Download Cash Holdings and Firm Value during Latin American Financial Crises

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Internal rate of return wikipedia , lookup

Present value wikipedia , lookup

Business valuation wikipedia , lookup

Financialization wikipedia , lookup

Pensions crisis wikipedia , lookup

1997 Asian financial crisis wikipedia , lookup

Financial crisis wikipedia , lookup

Corporate finance wikipedia , lookup

Global saving glut wikipedia , lookup

Transcript
Cash Holdings and Firm Value during Latin American Financial Crises
Susan Elkinawy*
Loyola Marymount University
Hilton Center for Business
One LMU Drive, MS 8385
Los Angeles, CA 90045-2659
Phone: (310) 338-2345
Fax: (310) 338-3000
[email protected]
Mark Stater
University of Georgia
Department of Public Administration and Policy
204 Baldwin Hall
Athens, GA 30602
Phone: (706) 542-2873
Fax: (706) 583-0610
[email protected]
June 2007
Abstract
We investigate the determinants of cash holdings and firm value of Latin American firms
during the Mexican crisis of 1994-1995 and the Brazilian crisis of 1999. We find that
each crisis alters the determinants of cash and firm value differently. Larger firms
increase their cash holdings during the Mexican crisis, while cross-listed firms increase
their holdings of cash during the Brazilian crisis. In addition, during the Brazilian crisis
we find an increase in ownership concentration is more value-enhancing for high-cash
firms. Our findings suggest that while illiquidity may become more severe during crises,
agency costs are also an important factor.
JEL classification: G30; G32
Keywords: Cash holdings; Firm Value; Corporate Governance; Financial crises;
Emerging markets
* Corresponding author.
Cash Holdings and Firm Value during Latin American Financial Crises
June 2007
Abstract
We investigate the determinants of cash holdings and firm value of Latin American firms
during the Mexican crisis of 1994-1995 and the Brazilian crisis of 1999. We find that
each crisis alters the determinants of cash and firm value differently. Larger firms
increase their cash holdings during the Mexican crisis, while cross-listed firms increase
their holdings of cash during the Brazilian crisis. In addition, during the Brazilian crisis
we find an increase in ownership concentration is more value-enhancing for high-cash
firms. Our findings suggest that while illiquidity may become more severe during crises,
agency costs are also an important factor.
1
1. Introduction
Corporate cash holdings can be characterized as both good and bad for
shareholders. On the upside, firms with high cash holdings have superior growth and
investment opportunities as well as better operating performance indicators than low-cash
firms (Opler, et al., 1999; Mikkelson and Partch, 2003). However, it is also the case that
governance arrangements that foster the accumulation of cash are less conducive to the
maximization of shareholder wealth. High cash holdings coincide with indicators of poor
corporate governance, such as weak country-level shareholder protection laws, larger
boards of directors, insider dominance of boards, and a high concentration of insider
ownership (Dittmar, Mahrt-Smith and Servaes, 2003; Ferreira and Vilela, 2004; Ozkan
and Ozkan, 2004; Kusnadi, 2005; Chang and Noorbakhsh, 2006). Thus, the downside of
cash holdings is that they can signify the presence of significant agency problems. Not
surprisingly, research has found a negative effect of cash holdings on firm value that is
most pronounced when shareholder protection is weak (Kalcheva and Lins, 2006;
Pinkowitz, Stulz, and Williamson, 2006).
Managers have stronger preferences for cash holdings than shareholders due to
the relative ease with which managers can spend cash on items that improve their own
utility but do not necessarily raise the value of the firm (Jensen, 1986). Cash is more
liquid and more difficult to monitor than the firm’s other financial and non-financial
assets. Moreover, the temptation to dissipate cash on pet projects may be especially
strong during financial crises in emerging markets because increased market uncertainty
can provide a screen for value-reducing self-interested behavior on the part of managers.
2
On the other hand, cash holdings may be beneficial to firms in surviving a severe market
downturn as reduced liquidity may increase the difficulty in securing alternate financing.
Our study examines the determinants of corporate cash holdings and firm value in
a decade when multiple crises hit Latin America and many countries in the region
underwent extensive liberalization reforms. Using panel data on publicly-traded nonfinancial firms in Argentina, Brazil, and Mexico from 1990 to 1999, we consider periods
of time encompassing the Mexican crisis of 1994-1995 and the Brazilian crisis of 1999.
The findings indicate that profitability, firm size, leverage, and measures of governance
all affect cash holdings and firm value during crisis and/or pre-crisis periods. While our
results suggests that liquidity constraints are more severe during crises, there is also
evidence that rising agency costs play a role in asset allocation decisions during financial
crises.
In both crises firms that should have greater access to alternate sources of
financing increase their cash holdings. In addition, we find that the relationship between
cash and Tobin’s Q differs based on the degree of ownership concentration.
A relatively small body of work has thus far examined how cash holdings respond
to crisis conditions.
Arslan, Florackis, and Ozkan (2006) consider the effects of
financing constraints on Turkish firms during the Turkish financial crisis of 2001-2002
and find that more financially-constrained firms exhibit greater investment-cash flow
sensitivity during the crisis than less constrained firms, suggesting liquidity is an
important consideration during market downturns.
Our study complements these
findings by examining the effect of liquidity during crises in a different developing
region. We also include a measure of ownership concentration to potentially shed light
on governance issues. By examining multiple markets over multiple crises, our findings
3
illuminate how governance and liquidity constraints are manifested across different
countries and time periods.
We find for example that high-cash firms in Brazil
experience higher Tobin’s Q values in both the Mexican and Brazilian crisis. In contrast,
high-cash Mexican firms experience lower values of Tobin’s Q during the Brazilian
crisis. Thus, our results suggest that financial crises differ in the way that they alter
managerial behavior and in the manner that the market values the firm. Moreover, these
differences are apparent not only across crises but also among individual countries.
The remainder of this paper is organized as follows. Section 2 discusses our
hypotheses concerning the effects of firm characteristics on cash holdings and firm value
during crisis and non-crisis periods, while Section 3 describes our data and reports the
descriptive statistics. Section 4 outlines our empirical methodology for predicting cash
and firm value, and Section 5 presents the results. Section 6 concludes.
2. Hypotheses
2.1 Hypotheses for Cash Holdings
The effects of firm characteristics on cash holdings are well-documented in the
finance literature. Firms where management is more accountable to shareholders (either
due to firm structure or country laws) have lower agency costs and tend to hold less cash
than firms with weaker governance (Dittmar, Mahrt-Smith, and Servaes, 2003; Chang
and Noorbakhsh, 2006). Opler, et al. (1999) identify features of firms that can foster
agency conflicts and produce high cash holdings as a result. These include dispersed
ownership, large firm size (since size is a deterrent to takeovers), and low debt due to less
monitoring by capital markets.
However, the evidence suggests that financial
characteristics associated with low holdings of cash and liquid assets include large firm
4
size, high leverage, the payment of dividends, high credit ratings, low ownership
concentration, and low costs of external financing (Kim, Mauer, and Sherman, 1998;
Opler, et al., 1999; Colquitt, Sommer, and Godwin, 1999; Pinkowitz and Williamson,
2001; Dittmar, Mahrt-Smith, and Servaes, 2003). In view of the strong consensus in the
literature on the determinants of cash holdings, we adopt the aforementioned
relationships as the expected effects of firm characteristics on cash holdings during precrisis time periods.
A distinguishing feature of our work is the explicit focus on how the determinants
of cash holdings differ between crisis and pre-crisis periods. Our motivating observation
is that market conditions become increasingly uncertain (i.e., information asymmetries
are stronger) during a crisis. To the extent that agency conflicts and liquidity constraints
are greater, this should create a tendency for firms to hold more cash during crises.
However, to disentangle governance explanations from liquidity explanations we expect
that firms that have limited access to alternative sources of financing will hold more cash
during crises. Thus, small firms and firms that are only listed on local stock exchanges
should increase their holdings of cash relatively more than large, cross-listed firms during
crises if liquidity is the main consideration.
We also expect that firms with more
dispersed ownership should have less need to hold cash than firms with more
concentrated ownership during a crisis due to the relatively greater liquidity of widelyheld shares.
2.2 Hypotheses for Firm Value
Firm characteristics that signify growth opportunities and good governance are
generally rewarded by the market, resulting in value premiums for these firms. Because
5
cash holdings not only signify higher agency costs but also arise from greater investment
opportunities (Opler, et al., 1999), the expected overall effect of cash holdings on firm
value is uncertain. During financial crises, the effect of cash on firm value is also
uncertain due to potentially greater liquidity needs.
In general, we expect firm value to decrease during a crisis for a given set of firm
characteristics because of increased market uncertainty and poor market performance.
These factors should reduce the demand for stocks in the affected region on world
markets. Because strong financial health should be highly desirable to investors in the
midst of financial turmoil, we expect crisis conditions to produce an increase in the
market valuation of firms that demonstrate good operating performance and desirable
governance characteristics. Thus, we expect the effects of characteristics such as high
profits, sufficient cash substitutes, high dividend payments, and cross listing to have more
positive effects on firm value during a crisis than before.
The effect of ownership concentration on firm valuation during crises is uncertain,
due to the degree of monitoring associated with share ownership. One issue is the
composition of the block owners. If block owners are primarily insiders, governance
problems could be greater due to the potential for greater expropriation. If, however, the
block owners are mostly outsiders they should provide better monitoring during crises.
The present data do not include information on whether the block owners are insiders or
outsiders. However, since family ownership is common in Latin American firms, there is
a reasonable probability that at least some of the block holders are insiders. This suggests
that firm values could drop with an increase in ownership concentration. On the other
6
hand, it is also possible that greater ownership dispersion makes monitoring of the firm
more difficult during crises due to collective action problems.
3. Data and Descriptive Statistics
3.1 Data
Our subsequent empirical analysis tests the hypotheses outlined in the previous
section. We obtain our firm financial data from Economatica, a company that specializes
in collecting Latin American firm data. This database tracks every publicly listed Latin
American company, providing more comprehensive coverage than databases that
generally only include large firms.
In addition, Economatica maintains financial
information on de-listed firms, eliminating the survivorship bias that is a problem with
many developed and emerging market databases. We focus on firms in three of the
largest Latin American markets: Argentina, Brazil, and Mexico. Information on cash
holdings of Chilean firms is unavailable and thus Chilean firms are excluded from our
analysis. We eliminate financial firms (NAIC code 52) and follow the literature as much
as possible subject to data availability constraints in our choice of financial variables.
The variables describing the firms in the sample therefore include cash holdings, Tobin’s
Q, total assets, leverage, net working capital, operating profit, a dummy variable
indicating whether or not a firm pays dividends in a particular year, and a dummy for the
presence of an American Depositary receipt (ADR) indicating whether or not a firm is
cross-listed on a U.S. stock exchange. Data on ADRs are obtained from the Bank of New
York. All variables are measured annually in U.S. dollars from 1990 through 1999.
More detailed descriptions of the variables are given in the Appendix.
7
We define two crisis analysis periods for purposes of testing our hypotheses. The
Mexican crisis analysis period is 1990-1995, with 1994 and 1995 designated the crisis
years. The Brazilian crisis analysis period is 1996-1999, with 1999 designated the crisis
year. Our approach is to compare the effects of firm characteristics on cash holdings and
firm value in pre-crisis versus crisis periods. For each crisis analysis period, we include
in the sample only those firms that have available data on all regression variables in at
least one pre-crisis year and at least one crisis year. Thus, we are only measuring the
effects for firms that survived the crisis. Doing this assures that we compare the same set
of firms in both pre-crisis and crisis periods, which allows us to capture how managerial
behavior is altered by the crises. Our final sample consists of 346 firms and 1,407 firmyear observations for the Mexican crisis and 403 firms and 1,278 firm-year observations
for the Brazilian crisis.
For a subset of firms in our database (i.e., 253 Brazilian firms and 761 firm-year
observations from 1996-1999), we are able to construct an ownership dispersion variable
referred to as “free float.” This variable measures the percentage of a firm’s outstanding
shares that are not held by blockholders (where blockholders are defined as shareholders
holding at least 5% of the firm’s shares). As mentioned earlier, the data do not designate
whether the blockholders are insiders or outsiders, so more concentrated ownership (i.e.,
lower free float) may or may not be an indicator of poor governance. However, because
it is generally well-known that there is a high incidence of family ownership among Latin
American firms, block holders of firms in this region are likely insiders and consequently
have greater incentive to expropriate from minority shareholders.
8
Figures 1 and 2 present the average cash to net asset ratio in each of our sample
countries during the Mexican and Brazilian crisis analysis periods, respectively. What is
most apparent from comparing the two graphs is the difference in the trend in each
period. In the Mexican period, cash holdings show a general upward trend prior to 1994
(the year the crisis began) with the exception of Mexico which experiences a sharp drop
in cash holdings in the early 1990s. In the Brazilian period cash holdings are decreasing
at a roughly linear rate in all three countries. During the Brazilian crisis, the rate of
decrease is largest in Brazil, the crisis country.
Figures 3 and 4 repeat the above analyses for Tobin’s Q. Again the graphs
display very different trends in the two crises. During the Mexican period Tobin’s Q
displays an inverted U-shaped pattern for Mexico and Brazil, but is decreasing
monotonically at a slow rate for Argentina. Over the Brazilian period the pattern of
Tobin’s Q is much flatter. Tobin’s Q is lowest in 1998, right before the Brazilian crisis.
Brazilian firms have the lowest values of Tobin’s Q over the entire 1996-1999 period.
The evidence from the cross-country comparisons suggests that the two crises
have noticeably different effects on cash and firm value. However, the effects for both
variables are more uniform across countries in the Brazilian than in the Mexican crisis.
3.2 Descriptive Statistics
Table 1 presents descriptive statistics for the variables used in our empirical tests.
For each crisis, we conduct t-tests for differences in means between pre-crisis and duringcrisis periods. The results indicate that Latin American firms hold a significantly higher
proportion of their net assets in cash during the Mexican crisis (11.4%) than in the precrisis period (7.5%). In contrast, Latin American firms hold a lower percentage of their
9
net assets as cash during the Brazilian crisis (8.7%) than they do before the crisis
(13.8%). Other differences observed are that firms are significantly larger, more highly
leveraged, and increase their incidence of cross-listing between the Mexican crisis and
pre-crisis period but not during the corresponding Brazilian period. In the Brazilian
crisis, fewer firms pay dividends and share ownership becomes more concentrated
relative to the pre-crisis period. The point estimates also indicate no significant change in
Tobin’s Q during either crisis, perhaps due to the value-enhancing effects of the early
1990s liberalization and trade reforms cumulating over the decade. Though one might be
tempted to conclude that the quality of corporate governance is weakened by the
Brazilian crisis, it is important to note that the descriptive findings do not control for
changes in firm characteristics from pre-crisis to crisis periods that affect cash holdings
and firm value. The following empirical analysis applies the appropriate controls.
4. Empirical Methodology
4.1 Cash Holdings
Our empirical analysis is similar to that employed elsewhere in the literature,
except that we focus on how crises affect the determinants of cash holdings and firm
value. For each crisis, we estimate a linear regression model for the natural log of a
firm’s cash to net asset ratio1 as a function of firm governance and financial
characteristics, a crisis dummy, and a set of interaction terms between firm characteristics
and the crisis dummy. Letting the subscripts i and t denote firms and years, respectively,
our models are expressed as:
1
The natural log of the cash to net asset ratio is used as the dependent variable instead of the ratio itself in
order to assure that the dependent variable has range (-∞,∞) as must be the case for the classical regression
assumption of a normally distributed error term to hold.
10
ln C it = αMex _ crisist + x it β + [x it × Mex _ crisis t ]γ + λi + ε it , t ∈ {1990 − 95} [4.1]
ln C it = αBraz _ crisis t + x it β + [ x it × Braz _ crisis t ]γ + λi + ε it , t ∈ {1996 − 99} , [4.2]
where Cit is the ratio of cash and cash equivalents to non-cash assets for firm i in year t;
xit is a vector of firm variables that includes total assets (or size), leverage, net working
capital, operating profit, a dummy variable for whether the firm pays dividends in year t,
and a dummy for whether the firm is cross-listed as an ADR in year t;2 Mex_crisist is a
dummy variable that equals 1 if year t is one of the years of the Mexican crisis (i.e.,
1994-1995) and equals 0 if t is one of the years in our sample period preceding the
Mexican crisis (i.e., 1990-1993); Braz_crisist is a dummy variable that equals 1 if t is the
year of the Brazilian crisis (1999) and equals 0 if t is one of the years in the sample
period between the Mexican and Brazilian crises (1996-1998); λi is a firm fixed effect;3,4
and εit is a zero-mean idiosyncratic disturbance term that varies across firms and years
and is assumed uncorrelated with xit. Estimation is with fixed effects, which allows for
non-zero correlation between λi and xit. As mentioned previously, we only examine
firms that survived the crisis in order to compare the same set of firms and associated
managerial behavior pre and during-crisis. More complete definitions of the control
variables in xit are provided in the Appendix.
Equations [4.1] and [4.2] assume that there are no cross-country differences in the
intercepts of cash holdings either before or during the crisis. Likewise, they assume the
2
Tobin’s Q is not included in our cash holdings model because of possible endogeneity (i.e, cash holdings
may affect value because it is a signal of greater agency costs).
3
The firm fixed effect includes both observed and unobserved firm characteristics that do not vary over
time. Thus, in addition to characteristics that are not observed such as the composition of the board, the
fixed effect captures variables that are observed but exhibit no time variation, such as the firm’s country of
location. For this reason we do not include country dummy variables in equations [4.1] and [4.2].
4
In some cases, the ADR and dividend dummies are time-invariant for a given firm. However, for some
firms, ADR and dividend payment status change over time, which implies that the effects of these variables
can be estimated in a fixed effects model. However, their effects are identified solely based off of variation
for firms that experience changes in these variables over time.
11
effects of firm characteristics on cash holdings are the same in all three countries both
before and during the crisis.
These assumptions are necessary for fixed effects
estimation, but are quite strong given that access to liquidity can differ among countries
even within the same geographic region. Thus, we also estimate each of the equations
[4.1] and [4.2] separately for each country, thereby allowing each country to have its own
coefficient vector for each crisis.5
4.2 Methodology for Firm Value
Our technique for estimating firm value is similar to that for estimating cash
holdings. In particular, we estimate a fixed effects model of Tobin’s Q as a function of
firm characteristics for each crisis. The regression models are given by:
Qit = αMex _ crisis t + w it β + [w it × Mex _ crisis t ]γ + λi + ε it , t ∈ {1990 − 95}
[4.3]
Qit = αBraz _ crisist + w it β + [w it × Braz _ crisis t ]γ + λi + ε it , t ∈ {1996 − 99} , [4.4]
where Qit is Tobin’s Q for firm i in year t and wit is a vector of firm-level controls that
includes cash holdings in addition to the same variables that are in the cash regressions.
Thus, we assume that cash holdings directly affect firm value (e.g., because investors can
collectively react to cash holdings by buying or selling shares and affecting equity prices)
but not the reverse (i.e., the decision of how much cash to hold is made independently of
the market value of the firm). Although an important counterpoint is that uncontrolled
governance variables (e.g., board composition) could form a spurious link between value
and cash such that a cash regression would pick up a significant firm value effect (giving
the false appearance of a direct impact), our analysis expunges the impact of these
5
Note that the country-specific estimates may have lower statistical power than the estimates of the model
that pools firms from all three countries.
12
variables to the extent that they are roughly constant over time.
This is because
uncontrolled time-invariant items are subsumed into the fixed effect term λi. Based on
this logic, our approach is to include cash holdings directly in the value regressions [4.3]
– [4.4] but to exclude firm value from the cash regressions [4.1] – [4.2].
As in the cash holdings models, we estimate the determinants of firm value
allowing the effects of the firm characteristics to differ between pre-crisis and crisis
periods but not across countries. Holding characteristics constant, we assume that firm
value is the same across countries in each crisis analysis sub-period. This assumption
would be reasonable if these markets were completely integrated, but this state of affairs
is unlikely given that the markets are still developing and were only recently liberalized
at the time of our data. Thus, we also estimate each equation [4.3] and [4.4] separately
for each country. The cost of this approach is reduced degrees of freedom and lower
statistical power, but the benefit is that each country is has a unique vector of regression
slope coefficients. In addition, to measure the effects of ownership concentration, we
estimate versions of [4.2] and [4.4] for Brazil that include free float, since this variable is
available exclusively for Brazilian firms from 1996-1999.
5. Results
5.1 Results for Cash Holdings before and during the Mexican crisis
In Table 2 we report the results of our cash holdings regression [4.1], both pooled
across countries and separately for each country, for the Mexican crisis analysis period of
1990-1995. The first column shows the results for the regression with the three countries
combined and the second through fourth columns show the country-specific regressions.
13
The results in the first column indicate that cash holdings significantly decline
during the Mexican crisis, holding firm characteristics constant. This result may seem
surprising given that one typically associates higher agency costs (such as those expected
to occur during a crisis) with higher cash holdings. However, the result is consistent with
a dramatic decrease in access to liquidity during a crisis that prevents firms from holding
as much cash as they otherwise would in the presence of greater asymmetric information.
Furthermore, the phenomenon of firms holding less cash in the crisis is also consistent
with the dissipation of cash on managerial “pet projects,” and thus may be a reflection of
higher agency costs rather than agency costs that paradoxically decline (Harford, Mansi,
and Maxwell, 2005).
In the pre-Mexican crisis period, firm size has a significant positive effect on cash
holdings. The magnitude of the estimate implies that a 1 percent increase in total assets
results in a 0.38 percent increase in the cash to net asset ratio. During the crisis period,
the effect of size becomes even larger. Each 1 percent increase in total assets now raises
the cash to net asset ratio by 0.53 percent. Although large firms tend to have easier
access to alternate sources of financing and thus should have less need to hold cash than
smaller firms, note that Kusnadi (2005), Kalcheva and Lins (2006), and Niskanen and
Niskanen (2007) also document a positive size effect, which may reflect higher agency
costs in large firms. Thus, the results are consistent with relatively greater agency
problems during the crisis that foster the accumulation of more cash.
The remaining results in the first column indicate that several firm characteristics
have significant effects on cash holdings before and during the Mexican crisis. Net
working capital is a substitute for cash in the pre-crisis period, but becomes a weaker
14
substitute during the crisis. This may indicate an unwillingness to substitute non-cash for
cash assets during a crisis. Operating profit positively affects cash holdings in the precrisis period, and its effect in the crisis period does not undergo a significant change.
Firms that pay dividends hold more cash in the pre-crisis period, and this effect is also
unchanged during the crisis.
In the second through fourth columns of Table 2 we re-estimate equations [4.1]
and [4.2] separately for each country to allow for differences in slopes and intercepts
across countries. The results suggest that many of the results for the pooled sample are
driven by Brazilian firms, which is not surprising since both the number of firms and the
number of firm-year observations are considerably higher for Brazil than the other two
countries. While the positive effect of firm size on the cash/net-asset ratio during the
crisis is evident in all three countries, it is only significant in Brazil. The coefficient
estimate on the Mexican crisis dummy is smallest in absolute value for Mexico, so that
cash holdings decrease the least in the crisis country. Interestingly, none of the variables
has a significant effect on cash holdings in Mexico.
5.2 Results for Cash Holdings before and during the Brazilian crisis
Table 3 reports the results of the cash holdings regressions for the Brazilian
crisis. One additional variable is free float (a measure of ownership concentration) that is
only available for Brazilian firms. In order to make the regressions directly comparable,
the sample of Brazilian firms included in the pooled model and in the Brazil-only model
includes only firms with non-missing values of free float during at least one pre-crisis and
at least one crisis year.
15
Unlike the Mexican crisis, the Brazilian crisis (as measured by the crisis dummy
variable) has no significant effect on cash holdings. This suggests the effects of higher
agency costs and lower liquidity are roughly equal and offsetting in this crisis. Also
unlike the Mexican period, few of the remaining coefficient estimates are significant in
the pooled Brazilian crisis model.
In particular, firm size and dividends have
insignificant effects on cash in both the pre-crisis and during-crisis periods. A possible
explanation for the small t-statistics on many of the coefficients is that as Latin American
markets became more integrated with world markets during the 1990s due to financial
liberalization and free trade agreements, observable firm characteristics became relatively
less important components of asset allocation behavior relative to unobserved governance
characteristics (which are subsumed into our fixed effect term).
Although the effects of most financial characteristics are insignificant in the
Brazilian crisis, there are a few non-zero coefficients. Net working capital has a negative
effect on cash holdings that is of similar magnitude both before and during the crisis.
Cross listed firms hold less cash prior to the crisis but increase their cash holdings during
the crisis. Similar to the size effect documented during the Mexican crisis, the finding on
cross-listing is consistent with potential governance problems. Lins, Strickland, and
Zenner (2005) document that greater access to capital markets is an important benefit to
firms from developing markets that list on a U.S. exchange. This suggests that ADRs
have greater financing alternatives and should have less need to hold cash than nonADRs. Operating profit has a positive effect in the pre-crisis period that becomes even
greater during the crisis. This may be because managers have a greater opportunity to
expropriate additional firm wealth during the crisis. However, it could also suggest an
16
increased precautionary motive for holding cash by firms that have sufficient financial
resources to do so.
Focusing next on the country-specific results, we find that as in the pooled model
the Brazilian crisis dummy has an insignificant effect on cash holdings in each of the
three countries and that the effect of size is insignificant both before and during the crisis.
The results for the full sample mirror those for Brazil in that, among Brazilian firms, net
working capital and cross-listing have negative effects on cash in the pre-crisis period
while operating profit has a positive effect. Leverage is negatively associated with cash
in Argentina and Mexico prior to the crisis, suggesting that debt serves as a substitute for
cash. Higher firm profitability is associated with higher cash holdings in Argentina
during the Brazilian crisis. This is apparently the source of the full sample result.
Among Brazilian firms, free float is insignificant in the pre-crisis period,
indicating that ownership concentration has no effect on the cash/net-asset ratio.6
However, the negative coefficient on the free float*crisis interaction term indicates that
firms with more dispersed ownership hold less cash during the crisis than firms with
more concentrated ownership. The magnitude of the estimate implies that a one percent
increase in ownership concentration is predicted to increase the cash to net asset ratio by
0.56 percent ((-0.4730 – 1.0280)*0.375). This result supports the liquidity hypothesis
that firms with more dispersed ownership should have less need to hold cash than firms
with more concentrated ownership during a crisis. As mentioned earlier, higher cash
associated with highly concentrated ownership concentration could also reflect a
6
The Brazilian model was estimated without free float for comparability with the other countries, but the
coefficient estimates were virtually identical and so were not reported.
17
governance problem if the block owners are mainly insiders. Our Tobin’s Q models will
examine the valuation effects of concentrated ownership.
To summarize, the results of the cash models indicate that Latin American firm
managers alter their cash holdings differently in the Mexican and Brazilian crises. Cash
holdings fall during the Mexican crisis but not during the Brazilian crisis. While leverage
and dividends do not have significantly different effects on cash between crisis and noncrisis periods, the effects of other characteristics do differ between the crises. Larger
firms increase their holdings of cash more than smaller firms during the Mexican crisis,
while cross-listed firms increase their cash holdings more than non-cross-listed firms
during the Brazilian crisis. Both of these effects are consistent with potential agency
conflicts, since these types of firms should have less need to hold cash. In addition,
Brazilian firms with more concentrated ownership increase their holdings of cash during
the Brazilian crisis. The next section seeks to develop a more detailed understanding of
the cash holdings results by investigating the determinants of firm value.
5.3 Results for Tobin’s Q before and during the Mexican crisis
In the previous section we described the effects of financial crises on the
determinants of cash holdings. Because many of these effects are consistent with both
agency and liquidity explanations for cash holdings, we attempt in this section to further
investigate the implications of the findings by estimating Tobin’s Q both before and
during these crises. Tobin’s Q is a standard measure of firm value as it measures the
market value of the firm relative to the book value of assets.7 In essence, Tobin’s Q can
7
In this paper, we define Tobin’s Q as market value relative to net assets rather than total assets so that we
can control for total assets in the regression without the left and right hand sides being perfectly correlated.
18
be interpreted as the ratio of the future value of the firm to its current value.
Consequently, it will rise as the market value of the firm increases and fall as the book
value of assets increases. In particular, one may witness an increase in Tobin’s Q even if
market value is falling in the event that book value is falling even faster. Because of this
possibility, we also estimate regression models in which the market value of the firm
(unscaled by assets) is the dependent variable and discuss these results in the next
section.
The Tobin’s Q results for the Mexican crisis period are reported in Table 4.
According to the pooled sample results in the first column, the Mexican crisis has no
statistically distinguishable effect on Tobin’s Q. Although this may be an indication that
crises do not reduce market value, this seems unlikely given the devastating economic
effects of crises on developing markets. Hence, a more probable interpretation might be
that both market and book value fall by similar percentages during the crisis such that
Tobin’s Q remains roughly constant. The market value results we present subsequently
indicate that market value decreases during the Mexican crisis, as expected.
The positive coefficient on cash holdings indicates firms that hold more cash have
higher values of Tobin’s Q in the pre-crisis period. The magnitude of the coefficient
implies that in the pre-crisis period, a one percent increase in cash is associated with a
0.26 percent increase in Tobin’s Q. (Note that the cash coefficient divided by the sample
average of Tobin’s Q in the pre-crisis period is 0.2299 / 0.869 = 0.26). This positive
association between cash and value is consistent with studies that find that firms with
more growth opportunities hold more cash. However, as shown later when the unscaled
It turns out, however, that the results are essentially the same regardless of whether the denominator is
expressed as net or total assets.
19
version of market value is the dependent variable, the effect of cash holdings on market
value is negative during the Mexican crisis.
The remaining results in the first column suggest that in the pre-crisis period,
Tobin’s Q is lower for larger firms and higher for cross-listed firms. The value premium
for small firms could be attributed to more growth opportunities as well as higher agency
costs of large firms. The premium for cross-listed firms is likely due to greater visibility
and more stringent governance controls for ADR listings.
Under crisis conditions
investors appear to display a favorable view toward dividends but an aversion to highly
leveraged firms and those with high working capital. Since working capital is typically a
substitute for cash, the negative effect could reflect collection concerns associated with
current assets due to potentially more bad debt. Interestingly, the interaction between the
crisis dummy and ADR is negative, so that the effect of cross listing becomes less
positive during than before the crisis. Elkinawy (2005) finds that U.S.-based Latin
American mutual fund managers increased their holdings of ADRs during the Asian
crisis, particularly closed-end funds that are less liquidity-constrained relative to openend funds. However, investors do not increase their holdings of cross-listed firms during
the Brazilian crisis. Thus, the negative effect we document is consistent with potentially
increased selling of ADRs by liquidity-constrained foreign investors who are relatively
highly represented among the owners of these securities.
The country-specific results in the second through fourth columns of Table 4
illustrate that the Mexican crisis has no significant effect on Tobin’s Q in any of the
sample countries. However, cash holdings are positively associated with Tobin’s Q in
both Argentina and Brazil prior to the crisis, with the effect in Brazil becoming even
20
more positive during the crisis. The result for Brazil is suggestive of a liquidity premium
for firms with high growth opportunities during this crisis.
An inspection of the
remaining country-specific coefficient estimates reveals that the value premium for small
firms increases in Argentina during the Mexican crisis, while the negative effects of
leverage and net working capital during the crisis in the full sample are driven by
Brazilian firms. Finally, the value premium to dividend-paying firms during the Mexican
crisis is observed only for Brazil (but arises in the pooled results as well), while unlike
the pooled model, cross-listed Argentinean firms experience higher Tobin’s Q values
during the crisis.
5.4 Results for Tobin’s Q before and during the Brazilian crisis
The first column of Table 5 shows that unlike the Mexican crisis, Tobin’s Q
declines during the Brazilian crisis, but larger firms increase in value relative to the precrisis period. Also unlike the Mexican crisis, firms with high leverage and high net
working capital have relatively higher Tobin’s Q values during the Brazilian crisis, while
dividends have no effect. The positive effect of leverage may reflect the role of debt as a
mechanism for disciplining managerial behavior (Harvey, Lins, and Roper, 2004).
However, similar to the Mexican crisis cross-listed firms have lower Tobin’s Q values
during the crisis.
According to the country-level regression results, the Brazilian crisis significantly
reduces Tobin’s Q in Argentina and Mexico but not Brazil. High cash holdings among
Brazilian and Mexican firms are associated with higher valuations prior to the crisis, but
there are differences in the effect of cash on firm value between the two countries during
the crisis. In Brazil, cash has a greater positive effect on Tobin’s Q during the crisis, but
21
in Mexico the effect of cash is negative during the crisis. This might suggest that
concerns about managerial expropriation are relatively greater in Mexico while liquidity
is of relatively greater concern in Brazil. Brazilian firms that pay dividends have lower
Tobin’s Q values during the crisis than before. In all three countries large firms have
relatively higher valuations during the Brazilian crisis than before.
Column 3 indicates that the effect of the free float variable in Brazil is zero in
both the pre-crisis and crisis periods, indicating no market-to-book devaluation of
Brazilian firms with more concentrated ownership. It may be that the liberalization
reforms during the 1990s loosened the association between block holdings and inside
ownership among Brazilian firms.
Dittmar and Mahrt-Smith (2007) find that the accumulation of cash is only
problematic for poorly-governed firms, since these firms tend to dissipate cash on less
profitable investments. Although the present data do not indicate how cash is spent, we
attempt to examine the governance issue more closely by testing the cash-value link with
respect to ownership concentration. To do this, we create a three-way interaction among
cash holdings, free float, and the crisis dummy variable.
The coefficient on this
interaction term reflects the pre-to during-crisis difference in the effect of cash on firm
value for firms with differing levels of ownership concentration. In Column 4, the
insignificant coefficient on the cash*free-float interaction suggests that the effect of cash
on firm value is independent of ownership concentration prior to the Brazilian crisis.
However, the negative coefficient on the cash*free-float*crisis interaction term suggests
that during the crisis, cash has a value-reducing effect for firms with more dispersed
ownership. Although one might expect that if the block owners are insiders they are
22
more likely to expropriate from minority shareholders, La Porta, et al. (2002) find that
higher insider cash flow ownership is associated with higher valuation, particularly in
countries with low investor protection. They argue that because expropriation is costly,
higher insider cash flow ownership should actually lead to less expropriation. Another
possibility is that firms with more concentrated ownership are easier to monitor, a
characteristic particularly valuable during a turbulent time period. Finally, firms with
more dispersed ownership should have less need to hold cash for liquidity reasons. This
suggests that to the extent that these firms do hold cash, the cash is more likely to be used
contrary to shareholder interests during a crisis than before.
In summary, the firm value models indicate that cross-listed firms have lower
Tobin’s Q values in both the Mexican and Brazilian crisis while Brazilian firms with
more cash have higher Tobin’s Q values in both crises. However, similar to the cash
holdings models, we also observe many differences in the determinants of Tobin’s Q
between the Mexican and Brazilian crisis. Tobin’s Q falls during the Brazilian crisis but
not during the Mexican crisis. We also find that firms with higher leverage and higher
net working capital have lower values during the Mexican crisis, while leverage and net
working capital are value-enhancing during the Brazilian crisis. Finally, our analysis of
ownership concentration reveals that the degree of block ownership affects the link
between cash and firm value, suggesting that governance concerns become more
significant during crises.
7. Sensitivity Tests
We conduct a number of sensitivity tests to check the robustness of our results to
alternative specifications. Table 6 presents the results of regression models that use
23
market value (defined market capitalization plus short term and long term debt) rather
than Tobin’s Q as the dependent variable. The reason for doing this is to examine the
direct effects of the control variables on a single measure of value rather than on a ratio
of value measures. The results for the Mexican crisis in the first column indicate that as
expected, market value falls significantly during the crisis. Cash holdings have no effect
on market value before the crisis, but have a negative effect during the crisis. This is
consistent with the notion that investors penalize firms for accumulating more cash
during a crisis, when agency costs are likely to be higher. Although we also observe a
drop in firm value during the Brazilian crisis relative to the pre-crisis period, we do not
observe a significant effect of cash holdings on market value either before or during the
crisis. Thus, the positive effect of cash on Tobin’s Q in Table 5 is consistent with the fact
that cash is more valuable for firms with higher growth opportunities rather than simply
for firms with higher market values. The third and fourth columns in Table 6 indicate
that ownership concentration has no effect on firm value, with or without the interaction
with cash. These findings suggest that governance considerations are relatively more
important in assessing differences in the determinants of a firm’s growth opportunities
rather than in its market value during a crisis.
The next sensitivity test involves redefining the Brazilian crisis period as a twoyear rather than a one-year window. The rationale for the one-year window utilized in
our baseline regressions is that the crisis officially occurred in January 1999, so that
responses to the crisis by local managers based on our year-end data should be most
apparent in 1999. However, because it is possible that some of the crisis impacts were
either felt or anticipated in 1998, we re-estimate the Brazilian crisis cash holdings and
24
Tobin’s Q regressions with the crisis period lengthened to two years (1998-1999) and
report the results in Table 7. The results for cash holdings are very similar to those with
the one-year window except that the interaction terms are all insignificant. In particular,
the crisis still has no effect on cash holdings and, likewise, size has no effect either before
or during the crisis. When free float is added and the analysis is restricted to Brazil, we
again observe no crisis effect. Free float also has no effect before the crisis, but it does
have a negative effect on cash holdings during the crisis, as was observed when a oneyear crisis window was employed. The results for Tobin’s Q with the two-year window
are generally similar as those with the one-year window except that more variables are
now significant.
In unreported models, we also estimate industry-specific regressions for cash
holdings, where industries have been grouped into two very broad categories
(“manufacturing” and “service”) so that there are sufficient numbers of observations in
each industry category to obtain meaningful regression results. The following industries
are classified as “manufacturing” industries: agriculture, mining, utilities, transportation,
construction, and manufacturing (i.e., the two-digit NAIC code specifically designated as
manufacturing). The following industries are grouped together under the title of “service”
industries: wholesale, retail, information, real estate, professional, management,
administration, education, health, arts, accommodations, and all “other” industries not
previously listed. Financial firms continue to be excluded from our sample.
For the Mexican crisis, we find a negative effect of the crisis on cash holdings, a
positive effect of size, and a positive size-crisis interaction coefficient in manufacturing
but not service industries. The effects of net working capital and dividends also differ
25
between pre-crisis and during crisis periods for manufacturing but not service industries.
For the Brazilian crisis, the effects of firm characteristics on cash holdings are similar in
the pre-crisis period for manufacturing and service industries, but the during-crisis effects
are different in the two industry groupings. Specifically, the coefficient on the crisis
dummy is negative and the coefficients on the crisis-leverage interaction and the crisisnet working capital interaction are positive for service but insignificant for manufacturing
industries. One lesson that therefore emerges from the industry analysis is that the effects
of a crisis differ across industries, but that it is difficult to generalize the reactions of a
particular industry across different crises.
In other unreported models, we alter the specifications by defining cash holdings
as the ratio of cash to total assets instead of net assets. For each crisis, the results are
virtually identical regardless of how cash holdings are defined. The one exception is that
when the cash to total asset ratio is used, the interaction term between operating profit
and the crisis dummy becomes insignificant instead of positive. As another robustness
check, we remove assets from the denominator entirely and simply use absolute cash
holdings as the dependent variable. Analyzing the pooled sample of firms from all three
countries, the results for each crisis are again almost entirely unaffected.
Finally, to account for the time series nature of the data, we also estimate cash
holdings regressions (with cash to net assets as the dependent variable) with explicit time
trends for each crisis. For the Mexican crisis period we specify a cubic time trend
because of the shapes of the country averages in Figure 1. Although the time variables
are all significant indicating a cubic time trend does exist, the other coefficient estimates
are not notably affected. For the Brazilian crisis we use only a quadratic trend because
26
the time plots in Figure 2 are simpler (they appear almost linear) and because the analysis
period consists of just four years. The time variables are insignificant and the other
coefficient estimates are unaffected. Thus, our results appear robust to time controls.
8. Conclusion
Using panel data on non-financial firms in Argentina, Brazil, and Mexico during
the 1990s, we investigate whether financial crises in emerging markets alter the
determinants of corporate cash holdings and firm value. We find that Latin American
firm managers alter their cash holdings differently in the Mexican crisis of 1994-1995
than in the Brazilian crisis of 1999. Our Tobin’s Q regression models indicate that the
determinants of firm value also differ across crises. Overall, our findings suggest that the
effect of financial crises on firms is not uniform, even for crises originating in the same
geographic region.
The results we document are consistent with both decreases in
liquidity as well as increases in agency conflicts during crises.
Because more explicit firm-level governance measures are unavailable in these
data, a definitive disentangling of transaction and precautionary motives for holding cash
from agency explanations awaits further research. A recent Wall Street Journal article
indicates that foreign investors have quadrupled their ownership stake in emerging
markets over the past five years, and these investors are becoming more active in the
governance of these firms (Slater, 2007). As a result, the role of governance in asset
allocation decisions as emerging markets become increasingly integrated with world
markets is a high priority for future work.
27
Appendix. Definition of firm characteristics
The sample consists of all non-financial publicly traded firms with available data in Argentina,
Brazil, and Mexico. The source for firm financial data is Economatica. The ADR program is
obtained from the Bank of New York. All variables are measured annually in U.S. dollars as of
December 31.
Cash Holdings
(Cash & ST investments)/(Net assets),
where net assets = Total assets – Cash & ST investments
Tobin’s Q
(Market capitalization + ST debt + LT debt)/(Net assets)
Size
Total assets
Leverage
Total liabilities/Total assets
Net Working
Capital
(Current assets – Cash & ST investments – Current liabilities)/(Net assets)
Operating Profit
(Earnings before interest, taxes, depreciation, & amortization)/(Net assets)
Dividend Dummy
Dummy variable = 1 if the firm’s dividend yield was greater than zero
during the year and 0 otherwise
ADR Program
Dummy variable = 1 if firm has an American Depositary Receipt in a
given year and 0 otherwise
Free Float
Percentage of firm not owned by blockholders, defined as shareholders
who own five percent or more of the shares (only available for Brazilian
firms after 1995).
29
References
Arslan, O., C. Florackis, and A. Ozkan, 2006, The role of cash holdings in reducing investmentcash flow sensitivity: evidence from a financial crisis period in an emerging market.
Emerging Markets Review 7, 320-338.
Chang, K. and A. Noorbakhsh, 2006, Corporate cash holdings, foreign direct investment, and
corporate governance. Global Finance Journal 16, 302-316.
Colquitt, L.L., Sommer, D.W. and N.H. Godwin, 1999, Determinants of cash holdings by
property-liability insurers. Journal of Risk and Insurance 66 (3), 401-415.
Dittmar, A., and J. Mahrt-Smith, 2007, Corporate governance and the value of cash holdings.
Journal of Financial Economics 83, 599-634.
Dittmar, A., J. Mahrt-Smith, and H. Servaes, 2003, International corporate governance and
corporate cash holdings. Journal of Financial and Quantitative Analysis 38, 111-133.
Elkinawy, S., 2005, Mutual fund preferences for Latin American equities surrounding financial
crises. Emerging Markets Review 6, 211-237.
Ferreira, M.A., and A.S. Vilela, 2004, Why do firms hold cash? Evidence from EMU countries.
European Financial Management 10, 295-319.
Harford, J., S.A Mansi, and W.F. Maxwell, 2005, Corporate governance and firm cash holdings.
Working paper, University of Washington, Virginia Tech, and University of Arizona.
Harvey, C.R., K.V. Lins, and A.H. Roper, 2004, The effect of capital structure when expected
agency costs are extreme. Journal of Financial Economics 74, 3-30.
Jensen, M.C., 1986, Agency costs of free cash flow, corporate finance, and takeovers. American
Economic Review 76, 323-329.
Kalcheva, I. and K.V. Lins, 2006, International evidence on cash holdings and expected
managerial agency problems. Working paper, University of Utah.
Kim, C., D.C. Mauer, and A.E. Sherman, 1998, The determinants of corporate liquidity: theory
and evidence. Journal of Financial and Quantitative Analysis 33, 335-359.
Kusnadi, Y., 2005, Corporate governance mechanisms and corporate cash holdings. Working
paper, Hong Kong University of Science and Technology.
La Porta, R., F. Lopez-de Silanes, A. Shleifer, and R. Vishny, 2002, Investor protection and
corporate valuation. Journal of Finance 57, 1147-1170.
Lins, K.V., D. Strickland, and M. Zenner, 2005, Do non-U.S. firms issue equity on U.S. stock
exchanges to relax capital constraints? Journal of Financial and Quantitative Analysis 40,
109-133.
30
Mikkelson, W.H., and M.M. Partch, 2003, Do persistent large cash reserves hinder
performance? Journal of Financial and Quantitative Analysis 38, 275-294.
Niskanen, M. and J. Niskanen, 2007, Cash holdings in Finnish SMEs.
University of Kuopio, Finland.
Working paper,
Opler, T., L. Pinkowitz, R. Stulz, and R. Williamson, 1999, The determinants and implications of
corporate cash holdings. Journal of Financial Economics 52, 3-46.
Ozkan, A. and N. Ozkan, 2004, Corporate cash holdings: An empirical investigation of UK
companies. Journal of Banking and Finance 28 (9), 2103-2134.
Pinkowitz, L., R. Stulz, and R. Williamson, 2006, Does the contribution of corporate cash
holdings and dividends to firm value depend on governance? A cross-country analysis.
Journal of Finance, forthcoming.
Pinkowitz, L. and R. Williamson, 2001, Bank power and cash holdings: Evidence from Japan.
Review of Financial Studies 14 (4), 1059-1082.
Slater, J., Emerging market investors seek more influence. Wall Street Journal, April 11, 2007,
C3.
31
Figure 1: Average Cash Holdings By Country in
Mexican Crisis Analysis Period
0.2
Assets
Cash as Percentage of Net
0.25
0.15
Brazil
Mexico
0.1
Argentina
0.05
0
1990
1991
1992
1993
1994
1995
Year
Notes: Cash holdings are defined as the ratio of cash and short term investments to net assets. Net assets are defined as total assets minus cash
and short term investments (measured in U.S. dollars). The graphs display average values of cash holdings only among firms with available data
for all financial variables used in regressions in at least one pre-crisis and at least one crisis year. The series for Argentina and Mexico begin in
1991 instead of 1990. The number of observations for Argentina is 0, 6, 16, 15, 25, and 37 in the years 1990-95, respectively. The number of
observations for Brazil is 163, 164, 177, 179, 217, and 230 in the years 1990-95, respectively. The number of observations for Mexico is 0, 17, 2,
39, 56, and 64 in the years 1990-95, respectively.
0.25
0.2
of Net Assets
Cash Holdings as Percentage
Figure 2: Average Cash Holdings by Country in
Brazilian Crisis Analysis Period
Brazil
0.15
Mexico
0.1
Argentina
0.05
0
1996
1997
1998
1999
Year
Notes: Cash holdings are defined as the ratio of cash and short term investments to net assets. Net assets are defined as total assets minus cash
and short term investments (measured in U.S. dollars). The graphs display average values of cash holdings only among firms with available data
for all financial variables used in regressions in at least one of the pre-crisis and one of the during-crisis years. The series for Argentina and
Mexico begin in 1991 instead of 1990. The number of observations for Argentina is 34, 34, 33, and 42 in the years 1996-99, respectively. The
number of observations for Brazil is 165, 150, 193, and 253 in the years 1996-99, respectively. The number of observations for Mexico is 70, 94,
102, and 108 in the years 1996-99, respectively.
32
Tobin's Q
Figure 3: Average Tobin's Q by Country in
Mexican Crisis Analysis Period
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Mexico
Argentina
Brazil
1990
1991
1992
1993
1994
1995
Year
Notes: Tobin’s Q is the sum of market capitalization, short-term debt and long-term debt all divided by net assets (measured in U.S. dollars). The
graphs display average values of Tobin’s Q only among firms with available data for all financial variables used in regressions in at least one of
the pre-crisis and one of the during-crisis years. The series for Argentina and Mexico begin in 1991 instead of 1990. The number of observations
for Argentina is 0, 6, 16, 15, 25, and 37 in the years 1990-95, respectively. The number of observations for Brazil is 163, 164, 177, 179, 217, and
230 in the years 1990-95, respectively. The number of observations for Mexico is 0, 17, 2, 39, 56, and 64 in the years 1990-95, respectively.
Tobin's Q
Figure 4: Average Tobin's Q by Country in
Brazilian Crisis Analysis Period
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Mexico
Argentina
Brazil
1996
1997
1998
1999
Year
Notes: Tobin’s Q is the sum of market capitalization, short-term debt and long-term debt all divided by net assets (measured in U.S. dollars). The
graphs display average values of Tobin’s Q only among firms with available data for all financial variables used in regressions in at least one of
the pre-crisis and one of the during-crisis years. The series for Argentina and Mexico begin in 1991 instead of 1990. The number of observations
for Argentina is 34, 34, 33, and 42 in the years 1996-99, respectively. The number of observations for Brazil is 165, 150, 193, and 253 in the
years 1996-99, respectively. The number of observations for Mexico is 70, 94, 102, and 108 in the years 1996-99, respectively.
33
Table 1
Descriptive Statistics for Mexican and Brazilian Analysis Periods
Sample SubPeriod
Variables
ANALYSIS PERIOD FOR MEXICAN CRISIS
Pre-Crisis
Crisis
(1990 – 1993)
(1994 – 1995)
Mean/
Std.
Mean/
Std.
Median
Dev.
Median
Dev.
Cash Holdings
0.075 / 0.031
0.112
0.114 / 0.041
0.464
0.038**
0.026
0.138 / 0.052
0.449
0.087 / 0.035
0.176
-0.051**
0.027
Tobin’s Q
0.778 / 0.383
5.481
0.786 / 0.613
0.768
0.008
0.969
0.869 / 0.680
0.778
0.798 / 0.647
0.610
-0.071
0.107
1108.9 / 191.6
4510.4
1701.6 / 375.4
6805.3
592.6*
0.051
1900.7 / 554.8
6038.7
1437.2 / 371.2
3879.7
-463.5
0.158
0.180 / 0.146
0.157
0.208 / 0.170
0.167
0.028***
0.002
0.254 / 0.236
0.191
0.268 / 0.238
0.264
-0.014
0.283
-0.019 / -0.008
0.181
-0.040 / -0.019
0.471
-0.021
0.253
-0.024 / -0.011
0.384
-0.031 / -0.018
0.307
-0.007
0.761
0.087 / 0.077
0.127
0.094 / 0.092
0.105
0.007
0.297
0.107 / 0.107
0.143
0.115 / 0.118
0.109
0.007
0.363
Dividend
Dummy
0.585 / 1.0
0.493
0.553 / 1.0
0.498
-0.032
0.235
0.584 / 1.0
0.493
0.526 / 1.0
0.500
-0.058*
0.052
ADR Dummy
0.021 / 0.0
0.142
0.135 / 0.0
0.342
0.115***
0.000
0.241 / 0.0
0.428
0.243 / 0.0
0.430
0.002
0.937
Free Float
----
----
----
----
----
----
0.425 / 0.420
0.222
0.375 / 0.347
0.231
-0.050***
0.004
Number of
Firm-year obs.
778
Size ($MM)
Leverage
Net Working
Capital
Operating
Profit
629
CRISISPRE CRISIS
Diff. in
pmeans
value
ANALYSIS PERIOD FOR BRAZILIAN CRISIS
Pre-Crisis
Crisis
CRISIS(1996 – 1998)
(1999)
PRE CRISIS
Mean/
Std.
Mean/
Std.
Diff. in
pMedian
Dev.
Median
Dev.
means
value
875
403
Notes: For each crisis, the sample consists of non-financial firms in Argentina, Brazil, and Mexico with available values of all variables during at least one pre-crisis year and one crisis year. The Mexican
analysis period is from December 1990 to December 1995, where the Mexican crisis spans the years 1994-1995. The Brazilian analysis period is from December 1996 to December 1999, where the Brazilian
crisis spans the year 1999. Cash holdings are defined as the ratio of cash and short-term investments to net assets, where net assets are equal to total assets minus cash and short-term investments. Tobin’s Q is
defined as the sum of market capitalization, short-term debt, and long term debt all divided by net assets. Size is measured by total assets. Leverage is the ratio of total liabilities to total assets. Net working
capital is the difference between non-cash current assets and current liabilities all divided by net assets. EBITDA is earnings before interest, taxes, depreciation, and amortization divided by net assets. Dividend
dummy is a dummy variable equal to one if the firm’s dividend yield is greater than zero in a given year and zero otherwise. ADR Dummy is a dummy variable equal to one if the firm has an American
Depositary Receipt in a given year and zero otherwise. Free float is the percentage of shares outstanding not owned by blockholders (where a blockholder is defined as someone who owns at least 5% of shares
outstanding). Free float is only available for Brazilian firms and only during the Brazilian crisis analysis period. The number of observations on free float is 508 in the pre-Brazilian crisis period and 253 in the
Brazilian crisis period. All variables are measured annually in U.S. dollars as of December 31 of the given year. *** (**) (*) – Difference in means is significant at the 1% (5%) (10%) level
34
Table 2
Fixed Effects Models of Cash Holdings before and during the Mexican Crisis,
December 1990 – December 1995
Dependent variable: ln(Cash & ST Investments/Net Assets)
Pooled
Argentina
Coeff.
p-value
Coeff.
p-value
Size (natural log)
0.3791***
0.000
-0.5410
0.504
Leverage
-0.5832
0.247
1.4994
0.519
Net Working Capital
-1.2620***
0.001
-2.5360
0.253
Operating Profit
1.0945***
0.013
8.7071*
0.065
Dividend Dummy
0.2623**
0.017
-0.0854
0.896
ADR Dummy
0.2954
0.486
0.5062
0.697
Crisis Dummy
-2.8917***
0.003
-6.4235
0.234
Size*Crisis
0.1516***
0.003
0.3787
0.184
Leverage*Crisis
-0.4472
0.374
-2.0843
0.397
NWC*Crisis
0.8653**
0.025
3.7404
0.109
Operating Profit*Crisis
0.6449
0.331
-4.7104
0.350
Dividend*Crisis
-0.2043
0.211
0.1694
0.839
ADR Dummy*Crisis
-0.4686
0.235
-0.7263
0.471
Constant
-15.7523***
0.000
0.9573
0.952
Within R2
Overall R2
Number of firms
No. of firm-year obs.
0.067
0.013
346
1,407
0.296
0.196
39
99
Brazil
Coeff.
p-value
0.3683***
0.001
-0.5081
0.366
-1.3295***
0.002
0.9203**
0.049
0.2449**
0.044
0.8203
0.547
-2.9997***
0.007
0.1529***
0.008
-0.0313
0.957
0.9611**
0.028
0.8740
0.232
-0.1610
0.395
-0.7900
0.566
-15.6265**
0.000
0.070
0.006
241
1,130
Mexico
Coeff.
p-value
0.3102
0.425
0.5140
0.713
0.3391
0.763
2.4604
0.287
0.3563
0.210
-0.2694
0.437
-1.6544
0.522
0.0969
0.454
-1.1035
0.246
0.7690
0.470
-0.8225
0.713
-0.4809
0.118
0.2108
0.528
-14.0535*
0.077
0.179
0.108
66
178
Notes: The sample consists of non-financial firms in Argentina, Brazil, and Mexico with available values of all variables during at
least one pre-crisis year and at least one crisis year. The Mexican crisis period is defined as the years 1994-1995. The coefficients on
the interaction terms represent the difference in the effect of the variable on cash holdings between the crisis and pre-crisis period.
Variable definitions are provided in the Appendix. *** (**) (*) – Coefficient estimate is significant at the 1% (5%) (10%) level.
35
Table 3
Fixed Effects Models of Cash Holdings before and during the Brazilian Crisis,
December 1996 – December 1999
Dependent variable: ln(Cash & ST Investments/Net Assets)
Pooled
Argentina
Coeff.
p-value
Coeff.
p-value
Size (natural log)
0.1057
0.371
-0.7461
0.134
Leverage
-0.5466
0.166
-3.1315*
0.076
Net Working Capital
-0.8924***
0.000
-0.7626
0.569
Operating Profit
1.9658***
0.000
-0.6901
0.470
Dividend Dummy
0.0572
0.596
0.1331
0.678
ADR Dummy
-0.7220***
0.003
0.7493
0.493
Free Float
------------Crisis Dummy
-0.1040
0.904
-1.0869
0.698
Size*Crisis
-0.0235
0.604
0.0118
0.938
Leverage*Crisis
0.4292
0.229
0.2395
0.876
NWC*Crisis
0.3795
0.193
-0.0980
0.937
Operating Profit*Crisis
1.1363*
0.077
7.0224**
0.024
Dividend*Crisis
-0.1182
0.434
0.0893
0.866
ADR Dummy*Crisis
0.3152**
0.046
0.3523
0.404
Free Float*Crisis
------------Constant
-9.7844***
0.000
7.3722
0.437
Within R2
Overall R2
Number of firms
No. of firm-year obs.
0.129
0.024
403
1,278
0.224
0.061
42
143
Brazil
Coeff.
p-value
0.2458
0.161
0.0363
0.944
-0.8627***
0.000
2.6086***
0.000
0.0588
0.723
-1.1523***
0.003
-0.4730
0.332
0.4022
0.755
-0.0254
0.697
0.3773
0.432
0.5292
0.175
0.6771
0.452
-0.2313
0.339
0.3750
0.179
-1.0280**
0.019
-12.7157**
0.000
0.189
0.035
253
761
Mexico
Coeff.
p-value
-0.0080
0.967
-1.3753*
0.089
0.5083
0.365
2.3607**
0.012
0.0809
0.523
-0.2402
0.341
-------0.6558
0.581
0.0211
0.741
-0.3933
0.489
0.1365
0.792
0.9578
0.253
-0.0990
0.604
0.1149
0.523
-------7.2634*
0.051
0.106
0.006
108
374
Notes: The sample consists of non-financial firms in Argentina, Brazil, and Mexico with available values of all variables during at
least one pre-crisis year and one crisis year. The Brazilian crisis period is defined as the year 1999. The coefficients on the interaction
terms represent the difference in the effect of the variable on cash holdings between the crisis and pre-crisis period. Variable
definitions are provided in the Appendix. *** (**) (*) – Coefficient estimate is significant at the 1% (5%) (10%) level
36
Table 4
Fixed Effects Models of Tobin’s Q before and during the Mexican Crisis,
December 1990 – December 1995
Dependent variable: Tobin’s Q = (Market Capitalization + ST Debt +LT Debt)/Net Assets)
Pooled
Argentina
Brazil
Coeff.
p-value
Coeff.
pCoeff.
p-value
value
Size (natural log)
-6.6136***
0.000
-0.2738** 0.042
-7.6707***
0.000
Leverage
0.4995
0.738
-0.3570 0.348
0.5409
0.746
Net Working Capital
0.8247
0.478
-0.4128 0.259
-0.1806
0.890
Operating Profit
-1.8666
0.153
3.3445*** 0.000
-2.1938
0.116
Dividend Dummy
-0.2621
0.422
-0.1987* 0.068
-0.2835
0.434
ADR Dummy
2.9123**
0.021
-0.6229*** 0.008
2.5329
0.531
Cash Holdings
0.2299**
0.024
0.0889*** 0.001
0.1892*
0.093
Crisis Dummy
3.7990
0.235
1.0589 0.275
3.8356
0.283
Size*Crisis
0.0329
0.830
-0.0852* 0.073
0.1338
0.437
Leverage*Crisis
-3.3863**
0.023
0.3717 0.363
-3.1795*
0.068
NWC*Crisis
-3.5629***
0.002
0.2140 0.578
-2.6544**
0.043
Operating Profit*Crisis
0.0816
0.968
-0.5431 0.567
0.7796
0.724
Dividend*Crisis
1.0907**
0.028
0.1062 0.439
1.1144*
0.055
ADR Dummy*Crisis
-3.5593***
0.002
0.3662* 0.058
-2.2409
0.584
Cash Holdings*Crisis
0.1085
0.380
-0.0428 0.276
0.2316*
0.095
Constant
130.178***
0.000
7.0294*** 0.008 148.786***
0.000
Within R2
Overall R2
Number of firms
No. of firm-year obs.
0.319
0.012
346
1,407
0.830
0.002
39
99
0.370
0.019
241
1,130
Mexico
Coeff.
p-value
0.3248
1.2584
1.0075
7.4466***
-0.2931
0.0527
0.1441
-0.7549
0.0172
0.3977
-0.2327
-1.8675
0.1832
-0.2058
-0.0144
-4.9483
0.456
0.106
66
178
Notes: The sample consists of non-financial firms in Argentina, Brazil, and Mexico with available values of all variables during at
least one pre-crisis year and one crisis year. The Mexican crisis period is defined as the years 1994-1995. The coefficients on the
interaction terms represent the difference in the effect of the variable on Tobin’s Q between the crisis and pre-crisis period. Variable
definitions are provided in the Appendix. *** (**) (*) – Coefficient estimate is significant at the 1% (5%) (10%) level
37
0.255
0.217
0.228
0.000
0.172
0.835
0.150
0.724
0.857
0.570
0.764
0.253
0.431
0.399
0.885
0.396
Table 5
Fixed Effects Models of Tobin’s Q before and during the Brazilian Crisis,
December 1996 – December 1999
Dependent variable: Tobin’s Q = (Market Capitalization + ST Debt +LT Debt)/Net Assets)
Pooled
Argentina
Brazil
Size (natural log)
Leverage
Net Working Capital
Operating Profit
Dividend Dummy
ADR Dummy
Cash Holdings
Free Float
Cash Holdings*Free Float
Crisis Dummy
Size*Crisis
Leverage*Crisis
NWC*Crisis
Operating Profit*Crisis
Dividend*Crisis
ADR Dummy*Crisis
Cash Holdings*Crisis
Free Float*Crisis
Cash*Free Float*Crisis
Constant
Within R2
Overall R2
Number of firms
No. of firm-year obs.
Coeff.
-0.2533***
-0.0894
-0.9986***
0.1742
0.1124**
-0.0656
0.0639***
-------2.2919***
0.0877***
0.4782***
0.4200***
1.3781***
0.0435
-0.1810**
-0.0321
------6.3547***
0.315
0.040
403
1,278
p-value
0.000
0.626
0.000
0.360
0.025
0.562
0.000
------0.000
0.000
0.004
0.000
0.002
0.540
0.014
0.132
------0.000
Coeff.
-0.2745*
0.3663
0.3309
0.4624*
0.0531
-0.3374
0.0462
-------2.0257**
0.0815*
0.2454
0.5358
-0.4401
-0.0755
0.0531
-0.0409
------6.6172**
0.248
0.060
42
143
p-value
0.066
0.474
0.395
0.095
0.565
0.290
0.136
------0.029
0.066
0.583
0.132
0.634
0.624
0.672
0.539
------0.021
Coeff.
-0.0782
0.0013
-1.0387***
0.0959
0.1265***
-0.1939*
0.0236*
0.1412
----0.5463
0.0439**
0.4643***
0.6039***
0.9236***
-0.1309*
0.0857
0.0398**
0.0226
---2.1722**
0.626
0.335
253
761
p-value
0.117
0.993
0.000
0.532
0.007
0.078
0.089
0.306
---0.163
0.018
0.001
0.000
0.000
0.063
0.277
0.019
0.856
---0.033
Brazil
(with cash*free float)
Coeff.
p-value
-0.0908*
0.071
0.0317
0.829
-1.0428***
0.000
0.1661
0.294
0.1282***
0.007
-0.1979*
0.072
0.0062
0.825
0.4380
0.347
0.0422
0.481
-0.2360
0.582
0.0488***
0.009
0.4436***
0.001
0.6303***
0.000
0.8707***
0.001
-0.1376*
0.051
0.0782
0.323
0.0890***
0.007
-1.0551*
0.088
-0.1347*
0.077
2.2881**
0.029
0.628
0.329
253
761
Mexico
Coeff.
-0.3051*
-0.5742
1.3080**
-0.7346
0.2900**
-0.1720
0.2080***
-------6.2558***
0.1937***
-0.3168
-0.2280
2.2146***
-0.2620
-0.1669
-0.2846***
------8.9947**
p-value
0.090
0.450
0.013
0.409
0.015
0.469
0.001
------0.000
0.002
0.561
0.639
0.006
0.143
0.322
0.000
------0.011
0.250
0.096
108
374
Notes: The sample consists of non-financial firms in Argentina, Brazil, and Mexico with available values of all variables during at least one pre-crisis year and one crisis year. The Brazilian crisis period
is defined as the year 1999. The coefficients on the interaction terms represent the difference in the effect of the variable on Tobin’s Q between the crisis and pre-crisis period. Variable definitions are
provided in the Appendix. *** (**) (*) – Coefficient estimate is significant at the 1% (5%) (10%) level
38
Table 6
Market Value Models for Mexican and Brazilian Crises
Mexican Crisis Analysis Period: December 1990 – December 1995
Brazilian Crisis Analysis Period: December 1996 – December 1999
Dependent variable: Market Value = Market Capitalization + ST Debt + LT Debt
Mexican Crisis
Brazilian Crisis
Brazilian Crisis
All Countries
All Countries
Brazil Only
(with free float)
Coeff.
p-value
Coeff.
p-value
Coeff.
p-value
Size (natural log)
0.3821***
0.000
0.6342***
0.000
0.6679***
0.002
Leverage
0.7084
0.116
2.4966***
0.000
2.6501***
0.000
Net Working Capital
1.0988***
0.002
-0.0663
0.728
-0.0859
0.677
Operating Profit
0.3156
0.423
0.6304
0.262
0.5609
0.397
Dividend Dummy
-0.0078
0.937
0.1494
0.314
0.1523
0.453
ADR Dummy
2.1164***
0.000
-0.5426
0.105
-0.3708
0.435
Cash Holdings
0.0151
0.623
0.0072
0.887
0.0146
0.807
Free Float
-------------0.8535
0.153
Cash Holdings*Free Float
------------------Crisis Dummy
-11.507***
0.000
-2.8472**
0.028
0.7488
0.658
Size*Crisis
0.5686***
0.000
0.1505**
0.016
-0.0497
0.533
Leverage*Crisis
-0.3548
0.430
-0.7133
0.146
-0.8004
0.172
NWC*Crisis
-1.1960***
0.001
-0.2892
0.471
-0.2202
0.643
Operating Profit*Crisis
-0.7876
0.192
2.7973***
0.002
3.0786***
0.006
Dividend *Crisis
0.0387
0.795
-0.2289
0.276
-0.2421
0.426
ADR Dummy*Crisis
-2.2749***
0.000
0.1500
0.493
-0.0570
0.867
Cash Holdings*Crisis
-0.0639*
0.087
-0.0007
0.991
-0.0163
0.824
Free Float*Crisis
------------0.5600
0.297
Cash*Free Float*Crisis
------------------Constant
-6.7324***
0.000
-12.0388***
0.000
-12.5641***
0.004
Within R2
Overall R2
Number of firms
No. of firm-year obs.
0.226
0.341
346
1,407
0.086
0.243
403
1,278
0.086
0.245
253
761
Brazilian Crisis
Brazil Only
(with cash*free float)
Coeff.
p-value
0.6583***
0.003
2.7253***
0.000
-0.0646
0.757
0.6586
0.336
0.1560
0.443
-0.3551
0.455
-0.1026
0.402
1.2471
0.536
0.2841
0.273
1.2450
0.502
-0.0352
0.663
-0.8594
0.145
-0.1709
0.722
2.9525***
0.009
-0.2349
0.441
-0.0924
0.787
0.0852
0.548
-1.3471
0.614
-0.2585
0.433
-13.2796***
0.003
0.088
0.244
253
761
Notes: The sample consists of non-financial firms in Argentina, Brazil, and Mexico with available values of all variables during at least one pre-crisis year and
one crisis year. The Brazilian crisis period is defined as the year 1999. The coefficients on the interaction terms represent the difference in the effect of the
variable on Tobin’s Q between the crisis and pre-crisis period. Variable definitions are provided in the Appendix. *** (**) (*) – Coefficient estimate is
significant at the 1% (5%) (10%) level
39
Table 7
Cash Holdings and Firm Value for Brazilian Crisis with Two-Year Crisis Period, December 1996 – December 1999
Dependent variables: Cash Holdings = ln(Cash & ST Investments/Net Assets)
Tobin’s Q = (Market Capitalization + ST Debt +LT Debt)/Net Assets)
Cash Holdings
Cash Holdings
All Countries
Brazil Only
(with free float)
Size (natural log)
Leverage
Net Working Capital
Operating Profit
Dividend Dummy
ADR Dummy
Cash Holdings
Free Float
Cash Holdings*Free Float
Crisis Dummy
Size*Crisis
Leverage*Crisis
NWC*Crisis
Operating Profit*Crisis
Dividend *Crisis
ADR Dummy*Crisis
Cash Holdings*Crisis
Free Float*Crisis
Cash*Free Float*Crisis
Constant
Within R2
Overall R2
Number of firms
No. of firm-year obs.
Coeff.
0.1430
-0.4724
-0.8243***
2.1485***
0.0696
-0.5646**
----------0.8700
0.0318
-0.1139
-0.1678
-0.2870
-0.0159
0.0921
----------10.5851***
0.113
0.023
432
1,355
p-value
0.209
0.262
0.000
0.000
0.561
0.023
---------0.300
0.472
0.759
0.487
0.602
0.907
0.527
---------0.000
Coeff.
0.3912*
0.0319
-0.8239***
3.9367***
-0.0249
-1.2371***
----0.1224
---0.9753
-0.0363
-0.2481
-0.1217
-2.0531**
0.0890
0.2374
----0.7655*
----15.8535**
0.162
0.037
267
797
p-value
0.015
0.958
0.000
0.000
0.894
0.003
---0.816
---0.479
0.601
0.659
0.696
0.018
0.688
0.388
---0.077
---0.000
Tobin’s Q
All Countries
Coeff.
-0.2168***
-0.3134*
-0.9544***
-0.2024***
0.1076**
0.2106*
0.0806***
-------3.0228***
0.1029***
0.7292***
0.4146***
0.8762***
0.0168
-0.3492***
-0.0675***
------5.8974***
0.377
0.096
432
1,355
p-value
0.000
0.087
0.000
0.394
0.040
0.052
0.000
------0.000
0.000
0.000
0.000
0.000
0.779
0.000
0.000
------0.000
Tobin’s Q
Brazil Only
(with free float)
Coeff.
-0.0475
-0.1431
-1.0720***
0.5617**
0.1629***
-0.0256
0.0023
0.1136
---0.3386
-0.0006
0.5752***
0.6898***
-0.1948
-0.1522**
0.0367
0.0404**
-0.1014
---1.4018
0.574
0.365
267
797
p-value
0.326
0.424
0.000
0.040
0.004
0.837
0.890
0.472
---0.445
0.976
0.001
0.000
0.466
0.026
0.654
0.015
0.439
---0.161
Tobin’s Q
Brazil Only
(with cash*free float)
Coeff.
-0.0554
-0.1253
-1.0787***
0.5872**
0.1619***
-0.0334
0.0150
-0.1219
-0.0285
0.4403
0.0027
0.5536***
0.6898***
-0.2126
-0.1548**
0.0305
0.0597*
-0.5260
-0.0518
1.6622
p-value
0.256
0.490
0.000
0.032
0.004
0.788
0.636
0.818
0.680
0.337
0.899
0.001
0.000
0.428
0.024
0.712
0.067
0.358
0.484
0.105
0.576
0.359
267
797
Notes: The sample consists of non-financial firms in Argentina, Brazil, and Mexico with available values of all variables during at least one pre-crisis year and one crisis year. The Brazilian crisis period is
defined as the years 1998-1999. The coefficients on the interaction terms represent the difference in the effect of the variable on Tobin’s Q between the crisis and pre-crisis period. Variable definitions are
provided in the Appendix. *** (**) (*) – Coefficient estimate is significant at the 1% (5%) (10%) level
40