Readjusting the speed of leverage adjustment during the COVID-19 pandemic?

1.1 Background and predictions

Before the onset of the pandemic, the US economy was on a major growth spurt. In such circumstances, analysts and economists may begin to wonder what could possibly turn things around. Without any possible thought about COVID-19, the only likely trigger for a recession would have been the growth in corporate debt [4]. Corporate debt reached a record high in the early months of 2020. However, the advent of the pandemic changed things around. We saw the country’s growth shrinking to 3.5% while corporate debt began to fall (see Figure 1). The change in corporate debt during the pandemic is likely to impact the SOA due to the uncertainty borne by the pandemic.

Several factors related to the COVID-19 pandemic have impacted the ideal level of leverage for firms. For instance, the pandemic-induced strains on corporate cash flows and the tightening of global liquidity constraints have influenced the market valuations of firms as the pandemic spread. Firms with higher cash reserves, lower debt and greater profits have shown more resilience to the pandemic than those with fewer resources (Haque & Varghese, 2021). Furthermore, the value-maximizing level of leverage for firms, particularly those with higher exposure to business risk, may have declined due to reduced growth prospects or increased risks from the COVID-19 shock. Business owners may increase collection efforts in anticipation of defaults and changes to short-term payment plans, which may result in an increase in non-compete litigation, particularly from startups, as laid-off professionals seek new employment opportunities.

The COVID-19 shock's reduction in the book value of equity has immediate consequences on firms' leverage ratios, which increase by approximately 6.7 to 8% points compared to a typical business scenario (Demmou et al., 2021). Additionally, firms that were severely impacted by the drop in demand due to social distancing, particularly those that did not de-lever based on risks from lockdowns, now have actual leverage ratios that significantly exceed their ideal level of leverage (Haque & Varghese, 2021).

Perhaps, the SOA toward a target ratio in the presence of the pandemic might have been driven by many factors, some of which are related to government interventions in the form of both fiscal and monetary stimulus to address the havoc induced by the pandemic. Tarkom (2021) shows that firms that receive government incentives in the form of investment tax credits and deferred taxes are efficient in managing their working capital. This implies that access to credit or liquidity may influence the rate and direction of adjustment toward the target ratio among US firms during the pandemic. Ho et al. (2021) confirm this from an international perspective.

Another important conduit through which the pandemic could alter the SOA is risk avoidance. Prior evidence (Cook & Tang, 2010; Hackbarth, Miao, & Morellec, 2006) suggests that both macroeconomic and firm characteristics are among other determining factors of adjustment costs. Thus, the rising uncertainty, either in the form of liquidity and/or default risk (Ho et al., 2021; Luo, 2021; de Vito & Gómez, 2020), may change the rate of convergence. Additionally, large deviations from the target ratio may expose firms to additional risks. As a result, with the many policies (e.g. social distancing and lockdowns) implemented to curb the spread of the pandemic which induces operational risks (Luo, 2021), drifting too far from target leverage may introduce costly adjustments.

Therefore, firms may be incentivized to adjust toward their target levels to avoid costly adjustments. It is expected that if the motive for convergence rate is due to risk avoidance, then low market valued (risky) firms, low cash holding (constrained) firms, and firms situated in states most affected by the pandemic will have a faster convergence rate than high valued firms, firms with excess cash and firms in the least affect states. The study finds support for these arguments. In line with the idea that practitioners pay much attention to book leverage and that they do not change their capital structure in response to stock price movement (Yin & Ritter, 2020), the convergence rate toward market leverage should be slower compared to that widely reported in the literature. Yin and Ritter argued that the estimated SOA is influenced by both passive (not related to financing choices) and active components (influenced by financing choices).

The authors argue that the upward bias and sensitivity of the estimated SOA towards the market target to the growth in market value drives the higher estimated SOA towards the market leverage target. Therefore, by separating high-valued firms from low-valued firms, differences in the estimated SOA toward market leverage should be expected. This is particularly true since there has been a sharp increase in the market value of firms during the pandemic. Hence, it is expected that the rate of convergence toward the market leverage target will be different for high-valued versus low-valued firms. Therefore, since high-valued firms and firms with excess cash are expected to adjust slowly toward their leverage target, this class of firms may drive the convergence rate for the market target down (Yin & Ritter, 2020). This hypothesis is also supported in this study.

Also, we predict that firms headquartered in states most affected by the pandemic may adjust faster to avoid additional risks stemming from the pandemic. The rationale for this focus is on the basis that fundamental business activities take place in close proximity to the headquarters (Chaney, Sraer, & Thesmar, 2012; Pham, Adrian, Garg, Phang, & Truong, 2021; Pirinsky & Wang, 2006). Thus, if the location of the headquartered firm is exposed to more risk, SOA will be faster to avoid any accumulating risks.

The full sample analysis shows an adjustment speed toward the book (market) leverage target to be 5.2% (9.2%) per quarter. These translate into an annual SOA of 20.8% (36.8%) which are consistent with prior findings. A subsample analysis indicates that the estimated SOA towards book target leverage during the pandemic is higher relative to the pre-pandemic by several percentage points. For instance, the quarterly SOA adjustments toward the book leverage target increased from 3.7% to 7.2% per quarter. However, the estimated SOA toward the market leverage target decreased during the pandemic relative to the pre-pandemic. For instance, the quarterly SOA adjustments toward the market leverage target decreased from a whopping 16.8% to 12.1%. This result supports the findings of Yin and Ritter (2020) who put it simply that “the estimated SOA towards market leverage is slower than you think” but counters the findings from international settings reported by Vo et al. (2021).

Second, the evidence shows that the observed low SOA towards the market leverage target is driven by the growth in the market value of firms (Yin & Ritter, 2020). For instance, for low-valued firms, there is a strikingly high SOA toward market leverage target from the pre-pandemic to the pandemic: 24.1% versus 29.9% per quarter which support the risk avoidance motive. However, for highly valued firms, not only did the SOA not matter before the pandemic, but the estimated SOA towards market leverage target is also only 3%. This can explain why the full sample SOA toward market leverage target is lower during the pandemic relative to the pre-pandemic case. The observed differences in market leverage targets can be attributed to the sensitivity of market leverage to the rise in market valuation during the pandemic.

Additionally, the evidence shows that (1) over-levered firms converge faster during the pandemic (2) low-valued firms relative to high-valued firms with excess cash adjusted faster during the pandemic (3) firms situated in states most affected by the pandemic adjusted faster. Lastly, we show that although the emergence of the pandemic caused a substantial deviation from the target ratio, this deviation is not only attributed to the pandemic but also the overall firm-level risk faced. A mediation analysis subsequently revealed that firm-level risk partially mediated how far the pandemic caused a deviation from target levels.

This paper makes an important contribution to the existing literature on the emerging discourse of COVID-19, corporate finance and capital structure (e.g. Faulkender et al., 2012; Flannery & Rangan, 2006; Krieger, Mauck, & Pruitt, 2021; Öztekin & Flannery, 2012; Tarkom, 2021; Vo et al., 2021) by documenting the (risk avoidance) deferential attitude of US firms towards leverage target SOA. The study extends prior literature’s determinant of SOA and suggests that excess cash determines how fast firms adjust towards target ratios in times of pandemic. The study further adds that not only are these changes in leverage SOA attributed to the pandemic but also the firm’s overall risk faced during the pandemic. The analysis offers a cautionary tale to the generalization of the SOA since the convergence rate is dependent on the growth in the market value of firms and the risk appetite of firms, suggesting that risk avoidance is core to the convergence rate during the pandemic.

2.1 Baseline leverage model

This paper aims to examine changes in the SOA as a result of the emergence of COVID-19 and offer possible reasons for such changes, as such, we employ a two-step approach since it offers flexibility and allows for the control of both firm and industry characteristics (An, Li, & Yu, 2015). The first step regresses (book and market) leverage on the determinants of leverage to estimate the (unobservable) target leverage. The target leverage from the first-step is used in the second-step to estimate a partial adjustment model. This approach is considered appropriate for our empirical setting since it will enable us to test the changes in SOA borne by the pandemic and the factors that caused a deviation from the target ratio.

The standard partial-adjustment leverage model often employed in the literature (e.g. Çolak, Gungoraydinoglu, & Öztekin, 2018; Faulkender et al., 2012; Flannery & Rangan, 2006; Lemmon, Roberts, & Zender, 2008; Luo, 2021; Öztekin & Flannery, 2012) is of the form

Rearranging Eq. (1) yields

The target ratio is unobservable; hence, it is estimated using a partial adjustment model following the restriction in Eq. (3)

3.1 Descriptive statistics

Table 1 reports the descriptive statistics of all the variables. The statistics show that the average book and market leverage is lower during the pandemic. Also, the target ratio for book and market leverage increased during the pandemic relative to the pre-pandemic. Similarly, as the average deviation from the target ratio for book leverage is higher during the pandemic, it marginally increased for the market leverage during the pandemic [6]. Test of means for the Pre-COVID and COVID statistics show that the (key) variables are different from each other.

3.2 Empirical findings

Table 2 Panel A reports the baseline regression using the specification in Eq. (1). For the full sample case, the SOA toward book leverage (Book Lev) is 5.2% per quarter (column 1), while the SOA towards market leverage (Market Lev) is 9.2% per quarter (column 4). These numbers translate into an annual SOA towards Book Lev and Market Lev of 20.8% and 36.8%, respectively. This finding is consistent with Faulkender et al. (2012) and Öztekin and Flannery (2012) who report an annual SOA towards Book Lev to be 21%. The large convergence rate for Market Lev reported in this study is also consistent with the findings by Flannery and Rangan (2006) who report an annual convergence rate greater than 30%.

For the subsample analysis, before and within the pandemic case, the SOA towards Book Lev (Market Lev) is higher (lower) for before and within the pandemic, respectively. That is, the SOA towards Book Lev increased from 3.7% to 7.2% per quarter (columns 2-3). This also translates into an annual convergence rate of 14.8% and 28.8%, respectively. However, the SOA towards Market Lev decreased from 16.8% to 12.1% per quarter (columns 5-6), translating into a whopping annual convergence rate of 67.2% and 48.4%, respectively [7]. The results suggest that the SOA towards book leverage and market leverage is significantly different, particularly in times of crisis. From Figure 2, the higher heterogeneity observed in market leverage changes compared to book leverage explains twice as much SOA towards market leverage.

While the increase in the SOA toward the book target is close and consistent with the findings by Vo et al. (2021), the results on adjustments toward the market target contradict their findings. This finding suggests that the attitude of US firms toward market leverage SOA in times of crisis is different from the rest of the world, as found by Vo et al. (2021). The results reveal that while the SOA toward the book target continues to increase from pre-pandemic to within-pandemic, the SOA towards the market target decreases. The finding is consistent with Yin and Ritter (2020) who show that the estimated SOA towards the market target is lower than the SOA towards the book target. With practitioners focusing on book leverage and that firms do not change their debt in response to stock price changes (Yin & Ritter, 2020), the change in the SOA toward market target found in this study and prior literature remains a puzzle. This is because, from Figure 3, the average market value of equity [8] increased sharply from lower levels pre-pandemic to higher levels within-pandemic. This is also associated with the sharp decline in market leverage.

Panel B report evidence using the baseline regression without zero-leveraged firms. Motivated by Strebulaev and Yang (2013) and Choi and Park (2022), we eliminated firms with zero debt in current liabilities and long-term debt and re-ran our baseline regression. This analysis rules out the effect of firms with zero leverage driving the results. The findings are quantitatively similar to that reported in Panel A except for the SOA for market leverage for the pre-pandemic case. This analysis shows that the results reported in Panel A are not driven by non-zero leveraged holding firms.

Panel C and D report findings by (1) eliminating healthcare firms and (2) examining distressed firms, respectively. This is to show that our results are robust when we exclude healthcare firms and also consider the behavior of distressed firms. Particularly, since healthcare firms received substantial federal government support during the pandemic, their attitude towards leverage adjustment will differ from other firms. Similar arguments hold for distressed firms due to the costly nature of the adjustment. We used dividend payout (Bhagat, Moyen, & Suh, 2005) as a proxy of distress due to the difficulty in paying dividends during the pandemic. Hence, firms that are unable to pay a dividend in a given quarter are considered distressed. Similar to before, we find that excluding healthcare firms from our sample did not alter the basic findings of this study. We find similar evidence when we also exclude firms in the airline industry (no reported). However, we notice that distressed firms strikingly adjusted their SOA faster during the pandemic which is consistent with the risk avoidance motive and our predictions.

To ascertain whether these effects vary for firms that receive government support or not, we conduct an additional test by examining the effect of government intervention on the SOA. First, we use investment tax credit (Tarkom, 2021) from the government to moderate the relationship between Book Dev and Market Dev on ΔBook Lev and ΔMarket Lev, respectively. We split the sample based on whether in a given quarter, firms received any government support in the form of investment credit or not. Tarkom (2021) shows that investment tax credit is critical for firms in managing effective working capital during the pandemic. Second, we specifically and randomly identified some firms that received government support during the pandemic. We then modeled the SOA for the identified firms (15 firms were selected, see the notes in Table A2). Our analysis (reported in Table A2 in the Appendix) shows that indeed for firms that received government support during the pandemic, the SOA is faster than those that did not receive any support, offering support to our argument that firms’ SOA changes in the face of government intervention.

We will provide details on how the growth in the market value of the firm is sensitive to the estimated SOA towards market leverage in the next section.

3.3 Growth in the market value of firms and the speed of adjustment

The results thus far in conjunction with Figures 1 and 3 show that the SOA towards market leverage is sensitive to the market value of equity which arguably distinguishes book leverage from market leverage. However, to explain this sensitivity of SOA to the market value of equity, important questions regarding the kind of firms that have been driving the leverage binge before and within the pandemic need to be explored. To do this, we created five portfolios based on the market value of firms. That is, we sort all stocks into five quintiles based on the market value of equity. Market value of equity is defined as the stock price times the number of shares outstanding. The first group top 20 consist of the first 20% most valued firms while the low 20 refers to the bottom 20% less valued firms. In between these two groups are three different groups following the same logic.

Table 3 reports the statistics for the leverage and asset position of each group. It is evident that while the market value of equity increased during the pandemic period, the low 20 group firms reduced their book and market leverage while the top 20 group firms had a marginal increase in book leverage but a reduction in market leverage. Additionally, it is interesting to note that while all other groups reduced both debts in current liabilities and long-term debt, firms in the top 20 group only reduced debt in current liabilities but increased debt in long-term debt, possibly to invest in technology to support remote work. It is thus not surprising that the top 20 firms hold more than 40% of both short-term and long-term debt. It is therefore expected that the SOA will be different across these groups.

Next, to examine how sensitive the SOA is to the growth in market value, we re-run the baseline model for each group to examine the variation in the SOA. The results are presented in different panels in Table 4. Panel A reports the findings for the low 20 group of firms. While the SOA for book leverage before the pandemic (3.6% per quarter) is similar to the baseline results reported in Tables 2 and it increased (4.2%) within the pandemic period even though the increase is smaller than that reported in Table 2. A more interesting finding is the SOA towards the market target. We find a high SOA during the pandemic for market leverage than for the pre-pandemic case: 29.9% versus 24.1% per quarter. The conversion rates for this group of firms are much higher compared to the baseline results in Table 2.

Similar to Panel A, Panel B report the same analysis but for the case of the next group of firms. Strikingly, the SOA for book leverage increased from 5% to 12.1% during the pre-pandemic to the pandemic, respectively. Similar to the findings for the Rank 1 firms, we find that the SOA during the pandemic period is marginally higher (27.8%) than in the pre-pandemic case (27.2%). The findings indicate that this class of firms tends to adjust faster towards both book and market leverage during the pandemic possibly due to risk avoidance motive.

Panel C reports the findings of the middle class of firms under this analysis. Similar to the results in Panel A and B, the SOA for both book and market leverage is higher during the pandemic relative to the pre-pandemic case. That is the SOA toward the book target increased from 5.5% to 6.8% while the SOA towards the market target is estimated to increase from 15.1% to 18.8%, also suggesting a faster convergence for medium-ranked market-valued firms.

The findings reported in Panel D show the estimated SOA for the group of firms next to the most valued firms. The results are different from the reported findings in Panels A, B and C. The estimated SOA towards book target increased from the pre-pandemic to the pandemic period. However, the case for the market leverage reverted for the estimated SOA during the pandemic period. For instance, while the previous three panels reported that the estimated SOA is higher during the pandemic relative to before the pandemic, we find that in Panel D, the estimated SOA during the pandemic is lower relative to before the pandemic. These findings are consistent with the baseline results reported in Table 2. The findings suggest that the SOA for high-valued firms is slower than less valued firms, supporting the idea that this group of firms has the capacity to pay and thus react differently toward the speed of convergence (Ho et al., 2021).

Finally, the results for the most valued group of firms in this sample are reported in Panel E. Interestingly, the results show that the SOA during the pre- and within-pandemic cases for both book and market leverage does not matter. However, the findings indicate that during the pandemic, the SOA toward the book target is higher (8.2%) relative to the SOA for market leverage (3%). The results for this class of firms are interesting and show that for highly valued firms, the SOA matters only in times of crisis. The results also deviate from the extant literature where the estimated SOA for market leverage has consistently been higher than book leverage and support the findings by Yin and Ritter (2020).

The analysis shows how sensitive SOA is to the growth in the market value of equity such that firms that are less valued (risky firms) adjust toward their leverage target much faster than firms that are highly valued (less risky firms). The analysis also offers some explanation as to why the baseline results show that although there is an increasing SOA for both book and market leverage, the SOA for the market increases at a decreasing rate during the pandemic. The results offer insight to the point that less valued firms adjust faster toward their leverage target, especially in times of crisis to prevent drifting too far off their target, since larger deviations expose the firm to more risks and costly adjustments. Thus, the decline in SOA for market leverage during the pandemic can be thought of as being driven by highly valued firms. Hence, a cautionary tale must be exercised when interpreting the SOA since convergence rates differ for larger-sized firms compared to smaller-sized firms. These findings are consistent with DeAngelo, DeAngelo, and Whited (2011) who argue that the SOA is not the same for all firms.

3.4 Speed of adjustment for over-levered and under-levered firms

This section refines the estimation from the baseline model specified in Eq. (1) to adjust for the asymmetry in the SOA between over-levered and under-levered firms. Existing evidence suggests that the speed of leverage convergence is the same for all firms except for DeAngelo et al. (2011) who argue otherwise. The pecking order theory posits an asymmetry in the SOA between over-levered and under-levered firms since the cost of funding is higher for over-levered firms (Byoun, 2008; Ho et al., 2021). Faulkender et al. (2012) argue that even when the cost of adjustment is the same for over-levered and under-levered firms, the potential benefits may differ and that the value of the firm also decreases with increasing leverage. It is expected that in crisis, over-levered firms may be at risk more and face costly financing relative to under-levered firms. Hence, the SOA is expected to be higher for over-levered firms.

The results are reported in Table 5. Surprisingly, the estimated SOA toward the book target for the pre-pandemic case is higher for under-levered firms: 8.2% versus 1.4% per quarter. However, during the pandemic period, the SOA toward book target is higher for over-levered firms than for under-levered firms: 9.5% versus 5.2% per quarter. On the other hand, the SOA toward the market target is strikingly higher for over-levered firms: 33.9% versus 5.4% per quarter before the pandemic and 31% versus 4.8% per quarter during the pandemic. While Faulkender et al. (2012) also report higher SOA for over-levered firms, the striking differences in the rate reported in this study can be attributed to the growth prospects of the firm, the tighter credit requirements and additional risks posed by the pandemic.

3.5 Cash and speed of adjustment

Prior evidence shows that US firms hold more cash and invest less due to uncertainty to prevent the destruction of shareholder wealth (e.g. Pinkowitz, Sturgess, & Williamson, 2013, 2016). If this is the case, then how does this behavior affect leverage SOA? Theories of corporate finance point to the imperative function of liquidity in the reduction of transaction costs (Dang, Moshirian, Wee, & Zhang, 2015; Ho et al., 2021). Faulkender et al. cite a firm’s incentive to access the capital market as another reason that affects the cost of leverage adjustment. They argue that “cash cow” firms may generate cash in excess of their profitable investment and may choose between paying off their debt, paying dividends, or repurchasing shares. Evidence suggests that the pandemic led to a reduction in dividend payout and an increase in cash holding (Krieger et al., 2021; Tut, 2021). Due to the increasing uncertainty borne by the pandemic and the worsening of cash flow (Haque & Varghese, 2021), firms with more cash fearing increasing adjustment costs from large deviations may want to converge to their optimal levels of leverage target faster.

Table 6 presents the empirical findings for firms with NPND (Strebulaev & Yang, 2013). Strebulaev and Yang (2013) argued that cash can sometimes be viewed as negative debt and defined NPND as firms with the sum of long-term debt plus debt in current liabilities less cash and short-term investment equal to or less than 0. We adopt this measure for this analysis. The findings are presented for the full sample, low-valued and most-valued firms. The results presented in Panel A (full sample) show that firms with more cash adjust faster towards their book target than those with less cash. However, firms with less cash are estimated to adjust faster toward market leverage than those with more cash. As discussed above, the behavior exhibited toward the SOA for market leverage is a result of the “size effect”.

For instance, in Panel B, we show that low-valued firms with more cash adjusted faster towards both book and market leverage. However, considering the case of highly valued firms (Panel C), we find that the SOA towards the market target is reduced. This could explain the lower SOA towards market leverage in the full sample case (Panel A). These results are expected since low-valued firms with more cash may want to adjust quickly in times of uncertainty to shirk away from increasing adjustment costs which may not be the case for high-valued firms. These results are also confirmed when we consider non-dividend-paying firms. Unreported results (available upon request) show that firms that did not pay dividends adjusted faster toward their target ratio than dividend-paying firms.

3.6 Speed of adjustment in states most affected by the pandemic

This section analyzes the SOA for firms headquartered in states most severely affected by the pandemic. We match the firm's location with the most affected states by the number of reported cases using the state of its headquartered. Evidence suggests that local stock returns increase during local unfavorable economic circumstances since local risk aversion increases with risk-sharing decreasing (Korniotis & Kumar, 2013). These states are also likely to get government support in times of crisis. Vo et al. (2021) find the SOA to be higher for firms situated in countries most affected by the pandemic.

Hence, it is expected that firms headquartered in such states will adjust faster toward their target ratio. To examine this effect, we split the sample into two groups: the top 25 most affected (high cases) and the bottom 25 least affected states (low cases) [9]. Next, we performed the basic analysis on the two groups for both book and market leverage. The results are presented in Table 7. The results confirm prior evidence (Vo et al., 2021) and show that the SOA increases for firms situated in states in terms of higher cases. Quantitively similar results were obtained when we ranked the states in terms of death cases. The results suggest that as the pandemic-related risk in the states in which the firm is headquartered increases, firms tend to adjust towards their target ratio faster to prevent any additional costs of adjustment emanating from deviating more from the target level.

3.7 COVID-19 exposure, firm-level risk and target deviation: a mediation analysis

The previous discussion emphasized that the emergence of the pandemic may potentially increase risk which could cause deviations from the target ratio thereby rendering firms willing to adjust quickly towards their target ratio to avoid costly adjustments as a result of large deviations. This section provides evidence of the degree of increase in firm-level risk brought about by the pandemic. Prior studies show that the COVID-19 shock increases stock market volatility (Baek & Lee, 2021). In this study, we use Hassan et al.'s (2020) firm-level risk and Covid-exposure measures [10] to conduct a mediation analysis and examine first, the degree of deviation from the target ratio induced by the exposure to the pandemic.

Second, how the exposure to the pandemic has increased the overall risk faced by the firm. Lastly, whether firm-level risk fully or partially mediates the effect of the pandemic on target deviation. This analysis is to help shed light on (1) the direct effect of the pandemic on target deviations which warrants an increase or decrease in convergence rate and (2) the indirect effect through which the pandemic affects target deviations through firm-level risk. We predict that if exposure to the pandemic has the potential increasing firm risk, then the exposure to the pandemic may (in)directly affect leverage deviations through overall firm risk.

The results are presented in Table 8 Panel A. The findings show that Ln(RISK) increase (decrease) book (market) deviations by 0.4% (−0.9%) (columns 1 and 4), respectively. However, an unreported subsample analysis (available upon request) shows that for low-valued firms, exposure to the pandemic caused an increase in both book and market deviation. Yet, for high-valued firms, while there was an increase in book deviation, there was a decline in market deviation. This potentially shows that the reduction in market deviation as reported in Table 8 Panel is driven by highly valued firms.

The results further show that exposure to the pandemic caused an increase in firm-level overall risk by roughly 23∼24% (columns 5 and 2). This is suggestive of the increase in SOA to shirk away from the costly nature of adjustments from larger deviations. Additionally, the results (column 3 and 6) shows that firm-level risk (Ln(RISK)) partially mediates the effect of Covid Exposure on how far firms deviate from their optimal level. That is, even though exposure to the pandemic has a significant impact on the overall level of firm risk, the firm risk does not fully absorb the effect of exposure to the pandemic to the deviations from the optimal leverage level.

The significance of this analysis demonstrates that a substantial proportion of the observed increase (decrease) in deviation from book (market) target during the pandemic can be explained partially by overall firm-level risk. This implies that although there is evidence that the pandemic caused a deviation from the target ratio which subsequently affected the SOA, this deviation cannot be solely attributed to the pandemic but also to the overall risk that firms face.

3.8 COVID-19 exposure, risk and SOA

We provide additional evidence on how firms exposed to the pandemic react toward their SOA. To do this, we split the sample into two based on the level of exposure. Firms with exposure values greater than the sample median values are classified as high exposure, otherwise low exposure. As a robustness check, we also use the variable COVID-Risk and split the sample into two following the same step as before. The results for this analysis are reported in Panels B and C of Table 8. Our findings show that firms that exhibited concerns of high exposure or high risk have a higher SOA as compared to low exposure or low risk firms. These findings also confirm our hypothesis that risk-avoidance may induce firms to accelerate how fast they return to their optimal level after deviating from it.