Real Estate Ownership by Non-Real Estate Firms: The Impact on Firm Returns Yongheng Deng and Joseph Gyourko 1 Zell/Lurie Real Estate Center at Wharton University of Pennsylvania Prepared for the Corporate Real Estate Alliance Seattle, WA, Meeting January 30, 1999 Preliminary: Please do not quote for attribution without permission. 1 Deng is Post-doctoral Research Fellow at the Zell/Lurie Real Estate Center at Wharton. Gyourko is Professor of Real Estate & Finance and Director of the Zell/Lurie Center at Wharton. Both authors gratefully acknowledge funding from the Corporate Real Estate Alliance and the Research Sponsors Program of the Zell/Lurie Center. This report was prepared for the Corporate Real Estate Alliance. All errors are the responsibility of the authors.
46 1998 OFHEO REPORT TO CONGRESS
Executive Summary Recent estimates (guesstimates is perhaps a better word) suggest that the amount of real estate owned by firms and corporations is approaching $2 trillion. A debate is emerging around the issue of whether non-real estate firms should own so much property. One prominent industry researcher, Peter Linneman, argues in a recent paper that it is harmful for non-real estate firms to commit so much of their scarce capital to investments outside their core competencies. Linneman and others also suggest that high cost of capital firms in particular should avoid committing to ownership of relatively low return buildings. Pure finance theory does not support the latter contention in particular. Capital budgeting principles imply that each investment project considered by the firm should be evaluated at its own opportunity cost of capital. Cash flows from risky projects should be discounted more severely than the cash flows from less risky projects regardless of whether the firm itself has a high or low cost of capital. That is, the true cost of capital depends upon the use to which the capital is put. Thus, theory implies that high cost of capital firms should not suffer a penalty when they use their scarce capital to invest in and own relatively low return real estate as long as they receive a return high enough to compensate them for the risks of real estate ownership. Of course, the real world often does not conform perfectly to theory. Linneman s contentions could hold if either of the following factors is relevant: (a) investors believe that firms are more likely to earn higher risk-adjusted returns on investments in their areas of core competence than they are in generic real estate; if so, they could penalize firms that stray too much from investments in core competency projects; or (b) investors i
in high cost of capital firms especially do not fully perceive the lower risk associated with real property investment; if so, they effectively discount all projects at the company, not the project, cost of capital contrary to what finance theory suggests should drive corporate value. There is no amount of debate, however clever, that can resolve this issue. The problem is an empirical one, so there is a need to go directly to firm-level data to determine whether too much real estate ownership/investment harms a firm. A capital budgeting-based investigation probably will not be fruitful due to onerous data requirements. Project-by-project data on risk and return would be needed for each investment to see if the firm receives an appropriate risk-adjusted return in each case. Another approach, and the one taken in this paper, is to examine a firm s returns to see if they are correlated with the degree of real estate ownership/investment. More specifically, a capital asset pricing model (CAPM) is estimated for 381 firms across eight industries, many of them relatively capital-intensive. This controls for systematic risk across firms, the factor that asset pricing theory suggests drives return differences across firms. The study then examines the non-systematic, or idiosyncratic, component of return to see if it is associated with the concentration of real estate ownership. More specifically, we test for whether the idiosyncratic component of firm return is less for firms with relatively high levels of real estate ownership. If returns are lower for firms with relatively high concentrations of real estate ownership, then the evidence is consistent with some penalty being imposed by investors. The results show a consistent negative relation between the idiosyncratic component of firm return and the degree of real estate ownership. That is, within an ii
industry it is the case that a firm s return is lower if it has a relatively high fraction of its total assets (measured by book value) in real estate (also book value). The impact varies across industries, being greatest in the electronics industry. A pooled regression across industries finds a statistically and economically significant negative impact of relatively high real estate ownership on return. While there are no meaningful differences in returns for firms with very similar real estate concentration measures, for firms with concentration measures ten percentage points above average, implied excess annual returns are about one percentage point lower. This amounts to just under 10 percent of cumulative compound excess return over a ten year holding period. Specifications that allow for threshold effects yield the largest impacts. For example, a firm with a real estate-to-total asset ratio that is above the median for all 381 firms in the sample has about a 4 point lower average annual return than a firm with a real estate-to-total asset ratio that is below the sample median. However, this particular result may be driven by a relatively few outliers, an issue that will be examined more closely in future versions. Unfortunately, we are unable to determine whether the penalty is being imposed for reasons suggested by Linneman or investor ignorance of the true risk profile of real estate ownership. This limitation aside, this is the first study of which we are aware to document a statistically significant negative relation between firm returns and the degree of real estate ownership over a long holding period. Work in progress is attempting to determine whether other correlates might explain the correlation. However, given how noisy return data are and the difficulty of finding any predictable correlation with returns, even such a preliminary result should be taken seriously. iii
46 1998 OFHEO REPORT TO CONGRESS
Introduction For many years operating companies around the world generally have owned their real estate assets. In the United States alone it is estimated that corporate users own nearly $2 trillion, or roughly half of all commercial property. Companies own not only their production facilities, but frequently their offices, warehouses, and retail outlets. Although many of these properties are suitable for a broad range of users, these operating companies choose to commit their scarce capital to the ownership and operation of real estate, rather than re-deploying this capital to their core operating businesses. In this study, we test the hypothesis that high levels of corporate real estate concentration in total assets for a non-real-estate company has a negative impact on the company s annual returns. We adopt an empirical approach that is familiar in the finance literature to analyze abnormal returns. Specifically, the analysis is conducted in the following two stages. In the first stage, we estimate the companies excess returns using a Sharpe-Lintner Capital Asset Pricing Model (CAPM). In the second stage, we analyze the estimated idiosyncratic components of returns from the CAPM to see whether they are influenced by the level of corporate real estate concentration. The empirical results suggest a statistically and economically significant relationship between the level of corporate real estate concentration (measured as the fraction of book value of assets held in real property form) and the company s total returns. The strength of this result varies across industries, being strongest among electronics firms. The finding also holds in a pooled regression across industries using a specification that allows for different industry intercepts. The Data The data sets used in this empirical study include the NYSE, AMEX, and NASDAQ monthly stock files maintained by the Center for Research in Security Prices (CRSP), and merged with the Standard & Poor s COMPUSTAT annual industrial files. 1
The CRSP Monthly Stock file provides detailed information on individual securities, most importantly including return data. COMPUSTAT is a database of financial, statistical, and marketing information. It provides more than 300 annual Income Statement, Balance Sheet, Statement of Cash Flows, and supplemental data items on more than 7,500 publicly held companies. Based on the information from COMPUSTAT, we compute a variable labeled as RC (Real Estate Concentration) ratio which empirically measures corporate real estate concentration for a non-real estate firm. Specifically, RC ratio is the ratio of the company s Property, Plant, and Equipment (including building at cost plus land and improvements) divided by the company s Total Assets. Both numerator and denominator are book values. Using book numbers reduces endogeneity problems that can bias the estimations reported below. This real estate concentration measure is then merged into the return data from the CRSP monthly stock files. 2 Due to a variety of data limitations in both data bases, we confine our analysis to the period from 1984 to 1993. In addition, firms are included only if they have at least 60 months of consecutive monthly returns data, the standard in the finance literature for estimating stable betas. Furthermore, each firm must have balance sheet information about property, plant, and equipment, total assets, as well as the company s year-end equity market capitalization information. After data cleaning and merging, the final sample includes stock returns and balance sheet data from 381 firms in eight major industries. The eight industries covered are: 1. Food and Kindred Products Industry; 2 We also have experimented with three other concentration measures. One is very similar to RC, except that its numerator reflects the value of buildings at cost less accumulated depreciation. Use of this measure yields results very similar to those reported below. The other two measures are ratios of the book value of real estate (gross and net of accumulated depreciation) to total property, plant, and equipment. Use of these two measures yields qualitatively similar results, although missing data is more of problem here, so that the findings often are not statistically significant due to the smaller number of observations. 2
2. Printing, Publishing and Allied Industry; 3. Chemicals and Allied Products Industry; 4. Primary Metal Industry; 5. Industrial and Commercial Machinery, Computer Equipment Industry; 6. Electronics, Other Electronics, excluding Computer Industry; 7. Transportation Equipment Industry; 8. Instruments, Photo Goods, Watches Industry. A list of all firms used in the study is available from the authors. The Test Methodology Our test is based on an approach similar to the cross-sectional regression approach suggested by Fama and MacBeth (1973). The basic idea is that for each security, the total return can be broken down into idiosyncratic and systematic components. It is crucial to control for systematic risk (or beta) as theory suggests that is the primary reason why returns vary across firms. After controlling for risk differences across firms, we then examine whether the idiosyncratic component of return (i.e., that part not related to market risk) is related to the company s real estate concentration level. Specifically, in the first stage, we run a series of regressions using the Sharpe- Lintner CAPM model, such thatýþ# ERET = α + β EMKT + ε (1) it i i it it where ERET is the excess return over the risk-free return, which is measured as the difference between the company s monthly holding period return and the 30-day T-bill returns reported by CRSP monthly stock files; EMKT is the excess return on the market portfolio, which is measured as the difference between the monthly return on the CRSP value-weighted market portfolio and the 30-day T-bill return reported by CRSP monthly stock files; α is the idiosyncratic component of the excess return; β is the systematic component of the excess return; ε is an error term following a standard normal distribution; i indexes for firm, and t indexes for time period measured in month. 3
In the second stage, we test whether the idiosyncratic component of return is related to the degree of real estate ownership by regressing the estimated α i on the real estate concentration measure, such that ü f IND, RC, SIZE α i i i i = 1 6 (2) where IND is a vector of industry dummy variables, RC is our empirical measure of a company s real estate concentration level, as described in the previous section; and SIZE is the size of a firm, measured by the year-end capitalization reported by CRSP. 3 We suspect that the Alpha-RC relationship, if it exists at all, may vary across industries. Therefore, equation (2) is tested by industry. However, the limited number of firms which satisfy our data screening criteria in each industry may affect the power of our test. Alternatively, we also test the equation (2) using a pooled sample by combining all the industries together and controlling for industry fixed effects in our test. The Empirical Results Summary statistics on the key variables used in the analysis are reported in Table 1. Recall that these data are for the 1984-1993 time period and that returns are measured in excess of the CRSP value-weighted market return. There are a number of noteworthy features about these data. First, the mean monthly excess return across all industries is over eight-tenths of one percent (0.0085) or about 10.7 percent per year. There is a fairly wide range across industries, ranging from 0.4 percent per month in electronics to 1.4 percent per month in chemicals. However, the variation in excess returns within industry is even larger. For example, monthly excess returns vary from -52.8 percent to 229 percent in the Electronics industry, and from 77.1 percent to 524.5 percent in the Computer Equipment Industry. 4 3 Size is included because recent research has shown that it often is related to differences in firm returns. While there is no theoretical explanation for this correlation, we control for it in case it is related to the degree of real estate correlation. In any event, it could be that firms which reach a certain size tend to accumulate real estate. 4 Obviously, these are not averages but extreme values that occur in at least one monthly observation. 4
The average idiosyncratic component of monthly excess return, which soon will be the focus of our interest, is a very small 0.047 percent. As with industry returns, the variance about this mean is quite large, with the standard deviation being 1.16 percent. We test whether or not this variation in idiosyncratic return is related to the degree of real estate-related property, plant, and equipment that a company holds as a percentage of its total assets. That ratio, RC in our terminology averages just under 18 percent across all industries. The variation in this variable across and within industries is less than for returns. Finally, most of our industries are relatively high beta sectors of the economy. Only the Food industry has an estimated beta less than one and it is close at 0.96. The Electronics and Transportation Equipment industries carry the greatest systematic risk, with their betas exceeding 1.2. Our estimates are in line with previous research my Gibbons and Fama on systematic risk across industry groupings. Figure 1 plots, for each firm in the Electronics industry, each firm s mean excess return (EMKT) against its real estate concentration measure (RC). The raw data clearly do not indicate any relationship between excess returns and the degree of real estate concentration in assets in this industry or any other. 5 Figure 2 presents a very different picture with its plot of the idiosyncratic component of excess return in the Electronics industry against the measure of real estate concentration. The regression line draws the clear negative relationship. That is, once systematic risk is controlled for, the higher is an electronics firm s real estate concentration, the lower its idiosyncratic return component. Figure 3 then plots the level of the idiosyncratic component of return for firms in the Electronics industry by degree of real estate concentration. There is a fairly large difference depending upon whether the firm has a relatively low or relatively high degree of real estate ownership as a percentage of total assets. For those firms with RC values below the median for the industry, the idiosyncratic component of return is 0.496 percent per month, versus 0.848 percent per month for those with RC values above the 5 Plots for other industries are available upon request. None reveal any relationship between excess returns and real estate concentration. 5
industry median. The difference is a fairly wide 0.4 percent per month. The divergence is even wider if one divides the firms into those with very high RC values in the top quartile for the industry versus those in the bottom three quartiles. [See the bar graphs on the right side of Figure 3.] The difference in idiosyncratic return components widens to nearly 0.6 percent per month. This is both statistically and economically significant. Figure 4 then shows that this relationship is not peculiar to the Electronics industry. When we pool across all eight industries, a similar pattern results. The difference is widest in the Electronics industry, but it exists in other industries as well. Across all firms, the difference is 0.3 percent per month for those with RC values above the sample median versus those firms with RC values below the sample median. The difference widens to 0.4 percent per month when we compare firms in the top quartile to those in the bottom three quartiles. [Note: The idiosyncratic component of monthly excess returns are computed using the estimates from model 5 and model 6 of Table 3 for the pooled sample, and from models 5 and 6 of Table A-14, and Table A-16, for the Electronics Industry and the Photo Goods Industry, respectively. For the pooled sample, the Alpha for those firms with real estate concentrations equal or below the median (i.e., RC 50%) is computed as the weighted average of the industry fix-effect intercepts (weighted by the number of firms in the industry) adjusted by the log size effect. The Alpha for those firms with real estate concentrations above the median (i.e., RC > 50%) is computed using the previously computed Alpha (for RC 50%) adjusted by the estimated coefficient for the dummy variable indicating RC > 50% (i.e., -0.003056 from Model 5 in Table 3). Similarly, we compute the Alpha for those firms with real estate concentrations equal or below the top quartile (i.e., RC 75%) and the Alpha for those firms with real estate concentrations above the top quartile (i.e., RC > 75%) using the estimates reported in Model 6 from Table 3.] While these differences between high and low RC firms are fairly large, it is worth emphasizing that the implications are different for firms with similar degrees of 6
real estate concentration on their books. This can be seen by examining the coefficient from the specification that allows RC to vary continuously (first column, Table 2, pooled regression). The estimate of 0.00932 implies that the idiosyncratic component of excess return is only.00009 lower for a firm with a 19 percent versus an 18 percent RC ratio (i.e., for a small change about the mean value of the independent variable). When compounded over a year, this amounts to less than one-tenth of a percent in overall excess return. Thus, while the result is statistically significant, it is not so economically. When we consider the implication for a firm with a 28 versus an 18 percent RC ratio value, the implications are that average annual returns are about 1.1 percentage points lower for the high RC firm. This amounts to nearly 10 percent of excess returns on average. Thus, the data seem to indicate that there are threshold effects around which the impacts on return become much larger. It is clear that for firms with the average amount of real estate investment, small differences about that mean level are not associated with meaningful differences in return. Ten percentage point differences in real estate concentration ratios begin to be associated with meaningful differences in excess returns over time. And, comparing very high versus very low real estate concentration differences are associated with quite large return differences over time. However, these differences may be driven by a few outlier firms. Future versions of the paper will test more rigorously for this. Figures 5 and 6 then present how these threshold differences translate into annual excess returns for the Electronics industry and Pooled industry samples, respectively. Estimated annual excess returns are calculated based on equation (3) AERET 12 1 6 (3), = ERET + 1 1 where AERET is the annual excess returns, and ERET is computed based on equation (4): ERET = αü + ü β EMKT (4), 7
where α is the estimated idiosyncratic component of excess return from figures 4-6, β is the average of systematic risk estimated from the first stage regression, and the EMKT is the excess market returns. Figure 5 shows that for the Electronics industry, where the relationship is strongest, a firm with a relatively low degree of real estate concentration measured a percentage of total firm assets has a 4.3 percentage point higher annual excess return on average (i.e., 5.01-0.68). The penalty associated with real estate ownership is even wider if one divides the sample into those with RC values in the top quartile for the industry. In that case, those firms in the bottom three quartiles have implied annual excess returns that are 7.6 percentage per annum higher on average (i.e., 4.37-(-3.28)). Figure 6 reports the analogous findings for the Pooled sample of firms across all industries. These results indicate that on average a firm with real estate-to-total asset ratio above the median for all 381 firms in the sample has a 4 percentage point lower average annual return than a firm with a real estate-to-total asset ratio that is below the sample median. If a firm has a real estate-to-total asset ratio in the top quartile of all 381 firms in the sample, its average annual return is 5 points lower than the returns earned by the firms with real estate-to-total asset ratio in the bottom three quartiles of the sample. Conclusion There is a growing debate as to the wisdom of non-real estate companies investing large amounts of scarce capital in real estate. Corporate finance theory does not suggest firms should be penalized for investing in real estate per se as long as they obtain the proper return for their investment. However, it could be that investors believe that companies are more likely to reap higher risk-adjusted returns on investments within the core competency of the firm. It also could be that investors do not understand the true risk profile of real estate investments, and, therefore, penalize high cost of capital companies for such investments. 8
If so, one should see some evidence that firms with relatively high degrees of real estate investment have lower returns. In this paper, we find such evidence. It is strongest in the Electronics, but holds in a pooled regression of firms across eight industries. Our differences are large when one divides the sample into high and low real estate concentration firms. The return difference is small for small differences in RC. Return differences become meaningful for larger differences in concentration of at least ten percentage points (nearly a standard deviation in the data). The size of the difference across the full range of RC value is so large that one reasonably should suspect that real estate is proxying for some other firm trait. Future versions of the paper will experiment with added correlates to see if the results are robust. However, the difference in the nature of the scatter plots reported in Figures 1 and 2 is strong evidence that there is a real relationship between the idiosyncratic component of firm return and the firm s degree of real estate ownership relative to other firms in its industry. 9
Table 1. Summary Statistics of Key Variables by Industry Industry No. of Obs. Means, (Standard Deviations), and [Minimum, Maximum] ERET ALPHA BETA RC RATIO SIZE Food 3,633 Publishing 3,000 Chemicals 7,322 Primary Metal 2,873 0.01268 0.00517 0.95995 0.22243 3.85712 (0.0954) (0.0102) (0.2114) (0.1157) (6.3906) [-0.410, 0.935] [-0.017, 0.029] [0.668, 1.609] [0.028, 0.641] [0.009, 33.00] 0.00807-0.00013 1.10759 0.15809 1.35414 (0.0916) (0.0046) (0.2592) (0.0542) (1.5261) [-0.473, 0.655] [-0.010, 0.011] [0.629, 1.670] [0.046, 0.262] [0.008, 6.16] 0.01376 0.00689 1.11944 0.19539 3.57710 (0.1143) (0.0139) (0.3327) (0.1503) (6.5842) [-0.616, 1.605] [-0.018, 0.094] [0.213, 2.215] [0.038, 1.606] [0.008, 30.17] 0.00556-0.00335 1.12698 0.21033 0.78346 (0.1257) (0.0089) (0.3531) (0.1986) (1.1991) [-0.556, 0.995] [-0.018, 0.021] [-0.130, 1.688] [0.008, 0.921] [0.009, 4.74] Computer Equipment 7,270 0.00809-0.00024 1.19624 0.16427 1.56910 (0.1520) (0.0144) (0.4846) (0.0644) (7.6492) [-0.771, 5.245] [-0.049, 0.050] [-1.617, 2.275] [0.016, 0.346] [0.002, 61.71] Electronics 7,425 0.00428-0.00444 1.2418 0.16562 1.63491 (0.1326) (0.0074) (0.3699) (0.0655) (5.2033) [-0.528, 2.294] [-0.020, 0.013] [0.409, 2.106] [0.022, 0.400] [0.005, 34.62] Transportation Equipment Photo Goods, Watches Pooled Samples 3,955 3,767 39,245 0.00741-0.00139 1.21466 0.16279 2.56880 (0.1100) (0.0088) (0.2896) (0.0585) (4.9032) [-0.523, 1.151] [-0.019, 0.027] [0.513, 1.788] [0.039, 0.278] [0.009, 24.72] 0.00734-0.00078 1.14724 0.17286 1.86006 (0.1252) (0.0097) (0.3089) (0.0652) (3.0058) [-0.520, 2.130] [-0.024, 0.021] [0.612, 1.995] [0.045, 0.292] [0.007, 14.36] 0.00852 0.00047 1.15337 0.17929 2.22272 (0.1242) (0.0116) (0.3611) (0.1064) (5.6892) [-0.771, 5.245] [-0.049, 0.094] [-1.617, 2.275] [0.008, 1.606] [0.002, 61.71] NOTES: ERET Monthly Excess Returns over 30-Day T-bill Returns. ALPHA Idiosyncratic Component of the Monthly Excess Returns estimated from the first stage regressions. BETA Systematic Risk of the Monthly Excess Returns estimated from the first stage regressions. RC RATIO Property, Plan and Equipment (building at cost plus land) divided by Total Assets, all at book value. SIZE Year-End Equity Market Capitalization, measured in billion $. 10
Table 2. Second Stage Estimation without Size Control for Pooled Samples Variable Model 1 Model 2 Model 3 Food 0.007315 0.006824 0.006308 (3.412) (3.530) (3.386) Publishing 0.001366 0.001445 0.000968 (0.631) (0.689) (0.475) Chemicals 0.008804 0.008457 0.008023 (5.506) (6.099) (6.141) Primary Metal -0.001523-0.001457-0.002112 (0.683) (0.683) (1.032) Computer Equipment 0.001265 0.001223 0.000774 (0.841) (0.877) (0.585) Electronics -0.002920-0.002851-0.003429 (1.936) (2.016) (2.593) Transportation 0.000116 0.000174-0.000422 Equipment (0.059) (0.093) (0.234) Instrument, Photo 0.000823 0.000967 0.000478 Goods, Watches (0.408) (0.498) (0.256) RC Ratio -0.009320 (1.967) RC > 50% -0.003054 (2.733) RC > 75% -0.003996 (3.207) R 2 0.1319 0.1402 0.1465 NOTES: t-ratios are in parenthesis. Sample size is 381. RC Ratio Property, Plan and Equipment (building at cost plus land) divided by Total Assets. RC > 50% A dummy variable, indicating that the company s RC Ratio is above the 50 percentile of the industry level. RC > 75% A dummy variable, indicating that the company s RC Ratio is above the 75 percentile of the industry level. 11
Table 3. Second Stage Estimation with Size Control for Pooled Samples Variable Model 4 Model 5 Model 6 Food 0.008475 0.008177 0.008539 (1.940) (1.914) (2.006) Publishing 0.002492 0.002757 0.003131 (0.582) (0.649) (0.740) Chemicals 0.009945 0.009787 0.010217 (2.443) (2.451) (2.565) Primary Metal -0.000470-0.000228-0.000083 (0.114) (0.056) (0.021) Computer Equipment 0.002285 0.002413 0.002736 (0.623) (0.665) (0.757) Electronics -0.001883-0.001641-0.001435 (0.506) (0.445) (0.391) Transportation 0.001244 0.001488 0.001744 Equipment (0.297) (0.359) (0.422) Instrument, Photo 0.001923 0.002249 0.002595 Goods, Watches (0.465) (0.549) (0.635) RC RATIO -0.009294 (1.959) RC > 50% -0.003056 (2.732) RC > 75% -0.004050 (3.239) LOG SIZE -0.000086-0.000100-0.000163 (0.305) (0.355) (0.583) R 2 0.1321 0.1405 0.1473 NOTES: t-ratios are in parenthesis. Sample size is 381. RC Ratio Property, Plan and Equipment (building at cost plus land) divided by Total Assets. RC > 50% A dummy variable, indicating that the company s RC Ratio is above the 50 percentile of the industry level. RC > 75% A dummy variable, indicating that the company s RC Ratio is above the 75 percentile of the industry level. Log Size Log of Year-End Equity Market Capitalization. 12
Appendix Tables: Results by Industry Table A-1. Second Stage Estimation in Food and Kindred Products Industry Panel (a). Without Size Control Variable Model 1 Model 2 Model 3 Intercept 0.006881 0.005929 0.005590 (1.916) (2.295) (2.704) RC Ratio -0.007437 (0.546) RC > 50% -0.001406 (0.401) RC > 75% -0.001483 (0.383) R 2 0.0090 0.0049 0.0044 Panel (b). With Size Control Variable Model 4 Model 5 Model 6 Intercept -0.021197-0.022149-0.021238 (1.871) (2.025) (2.044) RC Ratio -0.000489 (0.038) RC > 50% 0.000604 (0.182) RC > 75% -0.000234 (0.065) Log Size 0.001952 0.001990 0.001952 (2.591) (2.629) (2.626) R 2 0.1809 0.1817 0.1809 NOTES: t-ratios are in parenthesis. Sample size is 35. RC Ratio Property, Plan and Equipment (building at cost plus land) divided by Total Assets. RC>50% A dummy variable, indicating that the company s RC Ratio is above the industry median level. RC >75% A dummy variable, indicating that the company s RC Ratio is above 75 percentile of the industry level. Log Size Log of Year-End Equity Market Capitalization. 13
Table A-2. Second Stage Estimation in Printing, Publishing & Allied Industry Panel (a). Without Size Control Variable Model 1 Model 2 Model 3 Intercept 0.000824-0.000306 0.000070 (0.310) (0.247) (0.069) RC Ratio -0.005954 (0.381) RC > 50% 0.000332 (0.193) RC > 75% -0.000742 (0.385) R 2 0.0054 0.0014 0.0055 Panel (b). With Size Control Variable Model 4 Model 5 Model 6 Intercept 0.011900 0.009777 0.011859 (1.545) (1.383) (1.628) RC Ratio -0.008977 (0.584) RC > 50% 0.000223 (0.132) RC > 75% -0.001567 (0.809) Log Size -0.000805-0.000763-0.000879 (1.528) (1.448) (1.633) R 2 0.0873 0.0759 0.0980 NOTES: t-ratios are in parenthesis. Sample size is 29. RC Ratio Property, Plan and Equipment (building at cost plus land) divided by Total Assets. RC>50% A dummy variable, indicating that the company s RC Ratio is above the industry median level. RC >75% A dummy variable, indicating that the company s RC Ratio is above 75 percentile of the industry level. Log Size Log of Year-End Equity Market Capitalization. 14
Table A-3. Second Stage Estimation in Chemicals & Allied Products Industry Panel (a). Without Size Control Variable Model 1 Model 2 Model 3 Intercept 0.009064 0.010301 0.008249 (3.829) (4.547) (4.342) RC Ratio -0.010587 (1.253) RC > 50% -0.006643 (2.102) RC > 75% -0.004791 (1.343) R 2 0.0213 0.0578 0.0245 Panel (b). With Size Control Variable Model 4 Model 5 Model 6 Intercept 0.032433 0.032390 0.031861 (3.162) (3.232) (3.133) RC Ratio -0.010217 (1.246) RC > 50% -0.006167 (2.001) RC > 75% -0.004753 (1.375) Log Size -0.001758-0.001675-0.001772 (2.338) (2.259) (2.361) R 2 0.0913 0.1210 0.0955 NOTES: t-ratios are in parenthesis. Sample size is 74. RC Ratio Property, Plan and Equipment (building at cost plus land) divided by Total Assets. RC>50% A dummy variable, indicating that the company s RC Ratio is above the industry median level. RC >75% A dummy variable, indicating that the company s RC Ratio is above 75 percentile of the industry level. Log Size Log of Year-End Equity Market Capitalization. 15
Table A-4. Second Stage Estimation in Primary Metal Industry Panel (a). Without Size Control Variable Model 1 Model 2 Model 3 Intercept -0.002354-0.004797-0.002167 (0.984) (1.762) (1.086) RC Ratio -0.005086 (0.586) RC > 50% 0.002328 (0.673) RC > 75% -0.003821 (1.067) R 2 0.0126 0.0165 0.0405 Panel (b). With Size Control Variable Model 4 Model 5 Model 6 Intercept -0.012013-0.009040-0.010931 (0.910) (0.692) (0.855) RC Ratio -0.007421 (0.798) RC > 50% 0.001995 (0.546) RC > 75% -0.004240 (1.157) Log Size 0.000822 0.000361 0.000722 (0.744) (0.332) (0.694) R 2 0.0331 0.0207 0.0579 NOTES: t-ratios are in parenthesis. Sample size is 29. RC Ratio Property, Plan and Equipment (building at cost plus land) divided by Total Assets. RC>50% A dummy variable, indicating that the company s RC Ratio is above the industry median level. RC >75% A dummy variable, indicating that the company s RC Ratio is above 75 percentile of the industry level. Log Size Log of Year-End Equity Market Capitalization. 16
Table A-5. Second Stage Estimation in Industrial, Commercial Machinery, Computer Equipment Industry Panel (a). Without Size Control Variable Model 1 Model 2 Model 3 Intercept -0.002846 0.001275 0.000732 (0.613) (0.539) (0.370) RC Ratio 0.016157 (0.604) RC > 50% -0.003161 (0.924) RC > 75% -0.003829 (0.976) R 2 0.0053 0.0122 0.0136 Panel (b). With Size Control Variable Model 4 Model 5 Model 6 Intercept 0.004559 0.007476 0.008279 (0.393) (0.682) (0.752) RC Ratio 0.017123 (0.637) RC > 50% -0.002966 (0.859) RC > 75% -0.003912 (0.993) Log Size -0.000634-0.000528-0.000631 (0.697) (0.579) (0.697) R 2 0.0123 0.0171 0.0206 NOTES: t-ratios are in parenthesis. Sample size is 71. RC Ratio Property, Plan and Equipment (building at cost plus land) divided by Total Assets. RC>50% A dummy variable, indicating that the company s RC Ratio is above the industry median level. RC >75% A dummy variable, indicating that the company s RC Ratio is above 75 percentile of the industry level. Log Size Log of Year-End Equity Market Capitalization. 17
Table A-6. Second Stage Estimation in Electronics, Other Electronics, Excluding Computer Industry Panel (a). Without Size Control Variable Model 1 Model 2 Model 3 Intercept 0.001375-0.002528-0.002819 (0.610) (2.045) (2.981) RC Ratio -0.035635 (2.799) RC > 50% -0.003673 (2.146) RC > 75% -0.006400 (3.407) R 2 0.1006 0.0625 0.1440 Panel (b). With Size Control Variable Model 4 Model 5 Model 6 Intercept -0.004559-0.007420-0.008217 (0.850) (1.406) (1.680) RC Ratio -0.035825 (2.803) RC > 50% -0.003516 (2.043) RC > 75% -0.006347 (3.384) Log Size 0.000492 0.000397 0.000444 (1.218) (0.953) (1.124) R 2 0.1198 0.0749 0.1596 NOTES: t-ratios are in parenthesis. Sample size is 71. RC Ratio Property, Plan and Equipment (building at cost plus land) divided by Total Assets. RC>50% A dummy variable, indicating that the company s RC Ratio is above the industry median level. RC >75% A dummy variable, indicating that the company s RC Ratio is above 75 percentile of the industry level. Log Size Log of Year-End Equity Market Capitalization. 18
Table A-7. Second Stage Estimation in Transportation Equipment Industry Panel (a). Without Size Control Variable Model 1 Model 2 Model 3 Intercept -0.000933-0.000656-0.000397 (0.216) (0.314) (0.241) RC Ratio -0.002846 (0.113) RC > 50% -0.001438 (0.494) RC > 75% -0.004099 (1.230) R 2 0.0004 0.0069 0.0414 Panel (b). With Size Control Variable Model 4 Model 5 Model 6 Intercept -0.011983-0.011317-0.008804 (1.110) (1.156) (0.867) RC Ratio -0.000732 (0.029) RC > 50% -0.001361 (0.469) RC > 75% -0.003316 (0.954) Log Size 0.000813 0.000806 0.000624 (1.117) (1.114) (0.839) R 2 0.0357 0.0419 0.0609 NOTES: t-ratios are in parenthesis. Sample size is 37. RC Ratio Property, Plan and Equipment (building at cost plus land) divided by Total Assets. RC>50% A dummy variable, indicating that the company s RC Ratio is above the industry median level. RC >75% A dummy variable, indicating that the company s RC Ratio is above 75 percentile of the industry level. Log Size Log of Year-End Equity Market Capitalization. 19
Table A-8. Second Stage Estimation in Instrument; Photo Goods, Watches Industry Panel (a). Without Size Control Variable Model 1 Model 2 Model 3 Intercept 0.004110 0.001719 0.000278 (0.889) (0.692) (0.140) RC Ratio -0.028459 (1.130) RC > 50% -0.004369 (1.330) RC > 75% -0.003358 (0.947) R 2 0.0373 0.0509 0.0265 Panel (b). With Size Control Variable Model 4 Model 5 Model 6 Intercept 0.003787 0.002416 0.002213 (0.344) (0.232) (0.206) RC Ratio -0.028477 (1.114) RC > 50% -0.004383 (1.312) RC > 75% -0.003494 (0.951) Log Size 0.000025 0.000054-0.000147 (0.032) (0.069) (0.183) R 2 0.0373 0.0510 0.0275 NOTES: t-ratios are in parenthesis. Sample size is 35. RC Ratio Property, Plan and Equipment (building at cost plus land) divided by Total Assets. RC>50% A dummy variable, indicating that the company s RC Ratio is above the industry median level. RC >75% A dummy variable, indicating that the company s RC Ratio is above 75 percentile of the industry level. Log Size Log of Year-End Equity Market Capitalization. 20