Data

We use two types of data: recommendation reports by equity analysts, and historical financial statement data and stocks prices of publicly traded Mexican firms.

Recommendation reports by equity analysts

Recommendation reports were obtained during the first quarter of 2011 from ISI Emerging Markets (ISI).² All equity reports available on ISI, with an explicit buy, hold, or sell recommendation, were retrieved. The sample, 72 reports, from 40 firms, was restricted to non-financial firms trading on the Mexican Stock Exchange (MSE).³ Appendix 1 shows firms in the sample and presents some of their general characteristics including industry, annual revenues, and an indication of whether the firm was part of the MSE index or not. The MSE index contained 35 equity as of April 2011, with 6 financial firms and 29 non-financial firms. Of the 40 firms in our sample, for the regression analysis, we use the 29 non-financial firms that were part of the MSE index, implying that our sample represents the Mexican Stock Exchange market. Equity analysts are from J.P. Morgan, HSBC Global Research, Credit Suisse, among others.

Accounting data and stock prices

Accounting data and stock prices, from 1995 to 2011, are from Economatica. We use quarterly data, with variables from the income statement and the statement of cash flow annualized every quarter. That is, quarterly data are converted to trailing 12 months (TTM) data by adding figures from the current and previous three quarters to avoid within-year seasonality effects. This sample, used for the regression analysis, includes about 1150 firm-TTM observations.

Most preferred financial ratios

We use the recommendation reports sample to select the most preferred financial ratios. We find that equity analysts in the sample use standard templates, with the same financial ratios, regardless of the firm or industry covered. We assume they include financial ratios of their preference in their firm's standard template. Thus, we count the financial ratios used by analysts in the sample. While simple, this way of selecting the most used (preferred) financial ratios is straightforward and objective. Table 1 provides the frequency of use. Financial ratios are categorized as accounting ratios, valuation ratios or multiples; cash flow ratios, and other ratios. Accounting ratios are further grouped into profitability ratios and margins, and leverage or debt management ratios.⁴

We define "most preferred" ratios as those that are used by 50% or more of the analysts in the sample.⁵ This reduces the set from 36 ratios used in practice by equity analysts, according to the sample, to fourteen most preferred ratios, summarized in Table 2.⁶

Regressions

Our main interest is to examine whether the most preferred financial ratios used by equity analysts might predict stock performance. We use a two-way random effects model for our regression analysis. The regression model we estimate takes the form:

where RET is stock return at quarter t + n ; n is 4 for 1 year future returns or 8 for 2 years. FI are alternative sets of most preferred financial ratios (specified in the next section) at quarter t and log (MVE ), the logarithm of the market capitalization of the firm's equity, and BETA , the measure of systematic risk, act as control variables. Variables for Eq. (1) are computed using accounting data and stock prices obtained from Economatica, from 1995 to 2011, for firms included in the MSE index.

We control for serial correlation and heteroscedasticity using the Newey-West correction of standard errors.⁷

Table 3 provides descriptive statistics and Table 4 provides Pearson correlation coefficients of the main variables used in the study. These correlation coefficients help to avoid the inclusion of highly correlated variables in the models, as we explain next.

We have five FI categories (e.g., valuation, profit and margins, etc.). To reduce multicollinearity, alternative models are run with one variable per category. To avoid redundancy of information, and to have the number of variables/models manageable and results feasible to discuss in this article, we reduce the number of most preferred ratios from Table 2 following two criteria. (1) Financial ratios with high correlation coefficients are excluded from the models and (2) when two variables are similar, the most theoretically robust variable is included in the models.

First, ROE is statistically correlated with all financial ratios, but FCFY and Sales g (Table 4), so it is excluded from the regressions. Between EBITDA margin and EBIT margin, similar ratios, the former is chosen since it contains additional valuable information (depreciation and amortization). FV to EBITDA is chosen over price to earnings because the former is a more comprehensive measure than P /E , capturing both the relationship between market price and earnings and leverage. Net Debt to EBITDA is also excluded as the correlation between FV to EBITDA and Net Debt to EBITDA is very high (0.92, Table 4), and FV to EBITDA is used by 9 out of 10 equity analysts. Finally, dividends per share is considered redundant for the regressions as we have dividends yield and price to earnings per share. Thus, we use eight models for the empirical tests.⁸

Model 1: FV EBITDA, EBITDA margin, dividend yield, net debt to equity, EPS growth,

Model 2: FV EBITDA, EBITDA margin, dividend yield, net debt to equity, sales growth,

Model 3: FV EBITDA, EBITDA margin, FCF yield, net debt to equity, EPS growth,

Model 4: FV EBITDA, EBITDA margin, FCF yield, net debt to equity, sales growth,

Model 5: FV EBITDA, ROI, dividend yield, net debt to equity, EPS growth,

Model 6: FV EBITDA, ROI, dividend yield, net debt to equity, sales growth,

Model 7: FV EBITDA, ROI, FCF yield, net debt to equity, EPS growth, and

Model 8: FV EBITDA, ROI, FCF yield, net debt to equity, sales growth.

Most preferred financial ratios

Tables 1 and ² show the most preferred financial ratios used by equity analysts. This differs from what a typical financial textbook presents, especially pertaining to accounting ratios. Accounting ratios are usually classified in textbooks as profitability and margin ratios, asset turnover or efficiency ratios, liquidity, and leverage or management ratios. Equity analysts in this sample do not use liquidity or asset turnover ratios; instead, they concentrate on profitability and margins, and leverage. Probably equity analysts use cash flow related ratios as better substitutes for turnover and liquidity ratios.

Net debt to EBITDA and net debt to equity are the debt management ratios preferred by analysts. Textbooks, in general, do not mention net debt (defined as debt minus cash and cash equivalents), or net debt to EBITDA. This may be because net debt is used more by practitioners than by academics and instructors according to ^{Bates, Kahle, & Stulz (1999)}.

Concerning valuation ratios or multiples, price to earnings (PE), firm value (FV) to EBITDA, and price to book value (P /B ) are by far the most preferred multiples. Among cash flow ratios, dividend yield and free cash flow yield are most preferred.⁹ In general, cash flow ratios, are as common, in terms of frequency of use, as earnings and price related ratios. This supports the idea that cash flows are at least as important as earnings. Furthermore, we observed that, with one exception, all analysts in the sample included cash flow statements projections in their reports, implying that cash flow metrics are highly valued by equity analysts. Concerning this, ^{Call, Chen, & Tong (2009)} suggest that earnings forecasts of analysts accompanied by cash flow forecasts could be more accurate than earnings forecasts without cash flow forecasts.

In the "Other" category, earnings per share (EPS), sales growth, and dividends per share are most preferred. This is consistent with the findings in ^{Matsumoto et al. (1995)} who surveyed analysts in the U.S. In their study, analysts ranked EPS and sales growth rates as the most important financial ratios for both retailers and manufacturing firms. Finally, according to results in Table 2, EBITDA is common across ratio categories. This is surprising, as EBITDA could be a misleading metric (^{Stumpp, 2000}).

Financial ratios and stock returns

Table 5 provides regression results for all eight models with t + 4 stock returns (i.e., 1 year future returns) as the dependent variable. The market return (BETA ), to capture the firms' systematic betas from the market model (^{MacKinlay, 1997}), and the natural logarithm of market capitalization (log (MVE )) are included to control for risk and size.¹⁰ In addition, we controlled by industry using industry dummies.

With the exception of the estimate for net debt to equity (DEQ ) in models 7 and 8, all coefficients are statistically significant at 1% level. Furthermore, signs of coefficients are consistent across all eight models in Table 5.¹¹

The estimate for FV to EBITDA (FV EB ) is negative, consistent with the view that managers are pressured to decrease multiples every period so their stock be perceived by investors as less expensive. EBITDA margin (EB M ) and return on investment (ROI ) are positive (e.g., higher profitability ratios and margins increase stock returns). Dividend yield (DivY ) is negative; when dividends are paid, stock price declines. Furthermore, firms with high dividend yields are generally large, stable, mature firms with low growth prospects. This result is consistent with our sample: mature, low growth stocks that are part of the Mexican Stock Exchange index. These equities are considered to be "value" stocks with the bulk of their value coming from dividend payouts rather than from price growth.

For FCF yield (FCF Y ), the negative sign could be explained by the free cash flow hypothesis (^{Jensen, 1986}), which states that a higher free cash flow for a given price signals higher agency costs. The positive sign of net debt to equity implies that higher leverage increases firm value, as long as leverage is not unreasonably and unsustainably high (DEQ , for models 7 and 8, is statistically significant at 5%, unlike all coefficients/models, significant at 1%).¹² Finally, the signs of EPS growth (EPS g ) and Sales growth (Sales g ) are positive, as expected given the direct pressure managers have on these metrics every quarter they disclosure financial statements.

Overall, these results provide evidence that the most preferred financial ratios used in practice by equity analysts following firms on the MSE have predictive power; with the signs of coefficients as predicted by financial theory. The marginal predictive power of these models (marginal to an asset pricing model), as expected, is not very high (model 0, in Table 5, provides results of CAPM and size as control variable). It would not be possible to have a model with financial ratios having a high explanatory power, which would be the main purpose of asset pricing models, as stock price changes are very noisy.

As robustness tests (untabulated results), we also run regressions on the same variables, but using growth rates rather than levels. Results with growth variables are consistent with results with level variables. Furthermore, to test the persistency of results, we run regressions with the same variables using t + 8 stock returns (i.e., 2-years future returns). We did not find evidence on the predictive power of these ratios as time passes. That is, the magnitude of the selected financial ratios at time t could explain part of the variation of stock returns in t + 4, but when new information arrives (1 year later), the new information is incorporated into stock prices.