14.18, indicating that we are not suffering from a weak instrument problem. The results imply that a doubling of the relative power position of right-wings parties in government is associated with a 11.1% increase in policy-related uncertainty.
Table 8
Instrumental variables analysis
Dependent variable:
Investment
Log(EPU)
Investment
Log(EPU)
Full private firm sample
Matched private firm sample
(1)
(2)
(3)
(4)
Log(EPU)
− 0.035**
− 0.039*
(− 2.280)
(− 1.779)
Log(1 + Gov. right)
0.111***
0.080***
(3.765)
(3.265)
Sales growth
0.014***
− 0.006***
0.008**
− 0.003
(17.511)
(− 2.748)
(2.045)
(− 0.934)
Cash flow
0.134***
− 0.054**
0.112***
− 0.016
(22.809)
(− 2.373)
(5.278)
(− 0.601)
Size
0.085***
− 0.019
0.085***
− 0.017
(7.748)
(− 1.286)
(7.747)
(− 1.502)
GDP growth
− 0.000
− 0.068***
0.002
− 0.049**
(− 0.087)
(− 2.879)
(1.047)
(− 2.054)
Inflation
− 0.001
0.183***
− 0.001
0.152***
(− 0.709)
(3.450)
(− 0.331)
(2.737)
Firm fixed effects
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Adjusted R\(^{2}\)
0.076
0.745
0.061
0.761
Observations
291,415
291,415
9,456
9,456
In Regression 1, this table replicates our results from Table 1 for the full and matched private firm samples using a two-stage least squares approach with Log(1 + Gov. right) as an instrument for Log(EPU). Log(EPU) is the natural logarithm of a country’s economic policy uncertainty index. The Gov. right variable quantifies the relative dominance of right-wing parties in government based on their seat share in parliament. It is obtained from the Comparative Political Data Set 1960–2017 (CPDS). See Tables 1 and 2 for definitions of the other variables. Regression 2 reports the results of the first-stage regression. All regressions include year dummies as well as firm-fixed effects. Standard errors are clustered on both firm and year. Robust t-statistics are reported in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively
The second-stage regression estimates show that EPU, indeed, has an exogenous impact on investments of private firms that is not attributable to other distorting factors. The impact of EPU is still negative and significant. Therefore, we are confident that our previously presented results are not significantly compromised by endogeneity issues, but are in fact linked to the exogenous impact of the uncertainty surrounding future economic policy decisions.
6.3 Alternative matching procedure
Thus far, our matching procedure is comparable to that of Mortal and Reisel (2013), Asker et al. (2015), and Gilje and Taillard (2016). Specifically, based on a caliper of 0.05, each public firm is matched with a private firm that is most similar in terms of total assets in the same industry, country, and year. While this drastically reduces the difference in average firm size between these two types of firms, it unfortunately leads to an increase in heterogeneity in sales growth (see Table 1).
To mitigate the possibility that our results are in any way distorted by this circumstance, we create a new matched sample using an alternative matching procedure. It uses nearest neighbor matching in each year based on Sales growth, Cash flow, and Size in the same industry and country. Table 14 in the Appendix shows that in this newly created sample, the differences between public and private firms with respect to all three variables are smaller compared to our unmatched sample.
Regression 1 in Table 9 reports estimation results of Eq. (1). The coefficient on log(EPU) is − 0.055 (t-statistic = − 4.177), indicating that an increase in economic policy uncertainty continues to be significantly associated with depressing the investment behavior of public and private European firms. Compared with the coefficient on log(EPU) in regression 5 in Table 3 of − 0.051, this result suggests that the effect of EPU is almost independent of the matching procedure applied.
Table 9
Investment regressions based on alternatively matched sample
Dependent variable: Investment
(1)
(2)
Log(EPU)
− 0.055***
− 0.045***
(− 4.177)
(− 3.247)
Log(EPU) \(\times\) Public listed
− 0.016*
(− 1.718)
Public listed
0.129**
(2.180)
Sales growth
0.029***
0.029***
(9.836)
(9.849)
Cash flow
0.307***
0.307***
(11.104)
(11.106)
Size
0.110***
0.110***
(6.839)
(6.889)
GDP growth
− 0.002
− 0.002
(− 1.448)
(− 1.478)
Inflation
0.001
0.001
(0.219)
(0.248)
Firm fixed effects
Yes
Yes
Year dummies
Yes
Yes
Adjusted R\(^{2}\)
0.136
0.137
Observations
22,232
22,232
This table reports firm-level regression estimates of Eq. (1) based on an alternatively matched sample (controls = Sales growth, Cash flow, and Firm size; exact match = year, industry, and country; nearest neighbor matching). The dependent variable is Investment. The independent variables of interest are Log(EPU) and Public listed. Log(EPU) is the natural logarithm of a country’s economic policy uncertainty index. Public listed is an indicator variable that takes the value of 1 for each year the firm is identified as being listed on a stock exchange and 0 otherwise. See Tables 1 and 2 for further variable definitions and see Table 14 in the Appendix for descriptive statistics of the match sample. All regressions include year dummies as well as firm fixed effects. Standard errors are clustered on both firm and year. Robust t-statistics are reported in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively
To examine whether the two types of firms continue to differ significantly in their investment responses to policy uncertainty, a log(EPU)\(\times\)Public listed interaction term is introduced in regression 2. Similar to our baseline results, the coefficient on the interaction term is negative and statistically significant. It implies that public firms reduce their investments more than private firms to an increase in EPU. Therefore, we are confident that our baseline results are not simply attributable to the adoption of the matching approach of Mortal and Reisel (2013), Asker et al. (2015), and Gilje and Taillard (2016).
7 Conclusion
This paper studies whether Baker et al.’s (2016) economic policy uncertainty index can depress investment decisions of private firms and whether their reaction differs significantly from that of public ones. Our analyses are conducted in a large sample of public and private firms from nine European countries between 2009 and 2017.
The results show that private firms invest significantly less in times of uncertainty. This finding is robust in different specifications aimed at eliminating potential endogeneity concerns. Further, we observe that public firms reduce their investments by about 50% more than private firms in response to an increase in uncertainty. We attribute these results to the greater inclination of public firm management to avoid scrutiny by their shareholders, which fosters a larger degree of uncertainty averse decisions. When examining how the relationship between policy uncertainty and investment varies at different points in time, we find that unlike private firms, public firms experience a rebound effect in their investment levels. This variation in recovery responses might be connected to our observation that only public firms invest more efficiently when confronted with increases in policy-related uncertainty. This could encourage public firms to pursue investments at a higher rate than before whereas private firms refrain from offsetting their initial decline in later periods.
In terms of policy implications, our results suggest that if private companies do not always invest in the most efficient manner, they may benefit from adopting the decision-making processes of public companies in uncertain times, as this could help them identify the projects that justify immediate implementation the most. Therefore, when faced with uncertainty, private company decision-makers should proactively gather and consider feedback from their peers on future investments.
We see an avenue for future research to explore other channels that could help further explain the different investment responses of public and private firms to uncertainty. For example, Michaely and Roberts (2012) argue that public firms face greater public scrutiny than private firms because analysts only track and monitor listed companies.
Declarations
Conflict of interest
None of the authors have any conflict of interest to declare.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix for “Policy uncertainty and corporate investment: public versus private firms”
Description of variables
Variable
Type
Description
I. Country-level legal variables
Anti-self
Cross-section
This variable stands for the anti-self-dealing index constructed by Djankov et al. (2008) and measures the strength of minority shareholder protection against self-dealing. It includes both the ex ante strength of law that hinders self-dealing and the ex post enforcement of this law, once a self-dealing transaction occurred
Anti-dir
Cross-section
The anti-director rights index is constructed by La Porta et al. (1998). It measures the extent to which minority shareholders are protected in business decision-making processes. The index is constructed by adding up indicator variables for six sub-categories. A sub-category receives a value of one when: “(1) the country allows shareholders to mail their proxy vote to the firm, (2) shareholders are not required to deposit their shares prior to the general shareholders’ meeting, (3) cumulative voting or proportional representation of minorities in the board of directors is allowed, (4) an oppressed minorities mechanism is in place, (5) the minimum percentage of share capital that entitles a share-holder to call for an extraordinary shareholders’ meeting is less than or equal to 10 percent (the sample median), or (6) shareholders have pre-emptive rights that can be waived only by a shareholders’ vote” (La Porta et al. 1998, p. 1123)
Disclosure
Cross-section
The index of disclosure, as established by La Porta et al. (2006), measures the disclosure requirements of public firms in a given country. It is the arithmetic mean of six sub-categories, all of which are designed as indicator variables with being equal to one if certain requirements are meet and zero otherwise. The sub-categories measure whether a country requires a firm to deliver a prospectus to its investors when new securities are issued and whether this prospectus needs to include information on the directors‘ and officers‘ compensations, shareholder structure, insider ownership, irregular contracts, and any business dealings between the issuer and its officers and directors
Liability
Cross-section
The public firm liability standard index is created by La Porta et al. (2006) and is equal to the arithmetic mean of three sub-categories. The categories measure the liability standards of the issuing firm, the distributors, and the accountants in case that losses occur to the investor due to misleading information in the prospectus
II. Macroeconomic control variables
Predicted RGDP growth
Cross-section and annual time-series
This variable resembles the forecasted real GDP growth from the OECD database. The OECD’s forecast relies on expert- and model-based evaluations of the economic environment in specific countries and the world economy
CLI
Cross-section and annual time-series
This index stands for the Composite Leading Indicator and is developed by the OECD. It consists of multiple components, which reflect the business cycle. Components considered differ between countries according to the availability of data. For instance, the CLI for the United States includes, among several other things, the number of new homes started, the number of weekly hours worked in manufacturing, and stock prices (NYSE composite)
BCI
Cross-section and annual time-series
This index stands for the Business Confidence Indicator. It is based on the OECD’s Business Tendency Surveys and measures managers’ expectations on the economic development by calculating the average score on questions about future trends in production, finished goods inventory, and order backlogs
CCI
Cross-section and annual time-series
This index stands for the Consumer Confidence Indicator. It is based on the OECD’s Consumer Opinion Surveys and measures how consumers assess the economic development. The index averages the scores on questions that compare the economic situation of the recent past with expectations on the immediate future
Economic condition, first PC
Cross-section and annual time-series
This variable represents the first principle component of the four above-mentioned macroeconomic control variables
Note: This table presents means and standard deviations of firms’ total assets (in millions of euros) for the basic matched sample and for various variations of our initial matching specification. A description of the construction of the sample and details on the basic matching procedure are given in Sect. 2. In rows 1, 2, and 3, the matched samples are based on three-digit SIC industry sectors. The matched sample in row 4 is based on two-digit SIC industry sectors and in row 5 on four-digit NAICS industry sectors. For comparing purposes, row 1 is a repetition of our basic matched sample that is based on a caliper of 0.05. Row 2 applies a caliper of 0.01 and row 3 a caliper of 0.09. Rows 4 and 5 use a caliper of 0.05. The final column presents pairwise differences in means of firms’ total assets
Table 11
Investment regressions with different matching approaches
Dependent variable: Investment
SIC3 & cal.=0.05
SIC3 & cal.=0.01
SIC3 & cal.=0.09
SIC2 & cal.=0.05
NAICS4 & cal.=0.05
(1)
(2)
(3)
(4)
(5)
Log(EPU)
− 0.033***
− 0.033*
− 0.027*
− 0.038**
− 0.042**
(− 2.666)
(− 1.837)
(− 1.902)
(− 2.319)
(− 2.256)
Log(EPU) \(\times\) Public listed
− 0.034***
− 0.029***
− 0.037***
− 0.030***
− 0.021**
(− 4.595)
(− 3.520)
(− 4.886)
(− 2.979)
(− 1.977)
Public listed
0.281***
0.005
0.238***
0.200***
0.174**
(4.449)
(0.039)
(3.922)
(3.124)
(2.392)
Sales growth
0.026***
0.031***
0.027***
0.026***
0.027***
(9.027)
(9.406)
(7.851)
(7.009)
(7.183)
Cash flow
0.200***
0.219***
0.226***
0.231***
0.185***
(9.672)
(7.646)
(10.865)
(10.612)
(6.703)
Size
0.105***
0.105***
0.106***
0.099***
0.111***
(6.845)
(7.397)
(7.082)
(6.337)
(6.362)
GDP growth
− 0.001
− 0.002
− 0.001
− 0.000
− 0.000
(− 0.689)
(− 1.103)
(− 1.048)
(− 0.142)
(− 0.142)
Inflation
0.000
0.001
− 0.000
− 0.001
− 0.002
(0.084)
(0.332)
(− 0.203)
(− 0.421)
(− 0.820)
Firm fixed effects
Yes
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Yes
Adjusted R\(^{2}\)
0.115
0.13
0.123
0.117
0.125
Observations
18,912
16,412
19,536
20,298
18,550
Note: This table reports firm-level regression estimates that capture cross-sectional differences between public and private firms for the basic matched sample and for various variations of our initial matching specification. The dependent variable is Investment. The independent variables of interest are Log(EPU) and Public listed. Log(EPU) is the natural logarithm of a country’s economic policy uncertainty index. Public listed is an indicator variable that takes the value of 1 for each year the firm is identified as being listed on a stock exchange and 0 otherwise. See Tables 1 and 2 for variable definitions and Table 11 for the composition of the samples. All regressions include year dummies as well as firm fixed effects. Standard errors are clustered on both firm and year. Robust t-statistics are reported in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively
Table 12
Full private firm sample: controlling for economic conditions
Dependent variable: Investment
(1)
(2)
(3)
(4)
(5)
(6)
Log(EPU)
− 0.031***
− 0.031***
− 0.031***
− 0.031***
− 0.031***
− 0.031***
(− 2.991)
(− 3.052)
(− 3.052)
(− 3.052)
(− 2.991)
(− 2.991)
Predicted RGDP growth
0.000
0.000
(0.345)
(0.345)
CLI
0.149***
0.029
(6.332)
(1.411)
BCI
− 0.031***
− 0.016
(− 6.332)
(− 0.855)
CCI
− 0.020***
− 0.006
(− 6.332)
(− 0.457)
Economic condition, first PC
0.000
(0.345)
Control variables
Yes
Yes
Yes
Yes
Yes
Yes
Firm fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Adjusted R\(^{2}\)
0.076
0.076
0.076
0.076
0.076
0.076
Observations
291,415
291,415
291,415
291,415
291,415
291,415
Note: This table reports firm-level regression estimates of Eq. (1) for our full private firm sample with additional economic condition controls, as described in the Appendix. The dependent variable is Investment. The independent variable of interest is Log(EPU). It is the natural logarithm of a country’s economic policy uncertainty index. The set of control variables includes Sales growth, Cash flow, Size, GDP growth, and Inflation. See Tables 1 and 2 for definitions of these variables. All regressions include year dummies as well as firm fixed effects. Standard errors are clustered on both firm and year. Robust t-statistics are reported in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively
Table 13
Matched private firm sample: controlling for economic conditions
Dependent variable: Investment
(1)
(2)
(3)
(4)
(5)
(6)
Log(EPU)
− 0.033***
− 0.033***
− 0.033***
− 0.033***
− 0.033***
− 0.033***
(− 3.594)
(− 3.556)
(− 3.556)
(− 3.556)
(− 3.594)
(− 3.594)
Predicted RGDP growth
− 0.000
− 0.000
(− 0.250)
(− 0.250)
CLI
0.106***
0.006
(4.108)
(0.303)
BCI
− 0.022***
0.009
(− 4.108)
(0.498)
CCI
− 0.014***
− 0.019
(− 4.108)
(− 1.600)
Economic condition, first PC
− 0.000
(− 0.250)
Control variables
Yes
Yes
Yes
Yes
Yes
Yes
Firm fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Adjusted R\(^{2}\)
0.061
0.061
0.061
0.061
0.061
0.061
Observations
9,456
9,456
9,456
9,456
9,456
9,456
This table reports firm-level regression estimates of Eq. (1) for our matched private firm sample with additional economic condition controls, as described in the Appendix. The dependent variable is Investment. The independent variable of interest is Log(EPU). It is the natural logarithm of a country’s economic policy uncertainty index. The set of control variables includes Sales growth, Cash flow, Size, GDP growth, and Inflation. See Tables 1 and 2 for definitions of these variables. All regressions include year dummies as well as firm fixed effects. Standard errors are clustered on both firm and year. Robust t-statistics are reported in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively
Table 14
Descriptive statistics for alternatively matched sample
Full sample
Matched sample
Public
Private
Difference
Public
Private
Difference
Investment
Mean
0.035
0.018
0.017
0.039
0.024
0.015
SD
0.151
0.116
0.156
0.136
Sales growth
Mean
0.164
0.140
0.024
0.174
0.196
− 0.023
SD
0.775
0.674
0.776
0.897
Cash flow
Mean
0.066
0.084
− 0.018
0.073
0.078
− 0.005
SD
0.121
0.108
0.121
0.114
Size
Mean
18.957
17.479
1.478
19.077
18.574
0.503
SD
1.659
1.076
1.659
1.496
No. of observations
13,128
291,415
11,116
11,116
No. of firms
2,301
56,802
2,016
5,626
Note: This table contains summary statistics for our full sample and alternatively matched sample that is used in Table 9. A description of the construction of the new sample and details on the matching procedure are given in Sect. 6. Samples are divided into calendar years; financial years ending between 1 January and 31 May are counted towards ending in the preceding calendar year. The table presents means and standard deviations of the main firm-level variables used in our analysis as well as pairwise differences in means. All firm-level data are obtained from the Amadeus Database by Bureau van Dijk. See Tables 1 and 2 for further variable definitions
This figure is the weighted average of the mean Investment values of public and private companies: \((13{,}128 \times 0.035 + 291{,}415 \times 0.018)/(13{,}128 + 291{,}415) = 0.019\).
This inference is based on the (lenient) assumption that private firms do not invest in the most efficient way at all times, i.e. have ex ante potential for improvement.
Our results remain qualitatively similar when we position the EPU index as the last variable in the ordering, or when we include election year dummies.
In Tables 12 and 13 in the Appendix, private firm Investment is regressed on each variable used to construct the Economic condition, first PC measure. Again, we find that the inclusion of these variables does not substantially alter the size or significance of our primary covariate.
Armingeon, Klaus, Virginia Wenger, Fiona Wiedemeier, Christian Isler, Laura Knöpfel, David Weisstanner and Sarah Engler (2019). Comparative Political Data Set 1960–2017. Institute of Political Science, University of Zürich.
