Liquidity and firm investment: Evidence for Latin America

Journal of Empirical Finance 20 2013 18-29 Journal of Empirical Finance Liquidity and firm investment: Evidence for Latin America Francisco Munoz Pontificia Universidad Cat

Journal of Empirical Finance 20 (2013) 18-29

Journal of Empirical Finance

Liquidity and firm investment: Evidence for Latin America

Francisco Munoz

Pontificia Universidad Catélica de Chile, Facultad de Ciencias Econémicas y Administrativas and Finance UC, Chile

Received 5 March 2012 Using a panel of Latin American firms, I find evidence that a higher trading volume and a Received in revised form 7 October 2012 higher industry-adjusted trading volume are associated with higher firm investment (PPE,

Accepted 13 October 2012 : : Total Assets, and Inventory) This relationship is higher in episodes where the firm decides

Available online 22 October 2012 to issue shares, and it is also greater for firms with tighter financial constraints and better

investment opportunities This evidence is consistent with a mispricing channel, where JEL classification: firms issue and invest the proceeds to take advantage of low cost of capital, or with a cost

an informational channel, where a liquid market helps a manager to take more efficient decisions, since this channel does not necessarily predict an increase in investment, but only more efficient investment

Firm investment

Stock market liquidity

Financial constraints

1 Introduction

The study of liquidity in the stock market has attracted much attention in empirical and theoretical literature in recent years Most recently, there has been a growing interest in studying the relationship that may exist between liquidity in the stock market and the real economy At a macroeconomic level, works such as Kaul and Kayacetin (2009), Beber et al (2010), and Naes et al (2011) show evidence at the aggregate and industry level of a positive relationship between stock market liquidity and real variables such as GDP and investment At the microeconomic level, the relationship between liquidity and the decisions of firms has been studied using factors such as the issuance of shares (Butler et al., 2005; Gilchrist et al., 2005; Lipson and Mortal, 2009), leverage (Bharath et al., 2009; Lesmond et al., 2008), and the performance of the firms (Fang et al., 2009) However, a study centered on the relationship between firm real investment and stock market liquidity has not been previously done

Specifically, works such as those of Butler et al (2005) and Lipson and Mortal (2009) study the relationship between liquidity and equity issuance decision, finding that firms with greater liquidity have lower issuance costs, thus using more funding through the issue of shares In this manner, firms with higher liquidity tend to have lower levels of leverage Moreover, Lesmond et al (2008) find firms that increase their level of leverage increase the bid-ask spread (reduced liquidity) Similarly, Bharath et al (2009) show that firms that use a higher percentage of financing through debt, have lower liquidity in the stock market Fang

et al (2009) focuses on the relationship between liquidity and firm performance, finding that firms with greater liquidity have a better performance measured as the market-to-book ratio of assets

Gilchrist et al (2005) finds that greater variance in the predictions of stock market analysts predicts greater actual investment and equity issuance, which is literature that is more related to my findings Similarly, Polk and Sapienza (2009) find, using firm

w ] appreciate the comments of Jaime Casassus, Augusto Castillo, Mario Giarda, Daniel Mufioz, Rodrigo Mufioz, David Ruiz, Francisco Urztia, one anonymous referee, and the seminar participants at the School of Management PUC and at the Annual Meeting 2011 of the Chilean Economic Society (SECHI) In particular! thank Borja Larrain and José Tessada for their comments, support, and encouragement Finally, | thank the partial financial support from Grupo Security through Finance UC

E-mail address: fmmunoz@uc.cl

0927-5398/$ - see front matter © 2012 Elsevier B.V All rights reserved

http://dx.doi.org/10.1016/j.jempfin.2012.10.001

level data for the US, that the investment is larger when the shares are overvalued, using discretionary accruals as a proxy for mispricing.' Fang et al (2012) studies the relation between stock market liquidity and firm innovation (which is related with long-term investment) They find that an increase in liquidity leads to a higher level of institutional ownership by transient and quasi-indexers which reduces innovation Thus, they found a negative relationship between innovation (long-term investment) and stock market liquidity

This paper seeks to provide a novel evidence in this direction, contributing to the small growing literature that studies the relationship between stock market liquidity and firms decision Specifically, I study this relationship using a panel of firms listed

on stock exchanges in four Latin American countries (Argentina, Brazil, Chile, and Mexico), using quarterly data from 1990 to

2010 I estimate an instrumental variable fixed effects panel model that corrects for endogeneity problems in the Tobins' Q and includes country-quarter fixed effects, finding a positive and significant relationship between firms’ investment and stock market liquidity, which is robust to the use of different measures of liquidity (trading volume and firm trading volume over industry trading volume) and investment (PPE, Total Assets, and Inventory) Furthermore, I found that this relationship is increased in episodes of share issuance

This evidence is consistent with a mispricing channel, where firms issue and invest to take advantage of low cost of capital (Gilchrist et al., 2005), or with a cost channel, where liquidity is associated with lower issuance costs although in a rational market (Butler et al., 2005) It is also related, although less, with a mispricing channel without equity issuance (Polk and Sapienza, 2009) and with a channel based on agency problems and information of stock prices, where more liquid markets help the manager to make better investment decisions (Admati and Pfleiderer, 2009; Edmans, 2009; Edmans and Manso, 2011; Khanna and Sonti, 2004; Maug, 1998) These results do not support channels that predict a negative relationship between stock market liquidity and firms’ investment, as in Stein (1988, 1989) and Porter (1992)

Moreover, as was noted before, liquidity has a positive relationship with investment, because the above facilitates the financing of investment Thus, it should be noted that those firms that have greater financial constraints, should be more sensitive

to liquidity, because liquidity enables external financing When estimating a regression that incorporates the interaction between liquidity and a dummy that identifies whether the firm has higher financial constraints, I find that liquidity has a greater relation with investment in firms with greater financial restrictions In turn, the relationship between liquidity and investment should be stronger in firms that have greater investment opportunities For example, there is evidence that firms that have greater investment opportunities are more sensitive to stock market conditions in deciding their investment (Zhang, 2007) The results found in this paper show that the effect of liquidity is higher in firms with greater investment opportunities, as proxied by the interaction of liquidity and a dummy that identifies whether the firm has greater investment opportunities This is consistent with

a view of liquidity acting as a catalyst for decisions to invest

It is important to note that the study of liquidity in emerging markets has not garnered much attention However, in this area two works by Bekaert et al (2007) and Lesmond (2005) stand out in this area The first finds predictability from liquidity to asset returns, while the second makes a study of differences in liquidity between different emerging countries, finding that countries with weaker political and legal institutions have a higher liquidity cost Unlike these works, this paper studies the relation between firm-level liquidity and firm decisions in the context of emerging markets

This paper is organized as follows: Section 2 discusses the various channels in which stock market liquidity might affect a firm's real investment, thus motivating the hypothesis of this study Section 3 presents the econometric strategy and data, while Section 4 shows the empirical results Finally, Section 5 presents the conclusion

2 Liquidity and firm investment

This section explains in more detail the different channels that relate stock market liquidity and firm's real investment I separate this explanations into three different channels that derive a neutral, a positive, and a negative relationship I will begin with the neutral channels First, there are models based on agency problems and information in the stock prices As Maug (1998) and Edmans (2009) propose, liquidity in the stock market facilitates the entry of blockholders Given this entry, Maug (1998) predicts more monitoring activities by investors in highly liquid firms Consequently, more liquid markets tend to support better management of the company In the same light, Admati and Pfleiderer (2009), Edmans (2009), and Edmans and Manso (2011) show that the act of trading on private information by blockholders can discipline managers when managerial compensation is closely tied to stock prices This is because investors collection of information and trading on that information make liquid stock price more informative Khanna and Sonti (2004) show that liquidity can be related positively to the performance of the company This suggest that greater liquidity stimulates the entry of informed investors, which makes the price more informative for the “stakeholders”, thus improving the results of operations and relaxing financial constraints All these models approach the relationship between liquidity and investment, but they do not predict a specific sign of such a relationship

Secondly, there are models related to asset mispricing, which predict a positive relationship between stock market liquidity and firm investment, as detailed in Gilchrist et al (2005), who focus on a mechanism based on share issuing They develop a model where the dispersion of beliefs and short-sale constraints can lead to stock market bubbles, these being exploited by firms issuing new shares with an inflated price

1 These are defined as abnormal differences between accounting profits of the business and cash flow.

20 F, Mufioz / Journal of Empirical Finance 20 (2013) 18-29

Similarly, I can use Miller's (1977) model to derive a positive relationship between trade volume and firm's investment In the original model, short selling is a constraint and agents have different opinions about the returns on investment The latter can be thought of as an overconfident investor, which gives more weight to his personal information than to the market information Since there are short sale constraints, pessimists agents (whom expect low returns on assets) are left out of the market and prices reflect the valuation of the optimist agents (whom expect high returns on assets) In this context, the firm must decide how many shares to issue in order to finance its project, and given that there is a difference of opinions among the agents, the demand for stocks will be positively affected by these disagreements Therefore, the number of shares that the firm decides to issue and the total investment will depend positively on this difference of opinion among investors

Recently, development on theoretical explanations about stock market liquidity has shown that disagreement among investors can explain observed patterns for trade volume (Banerjee and Kremer, forthcoming; Hong and Stein, 2007) Using this result and the prediction of Miller's model; more disagreement among investors is related with increased trade volume, generating a positive relationship between trade volume and firm's investment It is important to note that this model does not predict a causality of trade volume to investment, but that difference of opinions are captured through the trade volume, generating a positive relationship with investment In econometric terms one may think that the trading volume is a proxy for disagreement, which is the relevant variable that affects investment

Unlike this previous models presented, Polk and Sapienza (2009) based on the model of Stein (1996), present a theoretical model that relates the mispricing of stocks to investment, without the necessity of stock issuance These authors show that in a scenario where the stock price is overvalued and there are “stakeholders” who have short time horizons, the wealth of these will depend on the overvaluation Since the manager will decide the optimal level of capital which is maximizing the wealth of

“stakeholders”, price overvaluation will result in a higher capital level

Finally, the last positive channel is related to issuance cost (Butler et al., 2005) Banks should charge lower issuance fees to firms with more liquid stocks because it is easier for investment banks to place a seasoned equity offering in a liquid market than

to place it in an illiquid market Thus, this relationship predicts that higher liquidity is related with more issuance, that can be used to finance more investments by the firm

Now turning to the negative channels, Stein (1988, 1989) shows that in the presence of asymmetry information between managers and investors, takeover pressure could induce managers to sacrifice good long-term performance (like investment) for higher current profits to keep the stock from becoming undervalued Secondly, due to lower trading costs, high liquidity facilitates entry and exit of institutional investors who trade based on current earning news and whose trading may lead to misvaluation and underinvestment in innovation (Porter, 1992)

The main goal of this paper is to try to discriminate between these three main channels that govern the relationship between stock market liquidity and firm investment Also, I conduct a specific test for the equity issuance channel, evaluating if the relationship between stock market liquidity and firm investment increase when firms decide to issue equity

3 Empirical strategy

The relationship between stock market liquidity and firm investment will be tested in a panel of firms listed on stock exchanges in Argentina, Brazil, Chile, and Mexico for the period 1990-2010, using quarterly frequency data The main equation to

be estimated is:

where i stands for firm, c country, f quarter, and j the number of advances | will use three different alternative definitions of investment growth in total assets, growth in property, plant and equipment (PPE), and growth in inventories Long differences are used (1, 2, and 3 years) since the investment is not carried out immediately

Generally, the studies of U.S firms use capital expenditure as the measure of investment However, in the case of Latin America, this variable is not reported by firms in their financial statements Thus, results will be presented with three measures in order to obtain robust results for the definition of investment and to explore the effects of liquidity on different investment horizons There are conceptual differences between these three measures of investment The first is defined as the growth in total assets, which roughly captures investment The problem is that it includes investments at different horizons, and also includes changes in accounts that are not necessarily handled by the firm, such as receivable accounts The second measure is closer to the capital expenditures measure used in the U.S., since it is defined as the growth in fixed assets (property, plant and equipment) This should reflect investment decisions by the firm with a long-term horizon Finally, the change in inventories reflects short-term investments which are decided

by the firm

Liquidity will be measured in two different ways The main measure of liquidity will be trading volume This measure is created using daily data of the quantity of shares traded and the total number of shares of the firm As in Lesmond (2005), I eliminate days when total trading volume is greater than the total number of shares of the firm The measure is defined as:

Dg «Total Shares

where Dg is the number of days of transactions in the quarter

The second measure of liquidity is the trading volume adjusted by industry This is defined as the trading volume of the firm divided by the trading volume of the industry.’ The latter is created as the average trading volume of the firms belonging to that industry in each quarter Thus, I will be measuring whether the investment increases in those firms where differences of opinion are greater in respect to the average differences of opinion in its industry

One measurement issue in trading volume's calculation is that it is calculated over total shares issued and not over the total shares that are available to be traded (Free-Float) For this reason, the denominator of the trading volume may not be the correct one Donelli et al (forthcoming), shows that the ownership structure in Chile is quite stable over time If this holds for all other countries in the sample, | would not have problems in the estimation; there would only be a change of magnitude in the coefficients One potential problem of my measure is that although the theoretical models show a relationship between disagreement and trade volume, I must recognize that there may be other variables that generate changes in the trading volume The important thing is to see if this affects the relationship found between liquidity and investment In the first place, supporting the use of trading volume as a proxy for differences of opinion is the work of Sadka and Scherbina (2007), whom show that firms where disagreement among analysts about future earnings of companies is high also have a high trading volume, thus existing a positive correlation between trading volume and disagreement among analysts This result is also found by Thakor and Withed (2011) and Diether et al (2002), empirically validating trading volume as a proxy of differences of opinion.* However, one might think that variables related to the business cycle are generating changes in the trading volume In order to avoid this problem, quarter-country fixed effects were incorporated, which would remove any macroeconomic factor that could affect all the firms in each country the same Therefore, liquidity would give us net information of the effect of the cycle

An alternative are the studies that have used trading volume as a proxy for investment horizons and information on prices In the case of investment horizons, the work of Polk and Sapienza (2009) and Dong et al (2007) use market trading volume as proxy for investment horizons They find that when there is a greater number of short-term investors (high trading volume), the effect

of asset's mispricing over investment is higher However, the underlying reason of why trading volume represents investors with short term horizons seems to stem from an overconfidence among investors As shown in Cremers and Pareek (2010) as in Odean (1999), Barber and Odean (2000) and Grinblatt and Keloharju (2009), investors who trade most frequently are those who are overconfident This interpretation would thus be in line with this work, given that trading volume captures the overconfidence among investors, which is reflected when they trade more

With regard to information on prices, the empirical evidence is inconclusive about the relationship between trading volume and information On the one hand, Hou et al (2006) found a positive correlation between trading volume and information, using the R? from a regression of returns as its measure of information, while Ferreira et al (2011) find a negative relationship, when they estimate using PIN as its measure of information With respect to the relationship between investment and information, Chen et al (2007) found that in the presence of more informative prices, the effect of the asset price is higher on investment However, the direct relationship between information and investment is not clear Theoretically, it is suggested that more information on prices should lead to more efficient investment (Khanna and Sonti, 2004), which does not imply that investment should be higher or lower As follows, if trading volume only captures information on prices, its relation to investment is not conclusive, since it is not clear whether this means more information or even more so, it is also not clear if more information means more investment

The other controls are the standard regressors in the literature of firm investment.® | include leverage as measured by the total liabilities over total assets, Tobin's Q ratio defined as market-to-book assets,’ which reflects firm investment opportunities and cash flow,® that represents part of the financial constraints that a firm might face Also, I include fixed effects at the firm level (a), which seeks to capture firm-specific characteristics that do not vary over time, while country-quarter fixed effects (a) are include to capture business-cycle effects inherent to each country

As Bond and Van Reenen (2008) and Almeida et al (2010) show, there are problems if I estimate Eq (1) using a fixed effect estimator The main problem being that Tobin's Q could be an endogenous regressor This problem arises because the standard way of introducing stochastic variation into the Q model is to treat this parameter as stochastic, and to interpret the error term as reflecting adjustment costs This implies that Q is an endogenous variable since current shocks to adjustment costs will affect the current period's revenue and therefore the current value of the firm Furthermore, there can be an endogeneity problem due to measurement errors in Tobin's Q.°

Bond and Van Reenen (2008) and Almeida et al (2010) present different approaches to solve this problem A first strategy is to differentiate the model and use lags of the endogenous variable as instruments, this can be estimated using an IV-OLS estimator or

a GMM estimator The second strategy is to keep the model in levels and use lags of first differences as instruments, this is estimated using an IV-OLS estimator To choose between these two models I perform a weak instrument test for the first stage, in both cases I use two lags of the instruments Using the F-test for the excluded instruments, I find that for the first methodology it

? This measure is similar to adjusted trading volume proposed by Sadka and Scherbina (2007), except that they scale by the trading volume of all the market

3 It would be the trading volume of an “Equally-Weighted” portfolio at industry level

4 Other works have that used trading volume as a proxy for differences of opinion are Berkman et al (2009), Hong and Stein (2007), Diether et al (2002), among others

° This measure represents the probability that an informed agent traded on the market

© See Almeida and Campello (2007) and Polk and Sapienza (2009)

7 It is defined as (Stock Market Capitalization, + Total Debt,)/Total Assets,

® Defined as (Depreciation, + EBIT,)/ Total Assets,

° See Bond and Van Reenen (2008) and Almeida et al (2010)for a detailed discussion.

22 F, Mufioz / Journal of Empirical Finance 20 (2013) 18-29

is near 17, while for second case is 84 Thus, I reject the first set of instruments as valid while the second shows no sign of weak instruments Following Stock and Yogo (2005), I use a critical value of 19.93 for the weak instruments test with two instruments and that admits a 10% of maximal size in IV estimation Therefore, I estimate the main equation in levels with two lags of first differences as instruments In all the regressions I show the F-test for excluded instruments and the Hansen test for overidentification, both of which

As shown in Eq (1), the central hypothesis is that the parameter PB is positive and significant, reflecting that increased liquidity

is related to increased investment by the firm This will be evidence of the presence of the mispricing channel and issuance cost channel

To test between the different channels, I focus on scenarios where companies decided to issue equity and test if the relationship between investment and liquidity occurs more sharply To test this hypothesis, the base specification adds an interaction between a dummy that identifies if the firm issued shares and the measure of liquidity A positive and significant parameter would reflect the relevance of mispricing channels in the case of issuance (Gilchrist et al., 2005; Miller, 1977) and the cost of issuance channel (Butler

et al., 2005) Otherwise, there would be channels other than issuance that may generate this relationship (Polk and Sapienza, 2009)

If there is a positive relationship between liquidity and investment, it is because liquidity facilitates the financing of investment By separating firms according to financial constraints, it should be observed that those firms more financially constrained are more sensitive to liquidity Following Almeida and Campello (2007), companies were separated into large and small according to their total assets, with the object of capturing financial constraints.'' This separation is also supported by Beck et al (2008) who show evidence that there is a difference in funding between large and small companies in a country panel dataset, which includes the countries analyzed in this paper, finding that small firms tend to have less external financing In this way, liquidity could relax these differences for the small firms, thus encouraging further investment This will be tested by adding to the base specification a dummy representing whether the firm is large or not, and making it interact with the liquidity variable A significant and negative coefficient would represent that the effect of stock market liquidity is heterogenous by financial constraints Those firms with greater financial constraints will take advantage of the higher liquidity and will invest more

Finally, if liquidity encourages more investment, this effect should be more pronounced in those firms that have greater investment opportunities It is argued that firms with greater investment opportunities (“growth”) would have greater ability on the timing of their investment, while those with lower investment opportunities (“value”) tend to be more stable in their investment (Zhang, 2007) Thus, in a context where there is greater liquidity “growth” firms would take advantage of this scenario, investing more '” To test this, I include in the base regression a dummy indicating whether the firm is “value” or “growth”, interacted with the measure of liquidity If the coefficient is significant and negative, it would be evidence that the effect of liquidity is heterogenous between investment opportunities, and it is greater for “growth” firms than for “value” firms

The firm level data was obtained from software Economdtica, excluding financial industries.!* For the case in which a firm has more than one series of shares, I took the most traded series The data was winsorized at 2% top and bottom in order to eliminate possible problems of outliers This practice is quite common for firm-level data The sample contains around 5000 observations and 450 firms This panel is highly unbalanced, as the firms financial information is not reported in complete form for all the periods Table 1 presents the descriptive statistics of the variables of interest

4 Results

Before reviewing the empirical results, I present Fig 1 with the aim of showing that the proposed relationship between liquidity and investment is positive This figure shows the average investment (growth in fixed assets, PPE) at one year term, for percentiles 0-33, 33-66 and 66-100 of liquidity (trading volume) It can be clearly seen that investment is increasing in the liquidity of the stock market, supporting the central hypothesis of this research

Turning to the econometric analysis, Table 2 presents the results of the proposed specification in Eq (1), using trading volume as the measure of firm liquidity The coefficient of interest, 6, is significant and positive for the three definitions of investment at 1, 2 and

3 years This shows that firms which have a higher trading volume, incur higher investments This variable is economically significant For example, an increase in one standard deviation of trading volume leads to an increase in investment in fixed assets of 1.3% in one year, which is enough if we consider that the unconditional media of this investment is 6% (see Table 1) These orders of magnitude are maintained for other investments and different horizons Consequently, the hypothesis I present in this paper, stock market liquidity is related with more investment, has substantial supporting evidence

The rest of the controls, as in the case of leverage appears to be significant and negative This would be evidence in favor of the over-investment channel Firms with higher leverage level will require a greater cash flow to pay interest and capital, thereby reducing its capital to invest in new projects The coefficient of this variable is similar in magnitude and significance to the study done by Aivazian et al (2005) for the relationship between leverage and investment in the case of firms in Canada

10 Another methodology is the one proposed by Erikson and Whited (2000, 2002) which delivers consistent parameters, by estimating with GMM using third and fourth moments for identification However, Almeida et al (2010) by means of a Monte Carlo exercise pitted one estimator against the other, finding that this estimator is biased in the presence of fixed effects, heteroskedasticity and small samples, while using VI delivers unbiased parameters

"As in Beck et al (2008), the firms were divided for each quarter and each country between those who were above the total assets median (large) and those below (small)

12 Firms were separated in each country and each quarter according to the book-to-market of equity Firms with a book-to-market above the median are considered “value”and the firms that are below are considered “growth”

13 These are the “Fondos” and “Finanzas y Seguros” industries in Economdtica.

Table 1

Summary statistics

This table shows descriptive statistics of the variables of interest The sample consists of quarterly data from 1990 to 2010 for firms in Argentina, Brazil, Chile and Mexico All variables were winsorized at 2% in each tail The dependent variables are growth in total assets, growth in fixed assets (Property, Plant and Equipment) and growth in Inventories, where Al, ;= (1:4; —1,)/I The variables measuring liquidity are Trading Vol constructed as Number of Shares Traded/ Shares Outstanding and Firm Trad Vol./Ind constructed as Number of Shares Traded/Shares Outstanding of each firm over the industry average The control variable group includes Leverage defined as Total Liabilities over Total Assets, Tobin's Q defined as (Stock Market Capitalization + Total Liabilities) /Total Assets and Cash Flow = (EBIT+ Depreciation)/Total Assets

Investment

Liquidity

Controls

Tobin's Q, which reflects investment opportunities, turns out to be positive and significant Thus, firms with greater investment opportunities have increased investment This result is fairly standard in literature on the subject The coefficients are similar in significance and magnitude to those obtained by Polk and Sapienza (2009) for US firms

Finally, cash flow is significant and positive The idea behind this result is that the positive relationship reflects the financial constraints that face the firm If a company is restricted for foreign credit, cash flow will allow you to perform a greater investment (Fazzari et al., 1988) The results for this variable are similar to those found by Almeida and Campello (2007) both in significance and magnitude It should be noted that this parameter could be potentially biased as was stated by Cummins et al (2006) To test if the cash flows are related to financial constraints I followed Bond et al (2003), estimating the relationship between EBIT and cash flow I found a positive and strong relationship between them, which supports the suitability of cash flow as measure for financing constraints In the case of inventories, the effect is negative and not significant Studies like Benito (2005) and Carpenter et al (1998) are inconclusive about the sign and the significance of this variable

Table 3 presents the results for the measure of industry-adjusted trading volume One may see that the results hold for the case of the variable of interest, showing that firms which have a trade volume higher than the industry average, have higher investment The economic significance is maintained for this variable The results for the rest of the controls are maintained

Fig 1 Firm investment splited by firm trading volume percentiles.

24 F Muñoz / Journal of Empirical Finance 20 (2013) 18-29

Table 2

Using firm trading volume as liquidity measure

This table shows panel estimation instrumenting Tobin's Q with two lags and including firm fixed effects and country-quarter fixed effects, with robust standard errors that correct for heteroscedasticity and serial correlation at the firm level The sample consists of quarterly data from 1990 to 2010 for firms in Argentina, Brazil, Chile and Mexico All variables were winsorized at 2% in each tail The dependent variables are growth in total assets, growth in fixed assets (Property, Plant and Equipment) and growth in Inventories, where Al;,;=(i+;—)/ Controls include Trading Vol constructed as Number of Shares Traded/Shares Outstanding, Leverage defined as Total Liabilities over Total Assets, Tobin's Q defined as (Stock Market Capitalization + Total Liabilities) /Total Assets and Cash Flow = (EBIT+ Depreciation)/Total Assets

Variables Art+al/K Ar+sl/K Ar+lal/K At+al/K Ar+sl/K Ar+lal/K At+al/K At+sl/K Ar+lal/K

Trading Vol 6.794** 15.633** 22.493** 5.127** 9,226** 11.312 11.116*** 15.988** 20.692*

(3.133) (6.404) (10.051) (2.467) (4.580) (7.285) (3.991) (7.970) (11.891) Leverage —0.291*** —0.717*** —0.884*** —0,.294"* —0549*** —0.682*** —0.401*** —0.602*** —0.704**

(0.056) (0.119) (0.187) (0.050) (0.100) (0.148) (0.105) (0.204) (0.343) Tobin's Q 0.091* 0.182*** 0.275*** 0.106*** 0.182*** 0.249*** 0.094*** 0.162*** 0.224**

(0.055) (0.039) (0.066) (0.016) (0.033) (0.062) (0.024) (0.061 ) (0.099)

(0.013) (0.041 ) (0.025) (0.003) (0.008 ) (0.016) (0.012) (0.020) (0.023)

F test (excluded instruments) 85.62 85.62 85.62 85.62 85.62 85.62 85.62 85.62 85.62

Robust standard errors and clustered by firm in parentheses

*** p<0.01, ** p<0.05, * p<0.1

The above results suggest a positive relationship between stock market liquidity and firm-level investment, supporting the channels proposed by Gilchrist et al (2005), Butler et al (2005) and Polk and Sapienza (2009) However, these results differ in sign from Fang et al (2012) who found a negative relationship between liquidity and innovation (patents), which they relate to long-run investment This difference may arise for two reasons First, I use capital expenditures, while Fang et al (2012) used patents These two measures can differ because patents are mainly related with an efficient investment while capital expenditure can be in some cases an inefficient investment, such as empire building investment in Stein (2003) Accordingly, it may be in some cases that patents are lower and long-term capital expenditure is high Second, in these countries it is less probable to find a negative channel like the one suggested by Stein (1988, 1989) and Porter (1992), based on earnings management and career concerns based on earnings reports This is because in the markets that I study it is more common to find large controlling shareholders with long-term objective functions, due to lower stock market pressure and lower myopic behavior of managers (La Porta et al., 1999; Shleifer and Vishny, 1986)

Table 3

Using firm trading volume relative to industry trading volume as liquidity measure

This table shows panel estimation instrumenting Tobin's Q with two lags and including firm fixed effects and country-quarter fixed effects, with robust standard errors that correct for heteroscedasticity and serial correlation at the firm level The sample consists of quarterly data from 1990 to 2010 for firms in Argentina, Brazil, Chile and Mexico All variables were winsorized at 2% in each tail The dependent variables are growth in total assets, growth in fixed assets (Property, Plant and Equipment) and growth in Inventories, where AJ; j= (1:4.;— 1)/I; Controls include Firm Trad Vol./Ind constructed as Number of Shares Traded/Shares Outstanding of each firm over the industry average, leverage is defined as Total Liabilities over Total Assets, Tobin's Q defined as (Stock Market Capitalization + Total Liabilities )/Total Assets and Cash Flow = (EBIT-+ Depreciation )/Total Assets

Variables At+al/K Ar+sgl/K Ary 12l/K Ar+al/K Ar+sl/K Ary 1al/K At+al/K Ar+sl/K Ary 1al/K

Firm Trad Vol./Ind 0.009 0.026** 0.043** 0.010** 0.020** 0.024* 0.020*** 0.028* 0.034

(0.006) (0.011) (0.018) (0.004) (0.008) (0.013) (0.007) (0.015) (0.022) Leverage 0.284*** = —0.729*** —0,903*** —0.297*** —0556*** —0.692*** —0410** —0/614*' —0.7207”

(0.057) (0.121) (0.190) (0.049) (0.100) (0.147) (0.106) (0.204) (0.342) Tobin's Q 0.102*** 0.1857” 0.275*** 0.107*** 0.183*** 0.249*** 0.096*** 0.165*** 0.225**

(0.018) (0.039) (0.066) (0.016) (0.034) (0.062) (0.025) (0.062) (0.100)

(0.013) (0.041) (0.025) (0.003) (0.008) (0.017) (0.012) (0.020) (0.024)

F test (excluded instruments) 84.70 84.70 84.70 84.70 84.70 84.70 84.70 84.70 84.70

Robust standard errors and clustered by firm in parentheses

*** p<0.01, ** p<0.05, * p<0.1.

However, as was mentioned previously, it is important to study whether this result is obtained through the issuance of shares,

by testing in this manner which channel is operating Table 4 shows the results where I add the interaction between a dummy that identifies if the firm issued shares and the measure of liquidity This table is constructed using as measures of investment, the PPE and Inventories and both measures of liquidity It can be seen that liquidity has an impact on investment that is amplified in cases where there was an issue of shares This evidence is consistent with channels that propose that liquidity is related with firm's investment because it encourages equity issuance The coefficients show that the importance of liquidity in the case of issuance, on average, is more than double than in the case without issuance

It is important to recognize that this result does not hold when I use the change in total assets as a measure of investment One possible explanation, as mentioned previously, is that total assets include investments at different horizons, making the effect of liquidity be the average of different types of investment Moreover, the change in total assets includes several items that are not handled by the manager For such items liquidity should have no effect Another alternative is that this is evidence for other mechanisms that are also generating this relationship between liquidity and investment With the results obtained, I cannot rule out the mechanism proposed by Polk and Sapienza (2009), where the mispricing of assets directly affects the investment, without any share issuance However, the mechanisms proposed by Maug (1998), Khanna and Sonti (2004), Admati and Pfleiderer (2009), Edmans (2009), and Edmans and Manso (2011) are not easily to detect in the data This is because this studies suggest that higher liquidity leads to more efficient investment decisions by firms, which does not imply greater investment, as was found

in the results presented I also do not find evidence of a negative channel based on earnings management

Tables 5 and 6 show evidence regarding the differential effect for firms with greater financing constraints The results are consistent with the intuition presented in the previous section In those firms with larger financial constraints (small ones), liquidity is shown to have a closer relationship with investment than with firms with lower financial restrictions (large), supporting the evidence found by Beck et al (2008) This is reflected in the negative and significant coefficient on the interaction “Large X Liquidity” It is important to note that the effect on large firms (represented by the sum of both coefficients that incorporate liquidity) is not significant in majority of the cases This result is observed for the definitions of PPE and Inventories investment and for both measures

of liquidity These results are maintained if 1 separate firms by size of stock market capitalization, that is to say that low capitalization firms have a higher relationship between liquidity and investment Similar results are observed if] separate the sample between the

Table 4

Interaction with share issuance

This table shows panel estimation instrumenting Tobin's Q with two lags and including firm fixed effects and country-quarter fixed effects, with robust standard errors that correct for heteroscedasticity and serial correlation at the firm level The sample consists of quarterly data from 1990 to 2010 for firms in Argentina, Brazil, Chile and Mexico All variables were winsorized at 2% in each tail The dependent variables are growth in total assets, growth in fixed assets (Property, Plant and Equipment) and growth in Inventories, where Al;+;=(i+j;—1)/i Controls include Trading Vol constructed as Number of Shares Traded/Shares Outstanding, Issue X Trading Vol is the interaction between a dummy that identifies if the firm issued and Trading Vol., Firm Trad Vol./Ind constructed as Number of Shares Traded/Shares Outstanding of each firm over the industry average, Issue X Firm Trad Vol./Ind is the interaction between a dummy that identifies if the firm issued and Firm Trad Vol./Ind., Leverage is defined as Total Liabilities over Total Assets, Tobin's Q defined as (Stock Market Capitalization + Total Liabilities) /Total Assets and Cash Flow = (EBIT+ Depreciation )/Total Assets

Aryal/K Aresl/K Areal/K Apsal/K Atssl/K Apeal/K Arsal/K Atssl/K Atsviol/K Acpal/K Apisl/K Apsaal/K

Issue X 3.260 9.152” 10.9837 8.280” 10.719 13.878

Trading Vol (2.684) (4.955) (5.648) (4.281) (8.542) (11.020)

Vol./Ind

0.283*** 0.5327" 0.648*** 0.366" 0581⁄'* 0.6847 0286''* 0.538" 0658°'* 0.374*** 0594"** 0.699

firms

(p value)

(excluded

instruments)

Robust standard errors and clustered by firm in parentheses

** p<0.01, ** p<0.05, * p<0.1.

26 F, Mufioz / Journal of Empirical Finance 20 (2013) 18-29

Table 5

Interaction with financial constraints, using trading volume

This table shows panel estimation instrumenting Tobin's Q with two lags and including firm fixed effects and country-quarter fixed effects, with robust standard errors that correct for heteroscedasticity and serial correlation at the firm level The sample consists of quarterly data from 1990 to 2010 for firms in Argentina, Brazil, Chile and Mexico All variables were winsorized at 2% in each tail The dependent variables are growth in total assets, growth in fixed assets (Property, Plant and Equipment) and growth in Inventories, where Al;,;=(i+;—)/ Controls include Trading Vol constructed as Number of Shares Traded/Shares Outstanding, Large X Trading Vol is the interaction between a dummy that identifies whether the firm is larger by assets and trading volume, Leverage defined as Total Liabilities over Total Assets, Tobin's Q defined as (Stock Market Capitalization + Total Liabilities)/Total Assets and Cash Flow = (EBIT+ Depreciation)/Total Assets

Variables At+al/K At+sl/K Ar+lal/K Ar+al/K At+sgl/K Ar+lal/K Ar+al/K Arigl/K Ary 42l/K

Trading Vol 5.639* 11.864* 19.183** 6.980** 13.676*** 19.790** 14.579”** = 26.720*** 35.004***

(3.288) (6.144) (9.026) (2.966) (5.145) (7.960) (4.258) (8.472) (13.274) Large X Trading Vol 3.334 10.364 8.793 —5,.154* —12.216 —22.456*** —9,496** —29,032*** —37.827***

(4.282) (9.437) (12.483) (2.858) (5.341) (7.614) (4.304) (8.444) (12.261) Leverage —0.281*** = —0,722*** —0.888*'' —0.292*** —0.544*** —0.674*** —0.396*** —0.585*** —0.683””

(0.056) (0.118) (0.186) (0.049) (0.099) (0.146) (0.105) (0.205) (0.344) Tobin's Q 0.101*** 0.185*** 0.278*** 0.105*** 0.179*** 0.241*** 0.091*** 0.154*** 0.214**

(0.018) (0.039) (0.066) (0.015) (0.032) (0.060) (0.024) (0.059) (0.096)

(0.013) (0.041) (0.025) (0.003) (0.008) (0.016) (0.012) (0.020) (0.023)

Hansen test (p value) 0.450 0.936 0.681 0.674 0.150 0.448 0.827 0.326 0.616

F test (excluded instruments) 86.28 86.28 86.28 86.28 86.28 86.28 86.28 86.28 86.28

Robust standard errors and clustered by firm in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Table 6

Interaction with financial constraints, using firm trading volume relative to industry trading volume

This table shows panel estimation instrumenting Tobin's Q with two lags and including firm fixed effects and country-quarter fixed effects, with robust standard errors that correct for heteroscedasticity and serial correlation at the firm level The sample consists of quarterly data from 1990 to 2010 for firms in Argentina, Brazil, Chile and Mexico All variables were winsorized at 2% in each tail The dependent variables are growth in total assets, growth in fixed assets (Property, Plant and Equipment) and growth in Inventories, where Al,+;=(4;—1,)/I, Controls include Firm Trad Vol./Ind defined as Number of Shares Traded/Shares Outstanding of each firm over the industry average, Large X Firm Trad Vol./Ind is the interaction between a dummy that identifies whether the firm is larger by assets and Firm Trad Vol./Ind., Leverage defined as total assets Total liabilities, Tobin's Q defined as (Stock Market Capitalization + Total Liabilities)/Total Assets and Cash Flow = (EBIT + Depreciation) /Total Assets

Variables Aryal/K Ary sl/K Ary 12l/K Ar+tal/K Arysl/K Ary 12l/K Ar+al/K Ar+sl/K Ary 12l/K

Firm Trad Vol./Ind 0.006 0.016 0.030 0.014** 0.029*** 0.040** 0.026*** 0.052*** 0.058**

(0.007) (0.012) (0.020) (0.006) (0.011) (0.017) (0.008) (0.018) (0.029) Large X Firm Trad Vol./Ind 0.007 0.022 0.030 —0.011* —0.021* —0.036* —0.014 —0.059** —0.057

(0.008) (0.017) (0.026) (0.006) (0.012) (0.020) (0.012) (0.024) (0.037) Leverage —0.284*** —0.732*** —0.907*** —0.296*** —0553*** —0.687*** —0.407*** —0598*** —0.706**

(0.057) (0.121) (0.189) (0.049) (0.099) (0.146) (0.106) (0.206) (0.342) Tobin's Q 0.103*** 0.188*** 0.280*** 0.106*** 0.180*** 0.243*** 0.094*** 0.157*** 0.217**

(0.018) (0.040) (0.067) (0.015) (0.033) (0.061) (0.025) (0.060) (0.098)

(0.013) (0.040) (0.025) (0.003) (0.008 ) (0.016) (0.012) (0.020) (0.024)

F test (excluded instruments) 84.79 84.79 84.79 34.79 3479 84.79 34.79 3479 3479

Robust standard errors and clustered by firm in parentheses

*** p<0.01, ** p<0.05, * p<0.1

large and the small firms In this case stock market liquidity is shown to be insignificant for large firms in most of the regressions, while for small firms the coefficient is higher and highly significant !*

Finally, Tables 7 and 8 show whether the effect of liquidity on investment is greater for firms that have more investment opportunities (“growth”) It can be seen that the effect is more ample for “growth” firms, thus supporting the idea that these firms

™ These tables are available upon request

Table 7

Interaction with growth opportunities, using firm trading volume

This table shows panel estimation instrumenting Tobin's Q with two lags and including firm fixed effects and country-quarter fixed effects, with robust standard errors that correct for heteroscedasticity and serial correlation at the firm level The sample consists of quarterly data from 1990 to 2010 for firms in Argentina, Brazil, Chile and Mexico All variables were winsorized at 2% in each tail The dependent variables are growth in total assets, growth in fixed assets (Property, Plant and Equipment) and growth in Inventories, where Al,,;=(i+j;—4)/ Controls include Trading Vol constructed as Number of Shares Traded/Shares Outstanding, High B/M X Trading Vol is the interaction between a dummy that identifies whether the firm has a high Book-to-Market of assets and Trading Vol., Leverage is defined as Total Liabilities over Total Assets, Tobin's Q defined as (Stock Market Capitalization + Total Liabilities) /Total Assets and Cash Flow = (EBIT + Depreciation)/Total Assets

Variables Ar+al/K Ar+sgl/K Ary 1al/K Art+al/K Ar+sl/K Ar+lal/K Ar+al/K Ar+sl/K Ary 12l/K

Trading Vol 11.459*** 20.404*** 27.167** 7,.842*** 12.423** 14.102” 12.7007” 19.230”? 22.224

(3.967) (7.617) (11.745) (3.018) (5.189) (8.559) (4.310) (8.822) (13.680) High B/M X Trading Vol —13.692*** —14,704** —14,190 —8.051*** —9,871** — 8.454 —4.843 —9.919 —4.678

(4.118) (6.727) (9.752) (2.843) (4.363) (7.146) (4.054) (7.602) (10.380) Leverage — 0.283*** —0.719*** —0.883*** —0.296”' —0551*** —0.681*** —0402*** —0.602*** —0.702**

(0.055) (0.117) (0.186) (0.049) (0.100) (0.148) (0.105) (0.204) (0.344) Tobin's Q 0.089*** 0.169*** 0.262*** 0.099*** 0.173*** 0.241*** 0.089*** 0.153** 0.220**

(0.017) (0.039) (0.066) (0.015) (0.032) (0.061) (0.024) (0.060) (0.100)

(0.013) (0.041) (0.025) (0.003) (0.008) (0.017) (0.012) (0.020) (0.023)

F test (excluded instruments) 83.11 83.11 83.11 83.11 83.11 83.11 83.11 83.11 83.11

Robust standard errors and clustered by firm in parentheses

** p<0.01, ** p<0.05, * p<0.1

Table 8

Interaction with growth opportunities, using firm trading volume relative to industry trading volume

This table shows panel estimation instrumenting Tobin's Q with two lags and including firm fixed effects and country-quarter fixed effects, with robust standard errors that correct for heteroscedasticity and serial correlation at the firm level The sample consists of quarterly data from 1990 to 2010 for firms in Argentina, Brazil, Chile and Mexico All variables were winsorized at 2% in each tail The dependent variables are growth in total assets, growth in fixed assets (Property, Plant and Equipment) and growth in Inventorys, where AJl,+;=(/:4;—1,)// Controls include Firm Trad Vol./Ind built as Shares Traded/Shares Outstanding of each signature on the industry average, High B/M X Firm Trad Vol./Ind which is the interaction between a dummy that identifies whether the firm has a high Book-to-Market of assets and Firm Trad Vol./Ind., Leverage is defined as Total Liabilities over Total Assets, Tobin's Q defined as (Stock Market Capitalization + Total Liabilities) /Assets total and Cash Flow = (EBIT+ Depreciation)/Total Assets

Variables Ar+al/K Ar+sl/K Ar+lal/K Ar+al/K At+sl/K Ar+lal/K At+al/K At+sgl/K At+lal/K

Firm Trad Vol./Ind 0.020** 0.038** 0.055** 0.018*** 0.031*** 0.035** 0.026*** 0.0397” 0.043"

(0.008) (0.015) (0.022) (0.006) (0.011) (0.016) (0.008) (0.017) (0.026) High B/M X Firm Trad Vol./Ind —0.027*** —0.033** —0.029 —0.020*** —0,027*** —0.027” —0.014 —0.031 —0.025

(0.008) (0.016) (0.022) (0.006) (0.010) (0.015) (0.010) (0.021) (0.029) Leverage —0.292*** —0.740** —0.911*** —0,304*** —0565*** —0.699”*' —0.414'7”° —0.621*** —0.724””

(0.056) (0.120) (0.189) (0.049) (0.099) (0.146) (0.105) (0.204) (0.341) Tobin's Q 0.089*** 0.169*** 0.261*** 0.097*** 0.169*** 0.235*** 0.088*** 0.149** 0.213**

(0.018) (0.039) (0.066) (0.015) (0.033) (0.062) (0.025) (0.061) (0.102)

(0.013) (0.041) (0.025) (0.003) (0.008) (0.017) (0.012) (0.020) (0.024)

F test (excluded instruments) 79.59 79.59 79.59 79.59 79.59 79.59 79.59 79.59 79.59

Robust standard errors and clustered by firm in parentheses

*** p<0.01, ** p<0.05, * p<0.1

would be more sensitive to the characteristics of the market when making their investment, as stated by Zhang (2007) This result

is true for both measures of liquidity As in the case of the financial constraint regressions, I estimate different regressions for

“growth” and “value” firms finding that the stock market liquidity is highly significant for “growth” firms while not significant and

15 These tables are available upon request.