Research Design For Testing Cash Flow Sensitivity

Published Date: 02 Nov 2017

In testing the cash flow sensitivity of cash for alternative sub-groups, I use the following empirical model:

∆CashHoldings(i,t) = α0 + α1CashFlow(i,t) + α2Q(i,t) + α3Size(i,t) + ℮(i,t) , (1)

where i and t denote firm and time respectively, Q is the Tobin’s Q, which represents the market-to-book value of the firm, Size is the natural log of assets. The theory concerns the change in cash holdings in response to a difference in cash flows, captured by α1 in equation (1). The theory also states that cash policy of constrained firms should be affected by the advantages of future investment, which is captured by α2 in equation (1). The control of size is because of debates in economies of scale in cash management.

The first step of my analysis involves estimating equation (1) using the entire sample. Then, I order sample using firm-specific financing constraint characteristics. I progressively remove the lowest 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, and 75 percent of the observations in such a way to obtain 15 different classes of firm-year observations, where the firms in sub-samples become less constrained as I delete observations. Figure 1 shows how I obtain my sub-samples of firms except for dividend payouts and leverage.

insert Figure 1 here –

Having estimated the cash flow sensitivity of cash for each of these classes by means of equation (1), I use estimated class coefficients to test the null hypothesis that they do not statistically differ from the parameter estimated for the entire sample. As mentioned earlier, the analysis is performed with making no statements about the degree of financial constraints faced by the sub-samples I am progressively left with. I use five financing constraints criteria to partition my sample:

Scheme #1: I rank firms based on their cash dividends. My first class is the firms that are non-payers dividends. Then, I start deleting the lowest 5 percent of the rest in a way that I obtained fifteen classes of firms for which cash dividends is progressively increasing. Then, I estimate the cash flow sensitivity of cash across these classes and test the null hypothesis that they do not statistically differ from the parameter estimated for the entire sample.

Scheme #2: Secondly, I rank the firms based on their leverage levels. I compute the leverage as the ratio of total debt to total assets. Using the same procedure as above, I built fifteen classes of firms for which leverage is progressively increasing. The first class is the firms having zero leverage. I estimate and compare cash flow sensitivities of cash across these classes.

Scheme #3: Then, I rank firms based on their total assets which is taken as a proxy for firms’ size. I obtained fifteen classes of firms for which size is progressively increasing. Similar to proxy of cash dividends, I estimate the cash flow sensitivity of cash across these classes and test the null hypothesis.

Scheme #4: Cash holding is another financing constraint that I use. I obtained fifteen classes of firm-year observations for which cash holdings of firm is progressively increasing. Same as I other criteria, I estimate the cash flow sensitivity of cash across these classes and then I test the null hypothesis.

Scheme #5: Finally, I use an index of firm financial constrains based on results in Kaplan and Zingales (1997) and separate firms according to this measure (which is called "KZ index"). I obtained fifteen classes of firm-year observations for progressively increasing level of KZ index. Then, I estimate and compare cash flow sensitivities of cash across these classes. The KZ index refers to:

KZindex = -1.002 x CashFlow + 0.283 x Q + 3.139 x Leverage – 39.368 x Dividends

– 1.315 x CashHoldings (2)

Data

My analysis uses an unbalanced panel of publicly traded USA NASDAQ and NYSE firms and Mexico firms, from 1989-2008. The initial sample is the set of all firms for which data are available on Datastream. This database provides both accounting data for firms and market value of equity. I exclude financial firms from the sample, from those I choose only with at least three continuous time series observations during the sample period. These criteria provide for a total of 13,762 firm-year observations for 2,341 Nasdaq firms, 16,340 firm-year observations for 2,006 Nyse firms, and 1,182 firm-year observations for 190 Mexico firms.

I measure cash holdings as total cash and equivalent items, cash flow as net income plus depreciation and amortization, leverage as total debt. All variables, including cash dividends and Tobin’s Q, are normalized by total assets. Tobin’s Q is proxies by the market-to-book ratio, which has been measured as total liabilities. I use the total assets to proxy for the size of the firms. All financial variables are for the end of the fiscal year.

Results

Univariate Analysis

Table 1 reports descriptive statistics on the main financial variables used in the analysis for US Nasdaq, US Nyse, and Mexico firms respectively. Results of Panel A of the table show that the average cash holdings ratio of 2,341 firms in sample during the sample period is 27 percent for US Nasdaq firms. The average leverage ratio during the same period is 16 percent which is less than 25 percent. The average firm’s market-to-book ratio is about 2.24 and the average ratio of cash flows to total assets is about 0.01. In Panel B, results for 2,006 US Nyse firms are in average 8 percent of cash holdings, 27 percent of leverage, 1.78 of market-to-book ratios, and 0.10 of cash flow to total assets ratio. These values are not in line with those reported for the US Nasdaq firms since average cash holdings level is low. Finally, Panel C shows the results for Mexico firms which are holding in average 8 percent of cash holdings, 24 percent of leverage, 0.6 of market-to-book ratio, and 0.07 of cash flows to total assets ratio.

insert Table 1 here -

Table 2 reports correlation among variables. In line with expectations, cash flows of all firms are positively related to difference in cash holdings of the firms. Additionally, growth opportunities, which proxies by the market-to-book ratios are positive for all US firms. However, Panel C displays negative correlation with market-to-book ratio and the difference in cash holdings for Mexico firms. Also, size is negatively correlated with the difference in cash holdings of both US and Mexico firms. These results are in line with the findings provided for US firms in the current literature (see, Almeida et al., 2004).

insert Table 2 here -

In Table 3, I present the US Nasdaq firms’ mean values of the main variables of the analysis for sub-samples that are determined using the financing constraints proxies I use in this paper. For instance, in Panel I of Table 3, I report the mean values in columns A to E for five sub-samples that are generated by dividing equally the total sample into five sub-samples using the entire distribution of dividend payouts. Column A shows firm-year observations corresponding to the lowest dividend payout observations in the distribution, while column E represents the highest dividend payout observations. Besides having an unclear picture, the statistic suggests that cash holdings of the firms decrease as firms pay out higher dividends. This indicates that the degree of financing constraints reduces as move from the lowest to the highest dividend payout groups. Plus, higher dividend payouts are associated with greater firm size and smaller market-to-book ratio. This shows that smaller investment opportunities and larger size of firms seem to be less constrained compared to lower dividend ones.

insert Table 3 here –

In Panel II, I provide the average values by grouping firm-year observations using the leverage ratio of firms. To be consistent across all panels, I order observations from the lowest to the highest leverage ratios. Hence, column E in Panel II should represent the most constrained firms. According to results, higher leverage firms are larger, with smaller growth opportunities, proxies by the market-to-book ratio, and lower cash holdings. Although it is not a clear picture, the difference in cash holdings of firms decreases as leverage ratio increases. Taken together, the leverage ratio increases as firms accumulating lower amounts of cash relative to their total assets, which may also indicate that the degree of financing constraints reduces as moving from the highest to the lowest leverage ratio.

Panel III shows the results for the groups determined on the basis of the size. As size of the firm increases, the cash holding and market-to-book ratio decreases. Additionally, the larger firms have higher cash flows. This indicates that larger firms are financially less constrained. Panel IV and Panel V have the results for the groups determined on the basis of the cash holdings and KZ index, respectively. According to results, as cash holding of firm increases, the difference in cash and the market-to-book ratio increase, size, leverage and cash flow decrease. This indicates that firms that hold higher cash amounts are more constrained. Finally, results according to KZ index shows that, as the level of KZ index increases, cash holdings, difference in cash and cash flow decrease, and market-to-book and leverage ratios increase. Overall, the results regarding KZ index suggest that firms with higher KZ levels are more likely to be financially constrained.

Table 4 provides the results of mean values of proxies for US Nyse firms. Similar to Table 3, the mean values in columns A to E for five sub-samples are generated by dividing equally the total sample into five sub-samples using the entire distribution. Plus, observations are from lowest to highest as moving from A to E. The results reported for the groups that are sorted by dividend payout ratio, leverage ratio, size, cash holdings and KZ index respectively. According to table, US Nyse firms are less constrained when have higher dividend payout ratio and larger size, however, lower cash holdings, leverage, and KZ index levels.

Table 5 shows the mean values of the proxies of financing constraints for Mexico firms. According to Panel I, as firms pay out more amount of dividends to shareholders, their size become larger. Plus, cash holdings increases as firms get larger. Additionally, results from the leverage sorting criteria, as leverage increases, cash holding, market-to-book, dividends, and cash flows decrease, size and KZ index increases. Panel III in Table 5 provides the results for sorting order of size. As size of the firms increases, cash holdings, leverage, and cash flows increases. From Panel IV, it is seen that as firms getting financially more constrained, which as cash holdings increases, level of market-to-book, cash flows, size, and dividends increases. Finally, Panel V shows that high level of KZ index firms holding less cash, less cash flows, less market-to-book ratio, and less amounts of dividends. The Table indicates that financially less constrained firms have less cash, dividends, and size, while high leverage and KZ index.

Regression Results

Table 6 reports the results for US Nasdaq firms’ regression analysis estimating equation (1) for different classes of firms, to extend determined using financing constraints proxies and results from this testing procedure. The results are obtained by means of OLS pooled regression estimators. All models include time dummies. This procedure allow to test whether the direction of the relationship between the degree of financing constraints and the magnitude of the cash flow sensitivities of cash changes across sub-samples. Panel I shows the results for 15 classes obtained using dividend payouts as the ordering criterion. Class1 is the sub-sample I am left with when I remove all non paying dividends of observations in the distribution of dividends from the entire sample. Class 2 is the sample I am left with when I remove the lowest 5 percent of the observations among dividend paying firms. Similarly, Class 3 is the sample I am left with when I remove 10 percent of the observations and so on. Thus, Class 1 represents the relatively more constrained firms whereas firms in Class 15 are supposed to be the least constrained ones. That is, the degree of financing constraints should reduce as I move from Class 1 to Class 15 assuming that the dividend payout ration of firms is an appropriate constraints criterion. Second column of all Panels reports the level of the ordering criterion I am using to classify firms; third column reports the estimated cash flow sensitivities for all sub-samples; fourth column reports results from a test where the null hypothesis is that the cash flow sensitivity of cash estimated for the entire sample does not differ from the same parameter estimated for each sub-samples. When ordered by dividends, the null hypothesis is not rejected for class 12. All the remaining classes display a significantly lower or higher cash flow sensitivity of cash. This evidence suggests that the relationship between dividend payout ratio and the magnitude of the cash flow sensitivities of cash is inverse "U-shaped". To further confirm this result, I use the cash flow sensitivity of cash estimated for the largest class among all fifteen – namely, class 11 – to test the hypothesis that this parameter does not differ from that of all other classes of firms. Results are reported in the last column of Table 6.

Panel II of Table 6 reports the results where the sample is ordered by the leverage ratio of firms. Similar to dividend results in Panel I, results are supportive of an inverse U-shaped relationship between cash flow sensitivity of cash and leverage. Contrary to Panel I, Class 1 in Panel II represents the least constrained firms whereas firms in Class 15 are assumed to be the most constrained as their leverage ratio is the highest among all classes. Similar findings are reported in the following two panels, Panel IV and V, using cash holdings and KZ index as ordering criteria. Finally, Panel III gives the results when sorting by size criterion. The larger the firm, the less constrained is the firm.

Similar to US Nasdaq firms, I also re-estimate cash flow sensitivity of cash using the same proxies for US Nyse and Mexico firms. Table 7 and Table 8 provide results for US Nyse and Mexico firms, respectively. Taken all together, the results point an inverse U-shaped relationship between the cash flow sensitivity of cash and the sorting criteria for almost all the US and Mexico firms, which also indicates that the monotonicity condition is violated for all these criteria and countries.

insert Figure 2 here -

Robustness Checks

To check the robustness of my results I run a different measure of the empirical cash flow sensitivity of cash. The model forms with the insights from the literature on investment demand (e.g., Fazzari et al. (1988), Fazzari and Petersen (1993), and Calomiris et al.(1995)) and on cash management (Kim, Mauer, and Sherman (1998), Opler et al. (1999), and Harford (1999)). Plus, the model is the annual change in a firm’s cash to total assets as a function of capital expenditures, acquisitions, changes in noncash net working capital, and changes in short-term debt. All four additional variables are scaled by assets.

∆CashHoldingsi,t = α0 + α1 CashFlowi,t + α2 Qi,t + α3 Sizei,t + α4 Expendituresi,t

+ α5 Acquisitionsi,t + α6 ∆NWCi,t + α7 ∆ShortDebti,t + ℮i,t (3)

Because firms can use their cash holdings to pay investments and acquisition, α4 and α5 expect to be negative. The change in net working capital and the change in short-term debt should be controlled by cause of working capital to be a substitute for cash (Opler et al. (1999)). According to Almeida et al. (2004) it is noticeable to expect a larger estimate for α1 from the augmented equation (3) relative to that from equation (1) if cash flows really drive cash savings. This is because of adding alternative uses of funds to the model. To prove it, results for this augmented model of US NAsdaq and Mexico provided in Table 9. As it is expected, the coefficient α1 is larger while α4 and α5 are negative for US Nasdaq firms.

insert Table 9 here -

Implication for previous studies

My results suggest that the testing framework written by Almeida et al. (2004) should be rejected. More specifically, I reject the null hypothesis that the cash flow sensitivity of cash is monotonic in the main indicators used to proxy for financing constraints. The findings of my testing are consistent either with the hypothesis that the criteria used to classify firms are unrelated to financing constraints, or with the hypothesis that magnitude of the cash flow sensitivity of cash does not depend on financing constraints. If only the hypothesis that the criteria used to classify firms are unrelated to financing constraints holds, then future research should focus on finding a valid criterion to identify the firms’ financial status. Alternatively, if only the hypothesis that magnitude of the cash flow sensitivity of cash does not depend on financing constraints holds, then future literature should concentrate on further investigate the determinants of the cash flow sensitivity of cash.

Concluding Remarks

This paper studies the effectiveness of cash flow sensitivity of cash criterion as a measure of financing constraints in the US and Mexico.

I test to which extend the monotonicity condition, according to which the cash flow sensitivity of cash should always increase in the degree of financing constraints, is empirically relevant. For that purpose, by using alternative ordering criteria, I progressively remove the lowest classes from the sample, so that financing constraints should always go in the same direction, either increasing or decreasing. No statements have been done on whether financing constraints are decreasing or increasing through classes. I compare parameters of the cash flow sensitivity of cash estimated across the residual samples, and test the null hypothesis that these parameters do not statistically differ.

My findings show that monotonicity condition is empirically violated both in US and Mexico. Furthermore, I find an inverse "U-shaped" relationship between the cash flow sensitivity of cash and the extent of financing constraints. Such results, cast doubts on splitting samples of firms according to a priori measures of financing constraints faced by firms.

My results show that all financial indicators display a non-monotonic behaviour. In other words, proxies of financial indicators I use in this paper, which dividend payout ratio, leverage, size, cash holdings, and KZ index are not represent a good proxy for financing constraints. I come to the point that this might be either because the criteria used to classify firms are unrelated to financing constraints or the magnitude of the cash flow sensitivity of cash does not depend on financing constraints.