Potential Dividends and Actual Cash Flows: A Regional Latin American Analysis

We have examined the value that the market assigns to different components of the cash flow to equity including "potential" dividends. We study non financial publicly traded firms of five Latin American countries: Argentina, Brazil, Chile, Mexico and Peru during the period 1991-2007. The model includes the following variables: market value of equity, dividends paid, change in equity investment and change in liquid assets ("potential" dividends). These variables are regressed with actual equity value (time t) as the dependent variable, and the other variables as independent variables (including equity value) for the next period (time t 1). The applied tests have given robust results. The main conclusions of this work are three: Firstly, the market assigns less than one dollar to a future dollar for any of the variables studied as expected. Second, in particular, we found that the undistributed "potential" dividends or changes in liquid assets destroy value. I.e. the value of a dollar today in liquid assets in t 1 is negative. Third, we find that, a dollar invested in liquid assets has a negative Net Present Value and it is not a zero NPV investment. These findings confirm the problem of agency costs when not distributing cash flows. As a practical conclusion, the empirical evidence suggests that we should include, in the working capital, the liquid assets and leave out the practice of adding the book value of liquid assets. The only relevant cash flow is what an investor in fact receives from the equity investment: dividends and stock repurchases.


INTRODUCTION
We have examined the value that the market assigns to different components of the cash flow to equity including potential dividends. The theoretical and basic methodology of this work was developed by Vélez-Pareja and Magni (2009).
Our study is about non financial publicly traded firms in five Latin American countries: Argentina, Brazil, Chile, Mexico and Peru, during the period 1991-2007. The model includes the following variables: market value of equity, dividends paid, change in equity investment and change in liquid assets (potential dividends). These variables are regressed with actual equity value (time t) as a dependent variable and the other variables as independent variables (including equity value) for a next period (time t+1). Tests applied have given solid results.
We are concerned with the value the market assigns to the components of the Cash Flow to Equity, CFE, in particular to the change in liquid assets. Common practice assumes it is distributed among shareholders although in reality or in the typical financial model used to derive the cash flows, it is not because it is listed in the balance sheet.
There are three main conclusions in this work: in first place, the market assigns less than one dollar today to a future dollar for any of the variables studied as expected. Secondly, in particular, we found that potential dividends (changes in liquid assets) destroy value. In other words, the value of a dollar today in liquid assets in t+1, is negative. Third, we have found that, contrary to common knowledge and assumptions, a dollar invested in liquid assets has a negative Net Present Value (NPV) and they are not zero NPV investments. These findings confirm the agency costs of the problem when undistributed cash flows are kept. It also confirms that liquid assets should not be included in the cash flows while being listed in the balance sheet as current practice does.
According to Damodaran (2008, p. 106): "In the strictest sense, the only cash flow that an investor will receive from an equity investment in a publicly traded firm is the dividend that will be paid on the stock." We adhere to this definition. Usually practice and textbooks consider the working capital for cash flow calculation, excluding liquid assets. The net effect of this is the increase of the cash flows by a sum that is listed in the balance sheet; hence it is not effectively received by the shareholder. This amount is the increase in liquid assets. In section one we show how this occurs. As a result, equity and firm value are overvalued.
On the other hand, when we analyze the change in liquid assets, we find that it is part of the cash flow generated by the investment in liquid assets; hence, when included in the cash flows, they are discounted to present value. This is part of the above overvalued value. However, the same common practice assumes that liquid assets are zero NPV investments. For this reason, the actual book value of liquid assets is added to the present value of the cash flows.
Empirical evidence suggests that we should abolish the practice of exclud-ing from the working capital the liquid assets and adding the book value of liquid assets under the assumption that they are zero NPV investments. The only relevant cash flow is what effectively an investor receives from the equity investment: dividends and stock repurchases.
The article is organized as follows: in section one we define the problem and present a review of the current literature. In section two we describe in detail the theoretical model and propose the hypotheses. In section three we describe the data analyzed. In section four we analyze the data using ordinary linear regression OLS. In Section Five we present our conclusions.

BACKGROUND
In this Section we define the problem and present a literature review. What we find as a problem is an overvaluation of firm and equity value.

The problem
It is a well spread practice among authors, teachers and practitioners to include undistributed dividends as part of the Cash Flow to Equity (CFE) (and hence in the Free Cash Flow, FCF). This happens when in the process of calculating working capital only non cash items are included. Working Capital (WC), is defined as the difference between current assets (cash (C) + short-term investments (STI) + ac-counts receivable (AR) + inventories (Inv)) -current liabilities (accounts payable (AP)): The usual procedure to estimate the CFE in an indirect way departing from the income statement is the following: Where CFE is cash flow to equity; NI is net income; NFA t+1 is investment in fixed assets ( NFA t+1 = NFA t+1 -NFA t = investment in Fixed Assets t+1 -Depreciation t+1 ; and represents the so-called Net Capital Expenditure); WC is change in working capital; and D is change in book value of debt.
Another assumption usually made is that investment in liquid assets has a NPV equal to zero. For this reason, the present value of the cash flows is usually added to the book value of liquid assets.

A literature review
We agree with Vélez-Pareja and Magni (2009), on the "idea that potential dividends that are not distributed (and that are invested in liquid assets) should be neglected in firm valuation, because only distributed cash flows add value to shareholders" (p. 125). As these authors conclude, "(…) Cash Flow to Equity should include only the cash flow that is actually paid to shareholders" (p. 125). (dividends paid plus stock buy backs minus new equity investment). This is also supported by others DeAngelo, 2006, 2007;Fernández, 2002Fernández, , 2007Shrieves and Wachowicz 2001;Tham and Vélez-Pareja, 2004;Vélez-Pareja, 1999a, 1999b, 2004, 2005a, 2005b. Dechow, Richardson, and Sloan (2006) have found that: A common approach to corporate valuation is to discount the expected free cash flows generated by a firm's business operations. An implicit assumption with this approach is that the use of free cash flows is not important (i.e., whether they are retained as cash or distributed to debtholders or equityholders). Our results suggest that this assumption does not hold in practice. In particular, we find that cash retained by the firm tends to be less valuable because it is more likely to be associated with future declines in return on investment. Our results suggest that a superior approach to corporate valuation is to specifically forecast how much free cash flows will be retained by the firm and deduct this amount from the measure of free cash flows. Equivalently, a more appropriate approach is to directly discount forecasted cash distributions to debt and equity holders, after explicitly modeling the investment of retained cash flows. (p. 5) In a footnote they say: Most definitions of free cash flows add back the change in the cash balance. This assumes that the cash balance can be paid out to financiers and so the cash balance is a source of free cash flow. Our results suggest that increases in the cash balance are ge-nerally not paid out, but instead are reinvested in net assets. (p. 5) Boldfaces are ours. All these authors claim that cash flows should include only what debt and equity holders actually receive.
As mentioned by Vélez-Pareja and Magni (2009), some recognized authors such as Copeland et al. (1990Copeland et al. ( , 1994Copeland et al. ( , 2000; Benninga and Sarig (1997); Benninga (2006), Brealey and Myers (2003); Damodaran (1999Damodaran ( , 2006aDamodaran ( , 2006b) and most practitioners support the idea that the Cash Flow to Equity has to include undistributed dividends and the liquid assets are zero NPV investments. For instance, see Benninga (2006, p. 271), where he excludes cash and market securities, and p. 288, where he adds the book value of cash (cash is used as a plug to match the balance sheet and it includes cash in hand and market securities). See also Benninga and Sarig (1997, p. 36), where they define that "cash and marketable securities are the best example of working capital items that we exclude from our definition" of working capital. In p. 428-429 they state that "the value of non operating assets should be added to the value of the business to obtain an estimate of the whole firm", and "(…) non operating assets are effectively liquidity reserves that don't generate any positive NPV", Hence all we have to do is "to value such non operating assets at their current market price [italics in original]". Copeland et al. (1994), define operating working capital and exclude liquid assets, they express that "investment in short-term marketable securities is a zero net present value [italics added]" (p. 161) and its value is the book value of those assets. Copeland et al. (2000) say the same.
Vélez-Pareja and Magni (2009) present evidence from the current literature that dividends paid are valued more than potential dividends, and that Jensen's (1986)   , accepts the findings of these last authors and states that cash holdings create less value than dividends paid. In general, the idea is that one dollar of liquid assets in t+1 might have today a value of more than one dollar. Chu and Partington (2008) report repeatedly that, today, one dollar in dividends (in the future), is more worth than one dollar Consistent with imputation tax credits adding value to the dividend, 1 dollar face value of dividends was observed to have a market value significantly greater than its face value. The market value of the dividend varied depending on the proximity of observations to the ex-dividend date. (pp. [10][11][12]. Most of these authors (excepting Harford, 1999) consider that one future dollar in dividends (and in potential dividends) is worth more than or equal to one dollar today. This contradicts the elementary time value of the money concept: one dollar today is worth more than a dollar tomorrow. Or, putting it the other way around: one dollar tomorrow is less worth than one dollar today.
We interpret our results in a different way. Given our model, the coefficients are discount factors. Therefore, if we are considering $1 tomorrow, we expect that its value today be less than $1. In the next section we present the model we have tested in this work.
Vélez-Pareja and Magni (2009) present several arguments against the practice of including cash and cash holdings in the cash flows: Economic arguments underline that only flows of cash should be considered for valuation; theoretical arguments show how potential dividends lead to contradiction and to arbitrage losses. Empirical arguments, from recent studies, suggest that investors discount potential dividends with high discount rates, which means that changes in liquid assets are not value drivers (p. 123).
In short, there is a contradiction in considering cash flows something listed in the balance sheet and, at the same time, not received by claimholders.

The model
As mentioned in Vélez-Pareja and Magni (2009), our model is based on the financial theory of valuation of cash flows. We do not "fish out" variables to include in the model. The empirical evidence is tested with a theoretically correct financial model. The model is based on standard results of the corporate financial theory and, in particular, on Modigliani and Miller (1963). We use the following equation (5) where E is the market value of equity, Ke is the levered cost of equity, and CFE, the cash flow to equity. CFE is defined as Div -CS where Div is dividends paid and CS is the change in capital stock. 5 Dividends paid are what actually the equity holder receives from the firm and can put in the own pocket; CS is the increase (new equity investment) or decrease (repurchase of equity) in CS, Capital Stock.
Model (5) is a standard in valuation of cash flow (see, for example, Miller and Modigliani, 1961, eq. (2)). However, we would like to explain in detail the reason for E t+1 in the model. What the market does to value a stock is to discount the full stream of dividends from t+1 to ∞. As we cannot model that fact, we can assume that for year t+1 the market did exactly the same but from 5 ∆CS from the point of view of the equity holder, is positive when the firm buybacks equity and negative when the equity holder invests additional funds in the firm. t+2 to ∞. This is captured in the value of the stock (or market capitalization) for year t+1. Hence, when we include the dividends for year t+1 plus the market value at t+1, it is equivalent to having the full infinite stream of dividends from t+1 to ∞. That is what supposedly the market does. This is an incontrovertible formulation of the basic concept of time value of money and it is common knowledge. From the previous model we have designed an econometric model to test our hypothesis. The correct model is: What we are doing when using this model is to see the value of the firm from the point of view of the share holders. We have a model that values the equity, and hence the terms in the right hand side (RHS) of (6a), including the cash flow to equity, CFE and the value of equity in t+1. What eq. (6a) indicates is exactly the same as eq. (5) does. The difference is that instead of having a common discount rate, Ke, we consider that each element of the CFE might have a different discount rate. The common term (1/(+ke)) a discount factor) is broken down into three discount factors represented by the β's.
The definition of CFE in this model is clear and concise, and also supported by others authors. As Penman (2007) underlines, Owner's equity increases from value added in business activities (income) and decreases if there is a net payout to owners. Net payout is amounts paid to shareholders less amount received from share issues. As cash can be paid out in dividends or share repurchases, net payout is stock repurchases plus dividends minus proceeds from share issue [italics in original]. (p. 39) He also writes that "it is noncontroversial that the price of a security is expressed as the 'present value' of the expected future payoffs to holding the security" (Penman, 1992, p. 466), where "payoffs" unambiguously refers to "the payoffs for equity securities" (p. 466). His notions of "net cash flow to shareholders" (p. 239) or "net dividend" (p. 241) are consistent with our notion of cash flow to equity: In order to test the hypothesis that LA, the change in liquid assets, is irrelevant or destroys value, we falsify the correct model introducing the change in liquid assets, LA, as follows: In the two previous equations CS corresponds to change in Capital Stock (in the data we have changed the sign to the change in CS: an increase in CS means extra investment from the equity holders and it is an outflow; a decrease in CS means a buyback of stocks and it is an inflow to the share holders), Div is dividends paid, LA is undistributed potential dividends and equal to the change in liquid assets as shown above in equations (1) to (4). E means equity market value, and LA is liquid assets. In this model a proprietary approach is followed, where E t is measured as the declared market capitalization.
This model is consistent with the standard finance theory, but we do not intend to claim that it is fully explanatory, nor we try to make use of it for forecasting purposes. The models are meant to provide information on the relevance of the independent variables, in particular, the relevance/irrelevance of LA t+1 to value creation. To this end, the various betas are to be interpreted as the discount factors for the independent variables. In particular, β4 is the discount factor for change in liquid assets, i.e. it represents the value at time t of one dollar of LA t+1 available at time t+1.
Coefficients for the variables are interpreted as follows: an increase of $1 in any variable will be equivalent to an increase of $β in value in t. If β >0 (let us say, 0,80) then we say that one increase of $1 in a given variable (in t+1 ) it is equivalent to an increase of 0,80 of the value in t. The coefficient is a discount factor that brings back to t (present value) that $1 in t+1. This is exactly what we teach about the time value of money in our courses. On the contrary, if β<0 (in particular β 4 , the coefficient for LA in the falsified model) it means that it destroys value. Hence, if LA decreases (negative) and its coefficient is negative, at the end it would mean that it creates value instead of destroying value. The logic of this lies in the fact that if we invest the funds in cash in hand or in a very low return investment, shareholders will push the management to distribute that money instead of leaving it, say, in the bank. They consider that they can invest it at a higher rate.
What do we expect from coefficients β? We expect β 1 and β 2 to be positive and less than 1. We expect β 4 to be zero or negative. We expect β 3 to be negative or positive. This last comment deserves an explanation: as CS is depicted as a cash flow, when it has a negative value it means that the shareholders have increased their investment in the firm. When it is positive it means that the firm has bought some equity back and it is an inflow for the equity holders. If β 3 is negative, it indicates that the market is recognizing a value driver in the extra equity investment: this means that -$1 of equity investment in t+1 is equivalent to-β 3 today, and this is a positive value (-β 3 × -$1). At the same time it would imply that the market sees the repurchase of equity as a negative value driver, even if it represents an inflow to the equity holder. On the contrary, if β 3 is positive, it would mean that the market values equity repurchases as a value driver, and is not aware that a new equity investment might be a value driver.
In order to normalize the data and avoid problems of size, currency, time, etc., we will divide each variable by the book value of total equity (BVE) in t. We also deflate items in year t+1 with the corresponding inflation rate. As all variables will be divided by book value of total assets, the ratio E/BVE will represent Tobin's Q. These independent variables are the normalized proxies for components of equation (5).
Equations (6a) and (6b), the correct model, will be affected by BVE and ESTUDIOS GERENCIALES Vol. 25 No. 113 • Octubre -Diciembre de 2009 deflated by (1+π t+1 ) where π is inflation rate as follows: The falsified correct model is modified as follows: peat, we do not fish out variables. The variables are those stipulated by the correct model. In the falsified model we introduce a new ("strange" to the model) variable as they do in practice and recommend according to the literature review. And our hypothesis is that this new variable should be non-statistically significant.

The hypotheses
Our hypothesis may be phrased in a strong or in a soft version (similar to Vélez-Pareja and Magni, 2009): "Strong version: we expect β 6 [or β 4 in this article] to be statistically zero (or close to zero). This means that investors will try to set down the value of Div pot(t+1) [or LA t+1 in this article] by discounting it with an infinite, or at least at a very high discount rate, because they do not consider (undistributed) potential dividends relevant for valuation. Another condition might be that β 6 [or β 4 in this article] be negative and this would mean that potential dividends destroy value.
Soft version: we expect Div pot(t+1) [or LA t+1 in this article] to be evaluated much less than the actual dividends. In econometric terms, we expect β 6 [or β 4 in this article] to be much smaller than β 5 [or β 3 in this article]" (pp. 145-146). A negative β 4 is included in the soft version of our hypotheses and this would mean that potential dividends destroy value. In terms of NPV, it means that investment in liquid assets does not have a zero NPV.

DESCRIPTION OF DATA
The source of data was primarily Economatica, and in case of abnor- where all variables are now meant to be divided by book value of equity, the independent variables are also deflated. The deflation is made with t+1 , the inflation rate for year t+1. With this model, the value of the firm depends on the cash flows the owners of equity expect to receive in the future, and on potential dividends as well.
The components of model (7a) and (7b) depict the shareholders' prospect of future value and cash flows to equity. The dependent variable is Tobin's Q, whereas the independent variables will be a percent of the book value of equity in t. With this model we try to measure how much value is created for a given value of the independent variables in the following period.
It should be here noticed that the models do not include an intercept. The reason is that we start from a correct financial and theoretical model. When we use an intercept it is because we fish out variables until we have come to a proper model. In this case, we re-mal behavior of the data we double checked with other sources. Secondary sources for checking purposes and completion of missing data were: We had to fix some criteria to define how we would extract the data from the database, as follows: 1. Closing date: we found diverse dates for the fiscal year of the financial statements. However, all data were collected on December 31st. We have assumed that at this date there would be no signaling effect from dividends announcements.

The use of consolidated or not consolidated financial statements:
we decided to use consolidated financial statements.

Currency and inflation adjustments:
we extracted the financial statements in the local currency and without adjustments for inflation. The model provided "normalization" of the data dividing by BVE and deflating future values (in t+1) to the present period, t.

Dividends paid in cash:
we decided to include in the sample only those firms (and countries) that reported the dividends effectively paid in cash in the cash flow state-ment. This was the reason why Colombia and Venezuela were excluded from the sample.
In our analysis we only included non financial firms and stocks with a high market liquidity index. The financial industry (banks, insurance firms, and pension funds) was excluded because we do not have dividends paid from the cash flow statement. Additionally, there is not a clear cut distinction between cash in hand and cash as an inventory to operate. The concept of liquid assets in the financial firms is somewhat different from the non financial firms. In the financial firms cash plays a different role as in non financial firms. We could even say (although it is not part of the study) that, by definition, financial firms do not hold liquid assets (CDs, pension funds investments, etc.). Cash in the financial firms plays the role of inventory.
We cleaned up the data and eliminated all those cases that did not have information for some of the variables. We distinguish between no data and zero as value of the variable. The result of this cleaning up was that the final sample was seriously unbalanced. See Appendix A for a complete description of the data collected.
A statistical description of the collected variables follows including the deflation of terms in t+1 for the total sample is in Table 2 Table 3.

RESULTS
In this section we show the different regression analyses performed with the data collected.
We have run several tests to measure the consistency and robustness of the data: 1. Homoscedasticity: we applied a test for robustness and it eliminated some variables from the model 2. Autocorrelation (passed)

Data panel that eliminated the same variables as in 1
The sequence of tests applied to the data is shown in Appendix B. The results for the models after the test for robustness follow. Table 4 shows the statistics for the falsified model. Notice we are including the potential dividends, as Damodaran calls them. As can be seen in Appendix B, first we tested a model including the constant or intercept. It also can be noticed here, that the constant is not significant. The final results after applying the robustness test are in Table 4.  As we can see, β 4 , the coefficient for ΔLA t+1 /BVE t , is significant and has negative sign. This behavior of ΔLA t+1 /BVE t , is identical in all regressions (with robust adjustment) excluding (one by one) the extreme values for E t /BVE t , that are greater than its mean plus three standard deviations. This is what was expected to happen and we can see the trend in Exhibit A4, in Appendix A.
We ran data panel tests for the correct and falsified models with and without intercept. The test for the two models with random effects, maximum likelihood effects (MLE) and without intercept, is displayed in the Table 5.
We can observe once again that E t+1 /BVE and ΔLA t+1 /BVE t are significant at 5%.
In summary, we show the results with OLS with the robustness test and Data Panel with MLE in Table 6. dependent variable plus 3-sigma and in all the tests ∆LA is consistently statistically significant and negative. Our sample included all the valid values; we did not delete any outlier. Our conclusion is that funds kept in cash and/or in short term investments destroy value as expected. This compares with some current literature mentioned in section one that assigns a positive value today or even greater than 1 to one dollar of liquid assets in next year.
As it can be noticed we can say that the market is averse to the ∆LA as a value driver because β 4 is negative.   Table 6. Summary of OLS and data panel tests We can observe that ∆LA is statistically significant and its coefficient is negative. We must say that we have tested the model with the data that included the highest fifteen values (one by one) above the mean of the correlated to a lower (or loss of) value. This is what a negative coefficient for ∆LA means.
In tests using OLS and robust estimates with data within ±14 standard deviations dividends are statistically significant but the value of the coefficient is greater than 1, which is contrary to the financial theory of time value of money, although this result is consistent with the reported findings mentioned in the literature review.
Moreover, we see that ∆LA destroys value, since each dollar in ∆LA in t+1 destroys -$1,13 of value today. The rejection of Div and ∆CS might be interpreted as if the market did not value dividends and that the market did not fully measure the impact of that extra investment in the value of the stock. What market values most is the expected value of E. One possible explanation is shown in Table 2 where, on average for the dividends (0,06 on the average) and ∆CS (-0,05 on the average) are so low compared with equity (1,90 on the average) that seems that the market does not value dividends as value drivers. In Table 7 we can observe the return on equity market value. In other words, our strong version hypothesis is rejected and this means that the Net Present Value of the investment in liquid assets is not zero, contrary to what generalized knowledge says. Now we analyze selected ranges for the discount factor, DF, and out of this point we can infer selected ranges for the implicit discount rate, DR. We constructed the   Table 8. Range of discount factor (β) and implied return The meaning of the coefficients, as we have said before, is a discount factor that implies a discount rate.
In Table 9a we show the discount factor for the correct model for the independent variables and for ∆LA from the falsified model. Potential dividends and actual cash flows. A regional Latin American analysis In terms of discount rates for each of the components of the CFE we can say that the implicit cost of equity for each component is: Implicit discount rates are shown in Table 9b. In the same fashion, β 4 , the coefficient of ∆LA t+1 /BVE t implies a discount rate of −188,55%. This is an expected result and it might be interpreted as an aversion from the market to the potential dividends.
We examined the consistency of the coefficients and observed values for the theoretically correct model. We assume that the equity value can be calculated either considering each element of equation (1) has a discount rate, or that the sum of them (E t+1 + CFE t+1 ) has a single discount rate (Ke). The weighted average of betas (discount factor) is A, as follows: (9a) Equation (9a) means that in the model we assign a beta (a discount factor) to each component of equation (6a) in the RHS. In the left hand side (LHS), we consider the three elements of equation (6a) as a whole and discount it with an average discount factor, A.
The weighted average of betas should be the average discount factor for Ke.
In this case we accept it is a proper discount factor if: β 1 ×E t+1 +β 2 ×Div t+1 +β 3 ×CS t+1 E t+1 +Div t+1 +CS t+1 (9b) And restricted to 0<A<1, where A is the average discount factor for the correct model. An analysis of the values for A gives the following results (Table 10). This mean implicit Ke compares with the implicit Ke in the coefficient of 17% and with the mean return of 24%.
What is the behavior of the change in Capital Stock and Liquid Assets? This is relevant because it gives us an idea of the sign of the cash flow. A negative sign in ∆CS means that shareholders increased their invested capital. A negative sign in ∆LA means that the firm recovered investment in market securities and cash to use it in alternative investments or distributed it to the claim holders.
In Table 11a we show the number of cases in the sample that have negative observations. The sum of the ∆CS and ∆LA gives an idea of the relevance of the data (negative or positive) in the results. This is shown in Table 11b. A× E t+1 +Div t+1 +CS t+1 =β 1 ×E t+1 +β 2 ×Div t+1 +β 3 ×CS t+1 iv t+1 +CS t+1 =β 1 ×E t+1 +β 2 ×Div t+1 +β 3 ×CS t+1 When we observe the results in the previous tables we conclude that almost 77% of the observations show an increase in capital stock. And not only this, the absolute amount is much larger that the buybacks (237,07 vs. 58,66). The fact that the change in capital stock has been rejected from the model could be interpreted as if the market does not recognize that the increase in capital stock represents an opportunity to increase the value of the firm.
On the other hand, we can see that the decrease in liquid assets is much lower, almost 44%. In this case, the absolute value of decreases compared with the increased is much lower: (102,62 vs. 200,12). The fact that change in liquid assets is rejected from the model might be interpreted as if the market perceives an increase of liquid assets as a value destroyer.
As a summary: 1. We have chosen a model that is robust and that has been long time ago supported by current literature and practice. It is based on the findings of Miller (1958, 1963). The model is common to all present value analysis. We have tested a well proven and utilized model in the current financial valuation practice.
2. The model is statistically significant and has an R 2 of at least 0,92.
3. We have analyzed a database that is significant: 3.482 observations. 4. Given the linear structure of the model, we consider that OLS is an appropriate estimation tool for the model. We tested the regression results with additional tests such as autocorrelation 6 and homoscedasticity. As a final test we applied data panel to test the data.
5. Given that the model is theoretically correct, we have not added any additional independent variable. The model has only the variables included in the correct version. We have falsified the correct model with an independent variable that is used precisely for testing its relevance. We tested the intercept and it proved to be non significant.
6. Our prior expectation before gathering and analyzing the available data is partially confirmed with the analysis. The model was proposed in Vélez-Pareja and Magni (2009).

CONCLUDING REMARKS
We have found the empirical evidence that liquid assets destroy value. This confirms Jensen's (1986) free cash flow agency problem. We also found that the investment in liquid assets is not perceived by the market as a zero NPV investment.
On the other hand, we found that the market discounts the CFE with discount rates below 100%. Contrary to what is found in some literature, one dollar in the future is worth less than one dollar today as expected 6 Some tests cannot be applied, such as the Granger Test because we need a time series and our data have under the best of circumstances a sequence of 11 observations and it can be applied only to pairs of variables. See Appendix A, Table A4. according to the time value of money concept.
In summary, we can draw several conclusions: 1. This exploration is a confirmation of the Jensen's (1986) free cash flow agency costs problem. Excess cash should be distributed because its retention destroys value.
2. Investment in liquid assets does not have a zero NPV. It has a negative NPV.
3. Market values more the expected price of the share than dividends. In fact, the independent variable Dividends (Div) is rejected from the regression.
4. Analysts should not include change in Liquid Assets (LA) as cash flow, because it is a fiction that contradicts the empirical evidence of reality.

5.
There is an overvaluation of cash flows when the analysts include in the value the actual book value of liquid assets when it is assumed that liquid assets are zero NPV investments. While the change in liquid assets is discounted at the cost of capital when included in the FCF, that change, with contrary sign is discounted at usually a lower rate when considered as part of the cash flow of the liquid assets investment.
6. If analysts include ∆LA in the cash flow, this is to say that ∆LA is fully distributed, it should be consistency between the financial statements and the cash flows. In other words, the equity book value should show the fact that there is either a new equity investment, or a repurchase of equity.
The fact that idle cash or cash invested in low return investments destroys value is not solved by including those amounts that are listed in the balance sheet in the cash flows as if they were distributed to shareholders. The solution is to effectively distribute the funds to them.
As a practical and general conclusion, the empirical evidence suggests that we should include in the working capital the liquid assets, as well as eliminate the practice of adding the book value of liquid assets under the assumption that they are zero NPV investments. The only relevant cash flow is what effectively an investor receives from the equity investment: dividends and stock repurchases.
Our conclusions are in line with the thinking of DeAngelo and DeAngelo (2006,2007), who have devoted much of their researches mainly explaining that only distributed cash flows produce value.

DESCRIPTION OF THE SAMPLE
The final number of observations by country and by year is as in Table  A1.   Descriptive statistics for the sample are found in Table A3.

EXHIBIT A9
Histogram for E t / BVE t

Distribution of observations by country, year and firm
We analyzed the observations by year, country and firms and found that the number of firms by country with an uninterrupted sequence of observations, and with a minimum of 4 in a row, is more than 50% of the sample.

Peru 38
Total 264 Table A4. Firms by Country with uninterrupted sequence of minimum 4 in a row Table B1 shows the statistics for the theoretical (correct) and the falsified models. Observe that we are including the potential dividends, as Damodaran calls them. First, we tested a model with constant included. As can be noticed, the constant is not significant.  Table B1. Correct and falsified models tested for significance (OLS) with intercept.

APPENDIX B
As we are using a theoretically correct model, we consider it is complete and therefore an intercept should not be included. In Table B2 we show the OLS test without intercept. After applying tests for homoscedasticity we applied tests of robustness (Table B3). Intercept, Div/BE and ∆CS/BVE are rejected as non significant. After this test we ran the robust test again without an intercept. The results follow in Table B4.  Table B4. Correct and falsified models tested for significance (robust, no intercept) The previous tests show that the approach in order to model the theoretical model without intercept is correct. However, as seen before, two independent variables were rejected. Observe that ∆LA/BVE is significant and negative, which means it destroys value.
We ran the two models with the significant variables with and without intercept and with the robust test (Table B5). Again, intercept is rejected as non significant and coefficient for ∆LA/ BVE indicates value destruction.
Now the regression without intercept is as in Table B6.