Liquidity Crises and Corporate Cash Holdings in Chile


  • We thank Alejandro Micco, Luis Opazo, two anonymous referees, and seminar attendants at Universidad Alberto Hurtado, University of Chile, and the Central Bank of Chile for useful comments and suggestions.


This paper addresses the heterogonous effects of adverse liquidity shocks on corporate cash holdings in an emerging market. We use a large panel dataset with quarterly financial information for Chilean firms during the period 1996–2009. We find three main results. First, liquidity crises have had a negative and economically significant effect on cash holdings, but mainly for small firms; medium-sized and large firms have not been affected by liquidity crises. Second, liquidity crises reduce the ability of firms to adjust to optimal cash holdings. Finally, medium-sized firms are less able to adjust cash holdings compared to small and large firms.

I. Introduction

This paper addresses the heterogeneous effects of negative liquidity shocks on corporate cash holdings, analyzing the case of one particular emerging market. We are particularly interested in understanding the magnitude of the impact on firms' liquidity and the existence of potential heterogeneous effects according to firm size. We also explore whether liquidity crises reduce the ability of firms to adjust cash holdings and how the adjustment velocity differs by firm size. To address these questions, we use a large panel dataset with quarterly financial information for firms in Chile during the period 1996–2009. In general, little is known about the effects of liquidity crises on firm performance in developing countries. There are some previous studies analyzing the effects of the Asian crisis during the 1990s, but most of them emphasize different issues than those addressed in this paper.1

A relatively large amount of the literature has studied the determinants of firms' liquidity, specifically the factors affecting corporate cash holdings (Almeida, Campello, and Weisbach 2002; Dittmar, Mahrt-Smith, and Servaes 2003; Opler et al. 1999, 2001; Kim, Mauer, and Sherman 1998; Bruinshoofd and Kool 2004; Ozkan and Ozkan 2004). Few studies, however, have analyzed how firms adjust cash holdings when facing liquidity shocks such those experienced by many economies during the recent financial crisis. One exception is Elkinawy and Stater (2007) who investigate the determinants of cash holdings and firms' value in Argentina, Mexico, and Brazil. They analyze how two specific crises—Mexico in 1994–95 and Brazil in 1999—altered the determinants of these variables. However, they do not analyze how the crisis directly affected firms' liquidity.

This is important to study for several reasons. First, a financial crisis may reduce a firm's liquidity and negatively affect its chance of survival or increase its probability of debt default.2 Second, liquidity shocks may have amplified effects on aggregate activity and employment when firms cannot obtain enough cash to fulfill their short-term obligations. Third, the identification of differences in firms' exposure to liquidity shocks could be important for policymakers monitoring financial stability.

The main contribution of this paper is to shed light on how a financial crisis can affect firms' liquidity in the context of a developing country. As we mentioned before, most of the empirical literature in this regard has focused on countries with more developed financial markets. Although our study is based on information from Chilean firms, these implications can be extended to other emerging countries with similar characteristics, such as lower development of capital markets and strong financial links with the rest of the world. In addition, we add to the literature by looking at how firms respond differently to liquidity shocks. Our results suggest that smaller firms are most affected by liquidity shocks, but they tend to adjust their optimal cash holdings more rapidly. This has important implications for policies aimed at alleviating the effects of financial crises.

The remainder of the paper is structured as follows. In section II, we present our data and the main stylized facts on corporate cash holdings across firms and industries. In section III, we present the empirical methodology and discuss the theoretical foundations for our estimations. In section IV, we discuss our results under two alternative definitions of liquidity crises. In section V, we present our conclusions and ideas for future research.

II. Data Description and Main Fact

We obtain our data from listed firms at the Chilean Superintendency of Securities and Insurance (SVS), which is available on a quarterly basis from 1986 to 2009. According to Chilean regulation, companies that offer publicly traded stocks or debt instruments must be registered and file certain information. These firms must provide financial statements and other information on a quarterly basis and be audited once a year. This group also includes companies with more than 500 shareholders and those for which at least 10% of their subscribed capital is in the hands of at least 100 people. There are also cases of specific registry requirements3 and others of voluntary registration.4 Thus, this database provides more coverage than those that only include the largest and most traded firms. In terms of debt, this group of companies holds 100% of local bond issuance, over 70% of corporate external liabilities (bank loans and bonds), but only 20% of local bank loans.5

For our empirical analysis, we use financial data from the firms' nonconsolidated financial statements for the period 1996(Q1)–2009(Q1). Although we try to follow the methodology of previous literature as much as possible, we are subject to the availability of certain financial variables. For example, we were not able to find information on some variables such as dividend payments, advertising and R&D investment, and ownership structures, which previous literature has found to be significant determinants of cash holdings.

We dropped from our sample financial firms, companies with toll road and other infrastructure concessions (project finance), major state-owned enterprises (ENAP, CODELCO, METRO, and EFE),6 and small firms related to the educational, sports, and entertainment sectors. Many of the latter report voluntarily and/or only on a yearly basis. Our final sample consists of 479 firms and 15,402 observations (on average, 283 firms per quarter).7

One limitation of this dataset is that it is biased to large, public-issuing firms. Thus, our results may be not applicable to smaller firms, which tend to be important in terms of domestic bank lending. However, the firms in this sample represent a very large proportion of Chilean GDP.8 In addition, the sample represents the best source of information given its complete financial detail, standardized format, and the availability of quarterly financial statements (FECU). Thus, it contains more financial information than other widely used sources in Chile such as the National Annual Industry Survey (ENIA).

Another potential problem with this dataset is the use of nonconsolidated information. In theory, consolidated data is better suited for financial analysis as it shows the net financial position of a group of companies, eliminating the double counting that arises when individual data of various subsidiaries are aggregated. Indeed, simply adding the individual accounts of companies that belong to the same group leads to overstatement of financial costs and debt ratios, and the flows of financing and liabilities among the members of the group are counted twice. For the purpose of liquidity assessment, using only individual accounts can hide liquidity problems of subsidiaries that are part of the group or economic unit (consolidated) but they are not listed in the SVS. The opposite is also true. A parent company with liquidity problems at the individual level may actually have enough liquid assets in an unlisted subsidiary. The extent of the potential impact caused by this phenomenon is directly linked to the number of intermediate parent companies created by the groups. Despite that, we use nonconsolidated data because consolidated data is not available for the period 1996–2000, which includes the Asian crisis of 1997–98.9

III. Methodological Issues

We follow the previous literature on determinants of cash holdings, specifically Ozkan and Ozkan (2004), who estimate the following equation:

display math(1)

where Cash is firm cash holdings measured as the ratio of total cash and equivalent items to total assets; αi and αt are firm and time fixed effects, respectively; and the vector X includes variables suggested by theoretical models as important for explaining firm-specific differences in cash holdings.

Theoretical models emphasize two main determinants of cash holdings: transaction costs and precautionary motives. These two determinants define the type of variable that we include in the vector X. All of the explanatory variables are defined in Table 1. Table 2 presents the descriptive statistics of these variables. In terms of transaction costs, we use the size of firms to ascertain if large firms have less information asymmetry than small firms. Thus, they would have lower costs of external financing because of scale economies resulting from a substantial fixed cost component of security issuance costs (Barclay and Smith 1996). Following this argument, we expect that larger firms need to hold less cash.

Table 1. Variables Definition
CashCash and equivalent items / Total assets
Cash flowOperating profit / Total assets
VolatilityYearly standard deviation of industry sales growth
Leverage(Current liabilities + Noncurrent liabilities) / Total assets
Liquidity(Current assets − Current liabilities / Total assets) − Cash
Bank debt STShort-term bank borrowing / (Current liabilities + Noncurrent liabilities)
Bank debt LTLong-term bank borrowing / (Current liabilities + Noncurrent liabilities)
SizeThree categories defined from the size distribution: Small, Medium, and Large
CrisisDummy variable defined for quarters when the monetary policy rate is one standard deviation above its trend level
Quarterly dummy variables (53)For each quarter, from the first quarter of 1996 to the first quarter of 2009
Table 2. Descriptive Statistics
VariableObservationsMeanStandard Deviation
Cash flow14,1020.0011.301
Bank Debt ST14,1020.1210.183
Bank Debt LT14,1020.1550.231

The literature suggests firm leverage as another determinant of the cost of funding. In the presence of transaction costs, higher leverage may act as a substitute for cash holdings. Thus, an increase in debt financing would be associated with lower cash holdings. However, it has also been argued that high debt may also increase the probability of financial distress and, through this effect, increase cash holdings (Ozkan and Ozkan 2004).

Access to banking debt may also affect cash holdings decisions. Following the theory of Fama (1985) that banks have some comparative advantage in minimizing information costs and can access information that is not publicly available, bank loans can signal positive information about firms. Thus, firms with higher bank debt are expected to have better access to external finance and less need to hold cash. In our specifications, contrary to previous empirical works, we distinguish between short-term and long-term banking debt to analyze whether the term “debt composition” matters for liquidity. It can be argued that longer-term banking debt generates better signaling information than shorter-term banking debt, reducing the needs of cash holdings by a significant amount.

Finally, we include three other variables that may affect cash holdings: cash flow, assets liquidity, and sales volatility. Firms with higher cash flows, due to the nature of their business, are expected to have lower cash holdings because cash flow provides a ready and immediate source of liquidity. Meanwhile, greater liquid assets reduce cash holdings because the firm would be in a better position to fulfill short-term liquidity needs. Finally, in the case of sales volatility, a positive relationship is expected because a higher variability increases cash holdings as a way to hedge uncertainty (precautionary motive).

The dynamic specification in equation (1) has been justified by the existence of adjustment costs on some unobserved level of desired cash holdings (Ozkan and Ozkan 2004). These costs may impede instantaneous variations in cash holdings following changes in firm-specific characteristics and/or random shocks.

We extend this specification to analyze four main issues. First, we are interested in analyzing how liquidity crises episodes have affected overall cash holdings. We include a dummy variable for these episodes in the specification of equation (1). Second, we study how firm liquidity adjustments may be affected during a crisis. To do so, we include an interaction term between the lagged dependent variable and the dummy variable for liquidity crises. We hypothesize that adjustment costs could be exacerbated during liquidity crises when the financial system experiences severe restrictions to normal functioning.10 Third, we study whether firm size affects cash holding adjustments by introducing an interaction between lagged cash holdings and size dummy variables. Fourth, we analyze how the effect of the crises, if any, depends on firm size. We introduce interactions between the dummy for liquidity crises and size dummy variables.

Considering these four issues, the empirical model is specified as:

display math

where Crisis is a dummy variable for episodes of liquidity shocks, and Z is a set of two size dummy variables: one for medium-sized firms and the other for large-sized firms. We need to define episodes of liquidity crisis. To do so, we look at periods where the economy has been affected by a significant liquidity shock. The recent fall in liquidity following the Lehmans' bankruptcy in September 2008 is one example. The previous Asian crisis is another potential candidate. In order to define these episodes in a more formal way, we use deviations of the Monetary Policy Rate (MPR) from its trend using the Hodrick-Prescott filter. A liquidity crisis period is defined for those quarters when the MPR is over one standard deviation from its trend level.11

As shown in Figure 1, in addition to well-known crises, this procedure allows us to identify specific quarters during the period 2001–2 as another liquidity shock. Thus, under this definition, we have three crisis periods. The Asian shock analysis covers the period 1998(Q2)–(Q4). The second is the period 2001(Q3)–2(Q1). The third is the recent financial crisis and covers the period 2008(Q3)–(Q4).

Figure 1.

Monetary Policy Rate 1996(1Q)–2009(1Q)

Source: Central Bank of Chile and authors' calculations.

The estimation of this model using standard fixed effects presents some econometric challenges. First, there is an expected correlation between the error term and the lagged dependent variable. Second, most of the variables contained in the vector X are not strictly exogenous. One solution is to use the Arellano and Bover (1995) and Blundell and Bond (1998) estimator for dynamic panel data, known as System GMM. This estimator augments Arellano and Bond (1991) by making the additional assumption that the first differences of instrumental variables are uncorrelated with the fixed effects.12

A crucial assumption for the validity of GMM is that the instruments are exogenous. When the model is overidentified, the validity of these assumptions can be tested using the standard GMM test statistic for overidentifying restrictions, or a Sargan/Hansen test under the null that the implied moment conditions are valid (see Sargan 1958; Hansen 1982). In this context, another key assumption is that there is not a serial correlation in the disturbances εit. The null hypothesis is that there is not a second-order serial correlation in the first-differenced residuals (see Arellano and Bond 1991).13

Table 3. Panel Data Regressions: Effects of Liquidity Crisis on Cash Holdings
  1. Note: Robust t-statistics in parentheses.
  2. **, and * represent statistical significance at the 1% and 5% level, respectively. All regressions include unreported quarter-specific dummy variables. Following Roodman (2009), Cashit−2 and Xit−2 were used as instruments and the dimension of the instrument matrix was reduced by collapsing it horizontally in order to mitigate the instrument proliferation that weakens Sargan test of overidentifying restrictions.
Cash flow0.0000.0000.0000.000
Bank debt ST−0.027−0.028−0.025−0.025
Bank debt LT−0.015−0.015−0.016−0.016
Crisis*Cash(-1) 0.3090.3430.369
Cash(-1)*Medium  0.3130.318
Cash(-1)*Large  0.1990.201
Crisis*Medium   0.024
Crisis*Large   0.026
Sargan p-value0.0180.1260.3210.340
AR(1) p-value0.0000.0000.0000.000
AR(2) p-value0.1290.2740.2450.259

IV. Econometric Results

A. Main Results

In Table 3, we present our first set of results using the liquidity crisis definition corresponding to significant deviations from the monetary policy rate.14 Considering the main determinants of cash holdings, the evidence is mostly consistent with the results from previous works. The exception is the positive, but not significant, effect of cash flow on cash holdings. Like Ozkan and Ozkan (2004), among others, we find that the effect of leverage on cash holdings is negative and significant. We also find that a large proportion of liquid assets and banking debt (both short- and long-term) reduce cash holdings. Similar to previous studies, our results show that larger firms keep fewer cash holdings than small firms. Also, we find that an increment in industry volatility is associated with an increase in cash holdings.

The coefficient of lagged cash holdings is positive and significantly different from 0 at 1%, providing evidence that firms face important adjustment costs for reaching a target cash ratio. The adjustment coefficient (1-δ0) in column (1) is about 0.4, lower than similar estimations for British companies as reported by Ozkan and Ozkan (2004). This coefficient implies that approximately 84% of one liquidity shock is dissipated in four quarters.

In general, our results show that liquidity crises are associated with a reduction in corporate cash holdings, although the parameter is not significant in column (1). In columns (2), (3), and (4), we sequentially include the rest of the interaction terms for analyzing heterogeneities for the impact of liquidity crises and adjustment velocity across firms. We introduce interactions between lagged cash holdings and crises, lagged cash holdings and firm size, and firm size and crises. The effect of the other determinants of cash holdings (volatility, leverage, etc.) tend to be robust to the inclusion of these interactions.

In terms of our main questions, we find that the interaction of crises and lagged cash holdings is positive and significant, suggesting that adjustment costs are larger (or adjustment velocity lower) during episodes of negative liquidity shocks. This is consistent with our hypothesis that adjustment costs are exacerbated during liquidity crises when the financial system experiences severe restrictions to normal functioning.

The interactions between lagged cash holdings and dummy variables for medium-sized and large firms are positive and significant (column 4). Moreover, the parameter is higher for medium-sized firms. This implies that there is a nonlinear relationship between size and adjustment costs. Our results suggest that medium-sized firms have larger adjustment costs compared to small and large firms.

Finally, the interaction between crisis and size dummy variables are positive and significant, showing that medium-sized and large firms should be relatively less affected than small firms during episodes of liquidity crisis. However, as we show below, when we calculate the parameter for the effect of liquidity crises on different firm sizes, we find that the effect of liquidity crises on cash holdings is only statistically significant for small firms.

B. Differences across Firms

To shed light on the quantitative differences of the estimated effects across firms, we calculate the parameter of adjustment velocity for the entire sample and for the three different size categories.15 For all of these parameters, we also calculate the confidence interval from estimations of column (4) in Table 3 to look at whether differences are statistically and economically significant.

The parameters and the confidence interval for the adjustment velocity are shown in Figure 2. The velocity of adjustment is approximately 0.26 when it is evaluated at the sample mean. However, this coefficient is highly heterogeneous across firm size. We find that the adjustment coefficient is larger for small firms (0.43), than for medium-sized (0.12) and large (0.23) firms.16 Also, according to the confidence intervals displayed as the line crossing the bar, the values for small and large firms are statistically higher than the adjustment velocity calculated for medium-sized firms. This evidence is the same for both indicators of liquidity crisis. According to these coefficients, about 90% of any liquidity shock dissipates after one year for smaller firms. In contrast, approximately 40% and 60% of the liquidity shock is eliminated after one year in medium-sized and large firms, respectively.

Figure 2.

Adjustment Velocity by Firm Size

This nonmonotonic relationship between size and adjustment costs may be consistent with the differences in credit access across firm size. In fact, it can be argued that medium-sized firms are more dependent on external funding for financing liquidity than smaller firms, but do not have as many funding alternatives as larger firms. In this case, our findings would be consistent with the idea that medium-sized firms need to adjust more quickly than other types of firms.

We also present the calculations of the crisis dummy for the entire sample and according to firm size. These “crises effects” are shown in Figure 3. First, we find that a liquidity crisis is associated with a reduction in cash holding of between 0.4% and 1.5% of total assets in the short term, and about 1.7% to 5.4% in the long term.17 These figures are also economically significant considering that the mean of the dependent variable (cash holdings over total assets) is 5.2% and the standard deviation is 13.4% (see Table 2).

Figure 3.

Liquidity Crisis Effects by Firm Size

Interestingly, our results show that the negative effect of liquidity shocks is only significant for small firms. As shown in Figure 3, we can reject the hypothesis that the coefficient for small firms is the same as those for medium-sized and large firms. As is evident from this figure, the coefficients for medium-sized and large firms are not statistically significant. Thus, these results are in line with the hypothesis that smaller firms have lower access to credit and, consequently, could be more affected in episodes of liquidity shortages.

V. Conclusions

This paper uses a large panel dataset of Chilean firms to address how cash holding decisions may be affected by episodes of negative liquidity shocks. We are particularly motivated by the recent financial crisis and the lack of evidence for developing countries. The results are particularly interesting for countries whose financial markets are less developed and for whom it has been traditionally argued that smaller firms are disproportionally affected during episodes of liquidity crisis.

Our main results are similar to those from previous empirical analysis. They show that size, leverage, bank debt, and other liquid assets negatively affect cash holdings. We also find that higher industry volatility increases cash holdings, which is consistent with the idea that higher liquidity is partially motivated by precautionary motives. In general, episodes of liquidity crisis have a negative and economically significant effect on firms' cash holdings, mainly for small firms. We find that a liquidity crisis reduces cash holdings of small firms by 2.2% of total assets, an economically important effect considering that the mean of this variable is 4.9%.

These findings can have important policy implications for addressing liquidity problems in developing countries during a financial crisis. At least in the case of Chile, our results suggest that there should be specific policies oriented to alleviating liquidity problems for smaller firms. Related to this finding, the fact that monetary authorities implemented special interventions during these periods in order to inject liquidity into financial markets is a relevant issue for future research. This was especially true during the recent crisis when most public authorities around the world actively intervened to reduce the negative effects of the international liquidity shock. Without information on the magnitude and the timing of these interventions, it is difficult to understand what would have occurred in the absence of these policy measures. Following the analysis in this paper, it would be interesting to also study how these interventions have differentiated effects across firms.

We also analyzed whether firms differ in their ability to adjust cash holdings. The evidence reveals that there is a nonlinear relationship between size and adjustment costs; medium-sized firms have larger adjustment costs (or lower adjustment velocity) in comparison with small and large firms. This nonlinear relationship between adjustment costs and size may also be due to differences in credit access across firm size. In fact, it can be argued that medium-sized firms have larger adjustment costs because they are more dependent on external funding for financing liquidity than smaller firms, but do not have as many funding alternatives as larger firms. In this case, compared to firms in the other size segments, medium-sized firms would face more difficulties for adjusting cash holdings to desired levels. We acknowledge that this result can be specific to Chilean firms and this result might not directly extend to other developing countries. This could be an interesting question to be addressed using a panel of firms in different countries. Related to this issue, it could be also interesting to analyze the consequences of these differences in the adjustment costs on firm performance and how they can be reduced.


  1. 1

    See, for example, Claessens, Djankov, and Xu (2000), and Blalock, Gertler, and Levine (2008).

  2. 2

    See, for example, Jacobson et al. (2008) for evidence on the relationship between default and liquidity.

  3. 3

    Such as requirements for bus service companies of the new public transportation system in the city of Santiago (Plan Transantiago).

  4. 4

    In these cases, companies aim to have better access to different types of financing.

  5. 5

    Data are form Superintendency of Securities and Insurance, Superintendency of Banks and Financial Institutions, and Central Bank of Chile.

  6. 6

    These firms are not considered in our sample because they have government-specific guarantees and credit access terms that differ from the rest of the companies in the sample. These features provide them with enough support in the event of financial distress, even though they can present weak financial indicators.

  7. 7

    The maximum is 317 firms in the third quarter of 2007 and the minimum is 218 firms in the first quarter of 2009. This reduction at the end of the period is because 80 firms began to publish their financial statements under different standards (International Financial Reporting Standards, IFRS).

  8. 8

    The financial debt (banks and bonds) represented almost 41% of the Chilean GDP at the end of 2008.

  9. 9

    We have compared indicators using consolidated and nonconsolidated information for the period 2001–9 and, even though there are changes in cash holdings, there are not important differences for a firm or industry over time.

  10. 10

    In fact, the current financial crisis has shown that central banks intervene through liquidity injections to the financial system when there is liquidity shortage in the economy.

  11. 11

    We have used an alternative definition of liquidity crisis, taking those periods when the inter-banking loans rate is significantly above—one standard deviation—the monetary policy rate. The results are very similar to those shown here and are available upon request.

  12. 12

    The Arellano-Bover/Blundell-Bond estimator builds a system of two equations—the original equation as well as the transformed one—allowing the introduction of more instruments. This can improve efficiency.

  13. 13

    As we show in the last rows of Table 3, we cannot reject that the instruments are valid and that there is not second order autocorrelation.

  14. 14

    We have also used a continuous variable for crisis, defined as the deviation of the monetary policy rate from its estimated trend, but the results do not show a negative and significant relationship with cash holdings. This is consistent with the view in this paper that only large increases in interest rates are associated with a liquidity crisis.

  15. 15

    The parameter for the velocity of adjustment is given by (1 − δ0δ2Crisisδ3mMediumδ3lLarge), where Crisis, Medium, and Large are the sample mean of the dummy variables included in the estimation. This parameter is computed as (1 − δ0δ2Crisis) for small firms, (1 − δ0δ2Crisisδ3mMedium) for medium-sized firms, and (1 − δ0δ2Crisisδ3lLarge) for large firms.

  16. 16

    In this case, the parameter is calculated taking the sample mean of the explanatory variables and the corresponding value of the size dummy variable.

  17. 17

    In Figure 3, we only show the short-term parameters. The long-term parameters are obtained by dividing the short-term parameters by the coefficient of the estimated adjustment velocity.