Testing the impact of corporate ownership structure on investment

any dividend requirements. Last, there is an unregulated over-the-counter market, the ‘Hors Cote’. All stock exchange activities are subject to the oversight of the Commission de Operations en Bourse COB, a state-appointed body similar to the US Securities and Exchange Commission. In 1990, 70 of the companies were traded on the main market, representing 93 of the French market capitalization. However, the growth in the second market has brought significant changes in corporate ownership and financing. A large number of the companies traded on this market were originally owned by families, but they have been able to raise more equity than the minimum requirements thanks in part to the interest shown by institutional investors.

4. Testing the impact of corporate ownership structure on investment

4 . 1 . Sample construction This section analyses a sample of 123 non-financial French corporations, all publicly traded between 1983 and 1990. Subsidiaries of French and foreign compa- nies are excluded from this project. The years 1983 – 1986 are used as instruments for the regressions run on the period 1987 – 1990. Accounting and financial informa- tion, as well as details on the ownership structure have been obtained from the reports published by the French company DAFSA. Sources and uses of funds, including investment expenditures, are provided for all companies. DAFSA also publishes a list of all French groups having more then ten subsidiaries. This list is used in this study to separate group firms from independent firms. The sample is split in three categories based on their ownership characteristics. The first sub-sample consists of 37 groups with bank ties. A firm is defined as having bank ties if the percentage of its shares owned by French banks is above 2 on average over the sample years. The second sub-sample includes 64 groups without significant bank links. The third category is limited to firms that do not belong to a group. For such companies, bank ownership is very low. Table 1 describes the ownership characteristics of each category. Groups with bank ties have an average institutional ownership of 18 banks, insurance companies and government agencies combined. The average is only 2 for independent firms. These companies are largely owned by families or large private investors. Their size, measured by the beginning-of-period stock of tangible assets LagK, is on average nine times smaller than that of groups with bank ties. Most industry classifications are represented in each firm subcategory. Table 2 shows that firms with bank owners sustain higher levels of leverage. The average debt-to-assets ratio for groups with bank ties is 18 higher than that of groups without bank ownership, and 50 higher than that of independent firms. Second, financially constrained firms should accumulate liquid assets in order to compensate for their restricted access to external funds. Adverse selection problems that affect lenders in credit markets may also be mitigated by the existence of liquid assets and collaterals. Three measures of corporate liquidity support this hypothesis. Working capital, liquidity and retained earnings are much larger for companies that do not have bank links. t-tests and non-parametric tests between the three sub-samples confirm that these differences are statistically significant. 4 . 2 . A Tobin ’ s Q model with measures of corporate liquidity In this section, a Q model of investment is used, first on the whole sample, then on each category of firms. The reduced form equation used for empirical estimation is: I i,t K i,t − 1 = c + c i + c t + b · Q i,t − 1 + u i,t 1 I i,t , investment at time t for firm I; K i,t − 1 , physical capital stock at the end of period t − 1; c i , firm specific effect; c t , year dummy; Q i,t − 1 market value of equity + book value of debt − book value of inventory divided by book value of tangible assets at the beginning of period t. Cash flow working capital and lagged sales are then added to the model in order to estimate the effect of financing constraints on investment decisions. The resulting empirical specification is: I t K t − 1 = c + c i + c t + b Q i,t − 1 + g S i,t − 1 K i,t − 2 + d CF i,t K i,t − 1 + o WC i,t − 1 K i,t − 2 + u i,t 2 Table 1 Ownership structure by firm categories, 1987–1990 Groups without bank ties, Variable Independent firms, 80 Groups with bank ties, 148 obs. obs. 264 obs. LagK millions FRF 242.2 3851.1 Mean 2161.4 84.7 Median 755.9 398.2 Bank ownership 1 Mean 1 11 7 Median Institutional ownership 6 2 18 Mean Median 13 Nb. Institutional in6estors Mean 2.13 0.67 0.50 Median 2 Ownership by other firms 36 35 33 Mean Median 27 37 25 Family and indi6iduals 5 12 17 Mean Median 5 Table 2 Summary statistics by firm categories, 1987–1990 a With bank All firms Without bank Independent Variable In6estmentLagK 0.37 0.38 0.38 0.38 Mean 0.33 0.33 0.34 0.33 Median Cash flowLagK 0.70 0.69 0.69 Mean 0.70 0.51 0.56 0.58 0.61 Median OutputLagK 10.69 9.44 Mean 9.94 8.86 7.04 8.32 7.23 Median 7.20 Interest co6erage 0.32 0.48 0.29 0.34 Mean 0.18 0.22 Median 0.20 0.24 Payout ratio 0.06 Mean 0.13 0.09 0.10 0.11 0.09 0.10 Median 0.11 DebtLagK 2.10 1.81 1.78 1.38 Mean 1.15 1.23 1.55 Median 1.31 LiquidityLagK 1.73 1.93 1.10 1.58 Mean 1.13 1.57 Median 1.12 0.90 Ret. earningsLagK Mean 0.72 1.24 1.30 1.09 0.93 1.34 0.58 Median 0.85 Working cap.LagK 2.77 2.50 1.71 Mean 2.31 1.73 Median 2.45 1.76 1.55 a Cash flow, operating income+depreciation; payout ratio, dividendoperating income; interest cover- age, interestinterest+cash flow; liquidity, cash+ST securities+accounts receivable−accounts payable; working capital, liquidity+inventory. S t − 1 , lagged sales, CF t , cash flow in period t, WC t − 1 , beginning of period working capital. The two measures of liquidity included in this regression create new problems of estimation because they can also proxy for investment opportunities. Although the Q variable used in this model is supposed to control for profitability, its empirical value may be biased because of stock market imperfections. As a consequence, variables measuring liquidity may be significant for all categories of firms, only because they proxy for future performance. In order to address this problem, this study tries to reduce endogeneity in two ways. First, a variable measuring the sales level at time t − 1 is included in the regression in order to control for future demand for capital goods. Further, instruments based on lagged values of the dependent and independent variables are used to control for future profitability. The equations are then first-differenced in order to remove the firm’s specific effect. Estimation is made with the Generalized Method of Moments, since it allows for heteroskedasticity, autocorrelation, and the use of instrumental variables. 4 . 3 . Empirical results Tables 3 – 6 present detailed regression results for the whole sample and for each category of firms. Estimated coefficients, standard errors, and tests of the overiden- tifying restrictions are given for each model specification. Firm specific effects are removed by first differencing the equation. Year dummies are included as regressors and instruments in all equations. The instruments used in all equations are: Q at t − 1; cash flow ratio at t − 2 and t − 3; investment ratio at time t − 2 and t − 3; retained earnings ratio at time t − 2 and t − 3; liquidity ratio at t − 2 and t − 3 and working capital ratio at time t − 2 and t − 3. Table 3 GMM estimation of the Q model: all firms, 1987–1990 a C All firms Q i,t S i,t−1 K i,t−2 CF i,t K i,t−1 WC i,t−1 K i,t−2 Coefficient − 0.038 0.037 s.e. 0.015 0.025 x 5 2 = 2.361 Coefficient 0.279 0.014 0.005 0.017 0.128 0.029 s.e. x 8 2 = 7.655 − 0.031 0.029 Coefficient 0.037 s.e. 0.028 0.014 0.024 x 8 2 = 8.026 0.009 Coefficient 0.033 0.306 0.007 s.e. 0.026 0.024 0.013 0.116 x 8 2 = 4.929 − 0.008 0.021 Coefficient 0.008 0.242 0.025 0.119 s.e. 0.012 0.005 x 8 2 = 5.228 0.012 Coefficient − 0.028 0.026 0.009 0.022 0.037 0.013 s.e. 0.007 x 8 2 = 7.445 Coefficient 0.018 − 0.001 0.015 0.005 0.268 0.017 0.029 s.e. 0.035 0.131 0.007 x 7 2 = 4.501 a and indicate significance at the 5 and 1 levels. Table 4 GMM estimation of the O model: groups with bank ties, 1987-1990 a C Q i,t S i,t−1 K i,t−2 CF i,t K i,t−1 WC i,t−1 K i,t−2 With banks Coefficient 0.017 0.017 0.017 0.021 s.e. x 5 2 = 2.591 0.104 0.009 Coefficient − 0.003 0.011 0.018 0.104 s.e. x 8 2 = 5.508 Coefficient 0.004 0.008 0.010 0.014 0.017 s.e. 0.020 x 8 2 = 4.482 0.112 0.004 − 0.011 − 0.002 Coefficient 0.012 0.018 0.099 0.016 s.e. x 8 2 = 5.966 0.004 0.109 0.009 − 0.001 Coefficient 0.006 0.103 s.e. 0.018 0.011 x 8 2 = 5.094 0.026 − 0.028 0.015 − 0.036 Coefficient 0.013 0.013 0.037 0.022 s.e. x 8 2 = 3.440 0.019 Coefficient 0.144 − 0.001 − 0.054 0.015 0.008 0.102 0.017 0.024 s.e. 0.029 x 7 2 = 1.925 a and indicate significance at the 5 and 1 levels. As reported in the empirical literature, the Q coefficient is small but significant when estimated without any adjustment for capital markets imperfections. The lack of significance for the categories of groups with bank ties and independent firms is attributed to the small size of these two classes, since pooling observations from any two of these three sub-samples makes the statistical significance of Q reappear. Tables 3 – 6 also present results for the Q model augmented to include measures of liquidity. Under the hypothesis of perfect capital markets, even if the model generates biased results for each category, there should be no difference in the estimated liquidity coefficients of each class as long as the bias is the same. However, the assumption of frictionless markets does not hold for the sample studied here. Groups with bank ties behave differently than firms from the two other sub-samples. Their cash flow and working capital coefficients are small and not statistically significant. In contrast, cash flow matters for groups without bank ownership, and for independent firms. Working capital is significant only for the latter category. These results are consistent with the theoretical literature on agency conflicts and asymmetric information problems. First, the investment decisions of French firms with bank ties are not influenced by the availability of cash flow and working capital. This result could be attributed to the positive impact of bank shareholders who provide monitoring and easy access to debt. However, because French banks and institutional investors are often controlled by the state, the Q model of investment may not be the right structural approach for corporations with strong institutional ties. Such companies may be less responsive to private incentives such as market valuation and internal liquidity. As for the two other firm categories groups without bank ties, and independent companies, two major results appear from the regression analysis. First, the coefficient on the cash-flow variable is higher for large groups when lagged sales and working capital are included in the regression than for independent firms. This result may support the hypothesis that large corporations suffer from significant agency conflicts related to the use of free cash-flow. It is also consistent with the international findings of Kadapakkam et al. 1998 showing that large firms rely more on cash flow than small firms. Second, the working capital variable is only significant for the category of small firms. This is consistent with the hypothesis that independent firms suffer from strong adverse selection problems. They retain higher stocks of liquid assets to finance their capital expenditures. 4 . 4 . Corporate in6estment and public capital markets in France As explained in Section 3.2, France has three categories of stock markets with specific information requirements. Since these markets were started at different Table 5 GMM estimation of the Q model: groups without bank ties, 1987–1990 a c Q i,t WC i,t−1 K i,t−2 CF i,t K i,t−1 Without bank S i,t−1 K i,t−2 0.036 Coefficient − 0.063 0.040 s.e. 0.017 x 5 2 = 2.948 Coefficient 0.267 0.004 0.033 0.101 0.013 0.037 s.e. x 8 2 = 6.480 − 0.019 0.017 Coefficient 0.013 s.e. 0.032 0.014 0.034 x 8 2 = 11.637 Coefficient 0.071 0.425 0.004 0.042 0.136 0.040 0.015 0.039 s.e. x 8 2 = 2.463 0.021 0.013 Coefficient 0.009 0.296 0.008 0.100 0.011 0.034 s.e. x 8 2 = 5.342 Coefficient − 0.010 0.008 − 0.018 0.018 0.033 s.e. 0.048 0.015 0.010 x 8 2 = 11.830 0.001 Coefficient 0.001 0.046 0.424 0.067 0.045 0.019 0.013 0.149 0.057 s.e. x 7 2 = 2.172 a and indicate significance at the 5 and 1 levels. Table 6 GMM estimation of the Q model: independent firms, 1987–1990 a c Q i,t S i,t−1 K i,t−2 CF i,t K i,t−1 WC i,t−1 K i,t−2 Independent Coefficient − 0.047 0.065 0.051 0.050 s.e. x 5 2 = 6.835 0.307 − 0.010 Coefficient − 0.015 0.030 0.045 0.146 s.e. x 8 2 = 7.957 Coefficient 0.123 − 0.028 0.021 0.035 0.048 s.e. 0.046 x 8 2 = 6.025 0.347 − 0.046 0.073 0.008 Coefficient 0.035 0.044 0.144 0.035 s.e. x 8 2 = 5.168 0.007 0.294 − 0.006 − 0.014 Coefficient 0.008 0.143 s.e. 0.044 0.025 x 8 2 = 7.407 − 0.009 − 0.023 − 0.007 0.077 Coefficient 0.010 0.029 0.028 0.044 s.e. x 8 2 = 9.636 − 0.002 Coefficient 0.338 0.006 0.076 − 0.044 0.008 0.147 0.035 0.036 s.e. 0.045 x 7 2 = 5.090 a and indicate significance at the 5 and 1 levels. points in time, they also provide useful information on the age and size of their listed companies. In this section, I will use this criterion to proxy for the severity of information asymmetries between firms and external investors. The objective is to show that constrained firms accumulate liquid assets when they have poor access to public markets. In order to test this hypothesis, groups with bank ownership are removed from the sample, since their investment decisions are not sensitive to liquidity in this model. Further, they are often government controlled firms for which public trading has not been consistent. The remaining firms, groups without bank ties and independent firms, are split into two categories based on their stock exchange listing. The first sub-sample is restricted to companies traded on the monthly settlement market. As described in Section 3, they are all large and mature firms. The other sub-sample includes all other corporations traded on the cash, second market or over-the-counter markets. The monthly settlement market should allow companies to raise external funds more easily, since it deals only with well-estab- lished corporations. In contrast, smaller and less mature firms traded on the secondary markets should have limited access to external finance. Table 7 compares the characteristics of companies traded on the monthly settlement market to those of firms listed on smaller markets. A non-parametric Wilcoxon test shows that, although debt ratios for the two subsamples are not statistically different from each other, firms traded on the monthly settlement market have a lower interest coverage ratio. Moreover, these companies also retain lower stocks of liquidity and working capital. These results validate the hypothesis that information asymmetries between external providers of funds and a firm are mitigated when the company is traded on a mature and active stock market. Companies that do not benefit from a privileged position on capital markets face a higher cost of external finance and choose to rely on internal funds. Table 7 Summary statistics by categories of stock markets, 1987–1990 a Without bank ties other trading Without bank ties monthly Variable settlement, 192 obs. markets, 152 obs. 1 LagK 359.9 5112.2 Mean Median 74.4 955.0 In6estmentLagK 1 0.35 0.40 Mean 0.29 Median 0.35 1 Cash FlowLagK 0.76 Mean 0.62 0.61 0.56 Median 1 OutputLagK Mean 9.61 11.39 Median 7.00 7.55 Interest co6erage 1 0.60 0.17 Mean Median 0 16 0.22 1 Payout ratio Mean 0.04 0.11 0.11 Median 0.10 DebtLagK 1 Mean 1.64 1.74 Median 1.02 1.29 1 LiquidityLagK Mean 1.84 1.73 1.11 Median 1.47 1 Ret. earningsLagK Mean 1.22 1.27 0.88 Median 1.12 Working cap.LagK 2.48 2.63 Mean Median 1.92 1.77 a Cash flow, operating income+depreciation; payout ratio, dividendoperating income; interest cover- age, interestinterest+cash flow; liquidity, cash+ST securities+accounts receivable−accounts payable; working capital, liquidity+inventory. Table 8 GMM estimation of the O model by stock market categories, 1987–1990 a C Q i,t S i,t−1 K i,t−2 CF i,t K i,t−1 WC i,t−1 K i,t−2 Firm category Monthly settlement 0.002 0.236 0.013 0.025 Coefficient 0.013 0.034 0.007 0.152 s.e. x 8 2 = 13461 − 0.005 0.018 0.010 Coefficient − 0.009 0.010 s.e. 0.041 0.029 0.017 x 8 2 = 13172 0.007 0.032 0.000 0.249 0.018 Coefficient 0.009 0.147 0.015 0.035 0.033 s.e. x 7 2 = 10.634 Other markets 0.020 0.054 0.007 − 0.016 Coefficient 0.071 0.071 s.e. 0.040 0.015 x 8 2 = 5.090 − 0.008 0.022 − 0.003 0.188 Coefficient 0.009 0.018 0.074 0.042 s.e. x 8 2 = 4.697 − 0.004 Coefficient 0.122 0.019 0.202 − 0.007 0.011 0.148 0.035 0.083 s.e. 0.050 x 7 2 = 4.035 a and indicate significance at the 5 and 1 levels. Empirical analysis is then conducted with the reduced form investment equations used in the previous section. Table 8 shows that the investment decisions of firms traded on the monthly settlement market are not based on their level of working capital or cash flow. In contrast, companies traded on less-established markets make their amount of investment dependent on their accumulation of working capital. Importantly, when the sample is split based on trading patterns, the statistical significance of the cash flow variable disappears in the Q model. 4 . 5 . E6olution of corporate financing between 1983 and 1990 Economic changes and more active stock markets influenced the investment behavior of French companies between 1983 and 1990. As described in Section 3 of this paper, the ownership structure of French firms was altered by the privatization of banks and the growth of public shareholding. This section investigates whether these changes benefitted non-financial companies by decreasing their reliance on internal funds between 1983 and 1990. The sample used for that purpose consists of 109 French firms traded from 1981 to 1990. It excludes younger companies, especially the ones listed on the second market which opened in 1983. The data is then split into two categories. The first one consists of 49 firms which all had bank ties before 1986, when institutional ownership was strongly linked to government control. The other one consists of all companies without any bank ownership before 1986. The objective is to compare the period 1983 – 1985 to the period 1988 – 1990 using 81 – 82 and 86 – 87 as instruments. Table 9 describes in more details the ownership characteristics of the two categories, as well as their evolution between 1983 and 1990. Institutional ownership declined for firms with bank ties, and increased for firms without bank ties. This reflects the diversification strategy implemented by recently privatized institutional investors. Table 10 provides summary statistics for the two subsamples. All measures of performance and liquidity improved between 1983 and 1990, regardless of the firm category. However, non-parametric tests show that some of these ratios are statistically different between the two types of companies. First, measures of liquidity, retained earnings, and working capital are systematically higher for firms without bank ties. However, debt levels are equivalent across the two sections, with a larger interest coverage for the latter category. Second, although companies without bank links improved their investment, cash flow and output ratios at the same rate as firms with bank shareholders, their levels of liquid assets increased at a slower pace during the period 1988 – 1990. These statistics support the hypothesis that companies without bank ties gained better access to capital markets during the late 1980s, thereby decreasing their need to accumulate internal funds. A Q model of investment is then augmented with sales, cash flow and working capital variables. Table 11 provides separate regression results for the two firm categories and the two time periods. The instruments used for this model are: Q at Table 9 Comparison of ownership structures between 1983–85 and 1988–90 Firms without bank ties 60 firms Variable Firms with bank ties 49 firms LagK millions FRF 509.1 Median 1988–1990 236.1 Bank ownership Mean: 1983–1985 8 7 1 Mean: 1988–1990 Institutional ownership 14 Mean: 1983–1985 2 Mean: 1988–1990 12 4 Nb. Institutional in6estors Mean: 1983–1985 1.93 0.49 Mean: 1988–1990 1.44 0.70 Ownership by other firms 37 Mean: 1983–1985 38 36 44 Mean: 1988–1990 Family and indi6iduals Mean: 1983–1985 6 10 Mean: 1988–1990 12 6 Table 10 Summary statistics by firm categories, 1983–85 versus 1988–90 a Firms with bank ties Firms without bank ties Variable In6estmentLagK 0.27 0.27 1983–1985 0.34 0.33 1988–1990 Cash flowLagK 0.48 0.45 1983–1985 0.54 0.53 1988–1990 OutputLagK 1983–1985 7.78 7.59 7.14 6.88 1988–1990 Interest co6erage 1983–1985 0.26 0.22 0.21 0.19 1988–1990 Payout ratio 0.09 0.07 1983–1985 0.11 0.11 1988–1990 DebtLagK 1983–1985 1.26 1.21 1.39 1.29 1988–1990 LiquidityLagK 1983–1985 0.31 0.25 1.26 1.04 1988–1990 Ret. earningsLagK 0.91 0.31 1983–1985 1.77 0.68 1988–1990 Working cap.LagK 2.48 1983–1985 0.63 1988–1990 1.92 1.54 a Cash flow, operating income+depreciation; payout ratio, dividendoperating income; interest cover- age, interestinterest+cash flow; liquidity, cash+ST securities+accounts receivable−accounts payable; working capital, liquidity+inventory; all ratios reported in this table are median values. time t − 2; cash flow ratio at time t − 1 and t − 2; investment ratio at time t − 1 and t − 2; retained earnings ratio at time t − 1 and t − 2; liquidity ratio at time t − 2; working capital ratio at time t − 2; and payout ratio at time t − 1 and t − 2. GMM estimation of the coefficients leads to the rejection of the structural model for companies with bank links. This result validates the hypothesis that firms with bank ownership do not follow a neoclassical model of investment including stock market valuation. As mentioned before, this effect may be linked to the intervention of the French government in firms having significant links with banks and other institu- tional investors. In contrast, the Q model including liquidity variables is accepted for companies that did not have any bank ties before 1986. However, the impact of working capital on their investment decisions is only significant for the period 1983 – 1985, before the liberalization of the French economy. This result is consis- tent with the hypothesis that the same firms may have faced stronger problems of information asymmetries in the years preceding the expansion of the Paris Bourse.

5. Euler equation approach