Panels and Cross-sections 1 Paul A. Gompers Empirical Topics in Corporate Finance February 19, 2009
Panel and Cross Sectional Data Today look at panel and cross sectional data. Covers lots of finteresting ti papers and data sets. Methodological issues arise in the cross section and we will deal with those in a variety of settings.
Agenda Look at methodological papers as well as applications of cross sectional and panel data. Hopefully this examination will provide insights i into how to approach many of fthe most interesting problems in CF.
Estimating Standard Errors in Finance Panel Data Sets: Comparing Approaches Mitch Petersen RFS 2009
Papers Contribution Examines a variety of approaches to estimating standard errors and statistical significance in panel data sets Interesting look at a variety of papers published from 2001-2004. 2004 Only 42% of papers adjusted standard errors for possible dependence d in residuals. Many different approaches. Which are correct under what circumstances.
Overview OLS standard errors are unbiased when residuals are independent and identically distributed. Residuals in panel data may be correlated by firmspecific effects that are correlated across time. Firm effect. Residuals of a given year may be correlated across different firms (cross sectional dependence) Time effect.
Paper s Approach Simulate data that has either firm effect or a time effect. Test various estimation techniques. See how they deal with the simulated data. Then takes regression approaches to actual data and compares them.
Firm Fixed Effects Assumption of OLS is that cross product matrix has only non-zero numbers on the diagonal. Figure 1 Example of a firm effect. Cluster standard errors by firm.
OLS vs. Clustering by Time vs. FM with Firm Effect Simulate 5000 samples with 5000 observations. 500 firms and ten years of observations. Let the residual and independent variable variance due to the firm effect vary between 0 and 75%. 500 clusters by firm.
OLS vs. Clustering on Firm vs. Table 1 FM with Firm Effect Compare average coefficients, st. dev. of coefficient estimates, % significant, average SE clustered and % significant with clustered SE. Vary how much of the independent variable variation is due to firm effect and how much of the residual variation is due to firm effect. Figure 2 Compare OLS, Clustered by firm, and Fama-McBeth. Table 2- Fama-McBeth
OLS vs. Clustering by Time vs. FM with Time Effect Simulate 5000 samples with 5000 observations. Let the residual and independent d variable variance due to the time effect vary between 0 and 75%. Not this is the situation that FM developed FM for. Clustering will be by the 10 years.
OLS vs. Clustering by Time vs. FM with Time Effect Table 3 Compare OLS and Clustering by time. OLS does pretty poor job. Table 4 Using FM to estimate.
OLS vs. Clustering by Time vs. FM with Firm and Time Effect In many typical examples, could have both a firm and time effect. Figure 6, typical structure t with both. Can cluster by firm and time together. See Samuel Thompson s 2006 working paper for math.
OLS vs. Clustering by Time vs. FM with Firm and Time Effect Simulate 5000 samples with 5000 observations. Let the residual and independent d variable variance due to firm and time effect vary Table 5 Compare OLS, with and without firm dummies, Clustered by firm and time, GLS, and FM.
Real Data Table 6 Look at asset pricing application. Equity returns on asset tangibility. Different methods matter. OLS and firm clusters do poorly. Time and dfirm clustering and FM work well. Seems to say that for returns may be more affected by a time effect.
Recommendations Think about the structure of the panel data structure. What is the likely l source of dependence. d Comparing different methods may provide additional information about the research question.
Real Data Table 7 Capital structure regressions. OLS, clustering by time, and FM do poorly. Clustering by firm and clustering by firm and time do well. Says that within corporate finance a lot of the effects seem to have firm level persistence.
Market timing and capital structure Malcolm Baker Harvard Business School Jff Jeffrey Wurgler NYU
Research question Why do similar firms have different capital structures? t Temporary fluctuations in market value have a lasting impact on capital structure outcomes
Overview A new fact: Temporary fluctuations in market value have a lasting impact on capital structure outcomes Possible explanations: (1) Trade-off theories Taxes, costs of financial distress, and agency lead to an optimal leverage ratio (2) Pecking order Adverse selection dominates other considerations, leading to a pecking order Neither appears to explain the new fact Consistent with the idea that managers are motivated by market-timing
Some motivation Prevailing market values are the most important empirical determinant of financing decisions Equity issues IPOs: Loughran, Ritter, Rydqvist (1994), Pagano, Panetta, Zingales (1998) et al. SEOs: Asquith and Mullins (1986), Marsh (1982) et al. Debt issues and repurchases Debt: Marsh (1982) et al. Repurchases: Ikenberry, Lakonishok, Vermaelen (1995) et al. and capital structure is the sum of past financing decisions (accounting identity) So, capital structure may depend on the historical path of market valuations
Sample Trace the evolution of capital structure as firms mature Require a known IPO date Natural starting point for the historical path of market valuations And, can trace the determinants of capital structure as firms mature Capital structure E t / A t Compustat coverage from 1969 to 1998 Compustat coverage from 1969 to 1998 Empirical approaches IPO time Calendar time
Capital structure changes Changes in the equity ratio come in two forms Δ E A t = e A t + Δ E A e t (1) New issues and repurchases Active change in capital structure t (Table 3A) (2) Retained earnings Passive change in capital structure (Table 3B) Examine the determinants of each Market-to-book, asset tangibility, profitability, and size Table 3
Measures of M/B Summarize the historical path of valuations with a Summarize the historical path of valuations with a single statistic: (1) Maximum market-to-book ratio The highest year-end value from the IPO through t-1 (2) Weighted average market-to-book ratio The weights are the amount of external finance (debt plus equity) raised in each year from the IPO through t- 1 Financing events represent a practical opportunity to change capital structure
Temporary fluctuations in M/B and capital structure Cross-section regressions in Table 7 Cross section regressions in Table 7 t t efwa t t u B M b B M b a A E + + + + = x 1 k 2 1 1 (1) Include controls x t t efwa t t u B M b B M b B M b a A E + + + + + = x 1 k 0 3 2 1 1 ( ) Fixed assets intensity Profitability i i Firm size (2) b 2 captures the impact of temporary fluctuations in market value fluctuations in market value Control for endpoints at IPO and t-1
Is M/B a measure of mispricing? M/B predicts stock returns... Basu (1977, 1983) Fama and French (1992) LSV (1994) partly because of errors in expectations La Porta (1996) LLSV (1997) But M/B captures both mispricing and legitimate growth prospects p... Growth could be correlated with agency, asymmetric information, financial distress costs so control for the level lof M/B Starting point (IPO), ending point (t-1), both
Some robustness checks (1) Market value capital structure (2) Industry effects IPO year effects (3) Fama and French (2000) Five profitability lags (4) Outliers included Mature firms included (5) Tobit Table 8.
Economic significance Figures 1 Large relative to the other determinants Asset tangibility and size die Profitability emerges
Possible explanations (1) Trade-off theories Taxes, costs of financial distress, and agency lead to an optimal leverage ratio Ancillary prediction: Temporary fluctuations in market-to-book to (or anything else) should have a temporary impact (2) Pecking order Adverse selection dominates other considerations, leading to a pecking order Ancillary prediction: Temporary increases in marketto-book should lead to lower cash balances or higher future investment (3) Market timing i Managers believe they can time the market
Trade-off theories Taxes, costs of financial distress, and agency lead to an optimal capital structure Market-to-book could be connected to one or more inputs to the trade-off, and Costly financial i ldistress Debt overhang Agency Perhaps tax benefits may have some persistence, but Adjustment costs temporary fluctuations in market-to-book (or anything else) should have a temporary impact else) should have a temporary impact Table 9.
Long-term impact Temporary fluctuations in market-to-book to have a lasting impact on capital structure The half-life lif of the initial iti effect is at least ten years (Table 9) Hatched bars are the five percent lower bound
Pecking order Adverse selection dominates other considerations, leading to a pecking order Market-to-book is related to investment opportunities, but High market-to-book means investment opportunities exceed internally generated funds and debt capacity extra equity raised should be spent or at least be earmarked for future investment So, temporary increases in market-to-book should lead to lower cash balances or higher future investment t
Cash balances Temporary fluctuations in market-to-book to also have a lasting impact on cash Increases in market-to-book t khave a permanent impact on cash balances (Table 10) No lasting impact on investment (Table 11)
Market timing A variant of Myers and Majluf (1984) Like Myers-Majluf: Majluf: Managers have the incentive to try to time the market because they care more about existing shareholders Investors react to financing decisions, and this adverse selection dominates other considerations, so there is no optimal capital structure Unlike Myers-Majluf: Managers think that they can successfully time the market, believing (1) Shares are occasionally under or overvalued (2) Investors underreact to new issues
Other evidence of timing Managers admit to timing the market Graham and Harvey (2000) It looks like they re trying Marsh (1982) Pagano, Panetta, and Zingales (1998) Although investors recognize it... Asquith and Mullins (1986) they underreact Ritter (1991), Loughran and Ritter (1995) Baker and Wurgler (2000) Is it actually successful? A separate debate Market-timing attempts affect capital structure
Conclusions Managers try to time their equity issues and this influences capital structure outcomes: (1)Low leverage firms raised external finance when valuations were comparatively high High leverage firms raised external finance when valuations were comparatively low (2)Temporary fluctuations in market-to-book lead to lasting changes in capital structure and cash balances (3) Trade-off theories and pecking order do not appear to explain the results (4) Market timing fits the new fact and old facts
Testing Trade-Off and Pecking Order Predictions i about Dividends id d and ddebt Fama and French RFS 2002
Agenda Look at two competing theories of capital structure and dividends. Pecking order Trade-off theory Utilize Fama-McBeth techniques. Very important technique to understand.
Motivation Most previous studies are pure cross section or small panels. Results can be wildly overstated. Cross correlation can reduce standard errors. Correlation of the residuals across firms are ignored. Auto correlation can reduce standard errors. Panels can have residuals correlated across years. Fama-McBeth gives robust standard errors in these types of situations. Particularly when there are multiple observations on the same firm and you have unbalanced panels potentially.
Methodology Run a series of cross sectional regressions. Can be run annually, monthly, daily, etc. Only depends upon number of unique observations you have. Report the average coefficient i and test significance by using the standard deviation of the time series of coefficients. Also can report number of positive and negative coefficients. Recognize that there may be firm persistence. Arbitrarily argue that there is a need to increase t- statistics critical value by 2.5x, i.e., 5.00 to get significance.
Pecking Order Dividends Less attractive for less profitable firms, large current and expected investments, high leverage. Leverage Depends if one period or care about future financing. Volatility. Lower leverage for firms with large future investments. Low dividends and low leverage.
Trade-off Model Bankruptcy costs Taxes Agency costs Adjustment costs.
Independent Variables ET/A Pretax earnings to assets V/A = Market value to book value (future investments) t RD/A R&D to assets Dp/A Depreciation over assets Ln(A) log assets as proxy for volatility.
Dependent Variables Dividends D/A Dividends over assets Leverage Market leverage L/V Book leverage L/A
Regressions Data from 1965-1999. 1999 Table 1 Dividend payout ratio Use target leverage from leverage regression in Table 4. Table 3 Level of leverage. Strong evidence of pecking order. More profits yields less debt.
Sorting the sorts Table 5 Sort firms based on dividend paying or not and leverage. Find that low leverage nonpaying firms have better investment opportunities and more equity issuances. At odds with pecking order.
Conclusion Neither theory wins out. Perhaps a third theory is at work. Perhaps both have merits. Interesting techniques.
Conclusion Interesting paper. Highly technical with attention to detail. Firms don t take full advantage of potential tax benefit savings from debt.
Understanding the Determinants of Managerial lownership and dthe Link Between Ownership and Performance Himmelberg, Hubbard, and Palia JFE 1999
Motivation Lots of studies show non-linear relationship between firm value and inside ownership. Mork, Shleifer, and Vishny (1988). McConnell and Servaes (1990). Hermalin and Weisbach (1991). Problem is that ownership may be endogenous. All these studies are large cross-sections.
Game Plan Examine the determinants of equity holdings. In particular, does the availability of information to monitor and track company affect inside ownership? Create panel data set of inside ownership to control for unobserved firm factors that affect ownership. Ideally would use IV estimation, but no good instruments. t
Data Find firms on Compustat that have complete data on sales, book value, and stock prices from 1982-1984. 1984 Randomly select 600 firms. Get ownership data from proxies. Unbalanced panel. Table 1.
Summary Statistics Collect data on: Number of top managers (from proxies). Ownership of ft top managers (from proxies). Table 2 and Figure 1.
Determinants of Ownership Stock price volatility. Managerial risk aversion. Demsetz and Lehn (1985). Interpretation? Ease of monitoring. Size - ln(sales) Capital intensity - K/S. Measures scope for discretionary spending. R&D intensity -R&D/K Advertising intensity - Ad/K
Determinants of Ownership Ease of monitoring. i Cash flow. Gross investment rate.
Results Tbl Table 4(A) - Examines total t linside id ownership. Fixed effects matter. Size has differential impact in different specifications. Increasing capital intensity, less inside ownership. Increase volatility, less ownership. Increase R&D, increased ownership with fixed effects.
Results (2) Table 4 (B) - Examines average inside ownership per top manager. Size increasing in pooled regressions. No effect with FE. Increasing capital intensity, ty, less inside ownership per manager. Increase volatility, less ownership per manager. Increase R&D, increased ownership with fixed effects.
Ownership and Firm Value Create three categories of inside ownership based on Morck, et al. m1 = ownership if ownership<5% =.05 if ownership >5%. m2 = 0 if ownership<5% =ownership level minus 5% if ownership >5% and <20%. =.20 if ownership > 25%. m3 = 0 if ownership<25% =ownership level minus 25% if ownership >25%.
Results Table 5B. Without fixed effects, get non-linear relationship between ownership and Q. With FE, becomes insignificant. Even SIC dummies and other controls don t help.
Instrumental Variables Approach Perhaps we can solve the problem with IV estimation. Use size and size squared as well as volatility as instruments for ownership. Results in Table 6. Once again, without FE get inverse relationship between een ownership and Q. Goes away with FE.
Conclusions Inside ownership appears to be related to monitoring/agency environment. The relationship between ownership and Q appears to be driven by unobservable firm factors. Argue that each firm is choosing ownership optimally to maximize i Q.
Concerns Equity ownership perhaps not best measure of incentives. Need dto consider other elements of compensation and relationship to executives wealth. Why look at ownership of all top executives? Is looking at CEO better or worse?
Concerns Why not control for other governance provisions? Structure t of board of directors. Anti-takeover provisions. Institutional i ownership.
Extensions and Questions CEO ownership and compensation over time. CEOs in our sample own substantially ti more than seen in typical, large public firms. When do the CEO begin to sell equity in large quantities? Is there a performance difference? Do ownership and incentives even matter?
Extensions and Questions (2) Relationship to other corporate governance mechanisms. Boards of directors. Insider vs. outsider. Detailed description of types of outsiders. Role of block shareholders. h Changes over time. Who gets added and why? Real effects of directors?
Managing gwith Style: The Effect of Managers on Firm Policies Bertrand and Schoar QJE 2003
Motivation Most theories have treated managers as being undifferentiated. Real differences in managers may have impact on firm policies. Investment and financial. Look at: Financial leverage, investment-cf sensitivity, organizational strategy (R&D, advertising, diversification, and cost-cutting), and performance.
Key Issue How do you separate out manager effects from firm effects? Strategy: t Collect long panel of top execs at US firms. Look at those who change firms and estimate the Manager Fixed Effects
Sample Forbes 800 data from 1969 through 1999 and Execucomp from 1992 through 1999. Need managers to be in a firm for at tleast tthree years. Find 600 firms with over 500 managers that spend at lease three years at two different firms. Table I Firm sample characteristics. Table II data on managers and transitions.
Methodology Look at dependent variable as a function of firm and manager characteristics Utilize manager fixed effects. Extract out these fixed effects. Table III Look at investment and financial policies on CEO characteristics. CEOs matter. CFOs matter for financial policy. Table IV Look at organizational strategies. Once again, CEOs matter.
Individual Characteristics Look at two important characteristics. MBA. Birth year. Did you grow up in Depression. Tbl Table IX. Find that Depression babies are more conservative and MBAs are more aggressive.
Conclusion Manager characteristics matter. Where you were before. MBA and birth year. Nice paper. Clever idea and clever test.
US cross-listings i and the private benefits of control: evidence from dual-class firms Craig Doidge JFE 2003
Motivation Can you use the decision of dual-class foreign firms to list on the US on NYSE or Nasdaq to get estimates of value of private benefits?
Background Coffee (1999) and Stulz (1999) show that firms issuing ADRs on US causes stock price increases. Commitment to better governance.
Data Compile panel sample of 745 non-us firms that have dual class shares. 137 cross list on US 75 on NYSE or Nasdaq Table 1 - Summary
Voting Premium Analysis Look at voting premium of superior voting class shares. Tbl Table 2 Table 3 Look at voting premium based upon type of ADR and the level of protection in the domestic market. Utilize OLS and random effects model.
Announcement Period Returns Figure 1 Look at voting premium and cumulative return difference between low and high voting shares. Table 6 CAR analysis. Estimate market model from day -244 to day - 6. Also, just subtract return on the high voting shares minus low voting shares. Table 7 Regression of LMH on protection of shareholders variables.
Conclusion Bonding hypothesis seems to have support. Better protection leads to lower voting premium. Higher level of disclosure of ADRs leads to lower premium. Better protection of shareholders in domestic markets leads to lower premium.