Cross-border mergers and acquisitions: the undervaluation hypothesis.
The growing web of interdependencies in the global economy has developed new relationships between economic agents of different countries. In the last decade, an interesting phenomenon surfaced in the international market for corporate control. The number of foreign firms acquiring U.S. firms, in aggregate terms, has been larger than the number of U.S. firms taking over foreign companies. For instance, during the 1981-1990 period the average number of transactions per year involving a foreign bidder for a U.S. company was 218 and the yearly average dollar amount for the same period was $23.4 billion. We can contrast with this the average number of transactions and dollar amounts involving U.S. bidders for a foreign company which were 147 and $8.5 billion respectively. Thus, as Table 1 shows, U.S. companies have played mainly a target role in the cross-border market for corporate control. The exact motivations for observing U.S. firms as targets outnumbering bidders are many (e.g., macroeconomic factors, firm-specific financial characteristics, corporate strategic moves, political motives, and/or the possibility of a "good buy"). The focus of our study is on this final factor, management's quest for undervalued assets.
Table 1. Cross-border Merger And Acquisition Activity Involving U.S. Companies Year U.S. Target U.S. Bidder Transactions Billions ($) Transactions Billions ($) 1981 243 18.1 10 11.1 1982 153 5.1 121 0.8 1983 125 5.9 146 2.5 1984 151 15.5 147 2.6 1985 197 10.9 175 1.4 1986 264 24.5 180 5.2 1987 220 40.4 142 11.0 1988 307 55.5 151 14.5 1989 285 40.4 220 22.2 1990 266 33.0 266 18.0 Average 218 23.4 147 8.5 Source: Mergerstat Review
International mergers and acquisitions research focuses primarily on wealth transfers. For instance, Doukas and Travlos (1988), besides offering an excellent review of this literature, contrasts the returns to shareholders from U.S. and non-U.S. based firms expanding into foreign markets. Conn and Connell (1990) also include an extensive literature review of merger and acquisitions within their empirical study of wealth transfers between U.S. and British firms expansion into each other's markets. Outside of the wealth transfer research, empirical international merger and acquisition research is lagging behind its domestic (e.g., U.S.) counterpart which is rich in studies from the perspective of both sides of the negotiation table. In this tradition, Harris and Ravenscraft's (1991) linkage of the undervaluation, management inefficiency, and market imperfections hypotheses provides the theoretical foundation for our empirical testing. Thus, our contribution to the merger literature is the empirical validation of undervaluation as one of the key financial motivations underlying acquisitions in the international arena.
Under our hypotheses, we postulate that the existence of product and service market imperfections that cause frictions in the global market (such as transaction costs and costs associated with barriers to entry) contributes to favor the acquisition of a company already operating. This is because the amount paid for an existing company, as compared to the replacement cost of its assets, more than compensates for the costs that could have been incurred had the foreign firm started with brand new facilities. Thus, in order to minimize the acquisition costs, foreign firms should follow the same pattern of analysis as their domestic counterparts and search for undervalued and/or mismanaged companies as targets for their acquisitions.(1) The results of our undervaluation hypotheses testing, within the Tobin's q framework utilized by Servaes (1991) for the study of domestic mergers, support this viewpoint.(2) To our knowledge, there are no other studies on cross-border merger and acquisitions that validate the theoretical undervaluation hypothesis within an international setting.
Other domestic M&A studies, such as Palepu (1986) and Dietrich and Sorensen (1984), provide the foundation for our use of the logit methodology for predicting acquisition targets. Palepu (1986) also stresses the need to take into account the fact that the targets and bidders are oversampled and therefore the Maximum Likelihood estimators might be biased. We attempt to compensate for this problem by using a choice-based sample based on a Weighted Maximum Likelihood Estimator (hereafter WMLE) as explained in Appendix II and outlined by Manski and McFadden (1981).
Consistent with previous studies applied to the domestic market for corporate control (see for example Chappel and Cheng, 1984), we hypothesize that undervalued U.S. companies are more likely to be targets of foreign companies. Thus, our first hypothesis, Undervaluation-Target Hypothesis, is described as:
H1: The likelihood of a U.S. firm becoming a target increases when the firm is perceived as being undervalued.
Within the empirical analysis, we proxy this undervaluation with Tobin's q (i.e., the ratio of market value to replacement cost of assets of the U.S. firm), which is a continuation of the approach pioneered by Tobin (1969). Since then many other researchers have used Tobin's q as both a theoretical and an empirical tool to establish a relationship between the product or service markets and the capital markets. For instance, Chirinko (1987) concludes that the theoretical usefulness of Tobin's q stems from the fact that it incorporates forward-looking behavior, reflects optimal choices, and contains estimated coefficients that are readily identified.
Under this hypothesized relationship, investment (i.e. the addition to the stock of capital) is determined by the marginal "q", defined as the ratio of the discounted future revenues from an additional unit of capital to its net-of-tax purchase price. However, due to difficulties of empirically valuing a marginal "q", our study relies on an average "q." This proxy is supported by Tobin and Brainard (1977), who emphasized that the forces of continuity in the economy are strong and that we can expect that the same factors which raise or lower "q" on the margin will likewise raise or lower "q" on the average.
Assuming that the takeover decision is motivated by the same stimuli that encourage firms to grow internally, Chappell and Cheng (1984) were among the first to study the "q" ratio as a predictor of takeover targets. They found that the high abnormal returns experienced by acquirers before the merger are consistent with a high "q" ratio, signaling to the companies that it is time to expand. Nevertheless, they concluded that the effect of the "q" ratio is not always significant and that these effects vary. Holly and Longbottom (1988), using the same framework followed by Chappell and Cheng (1984), analyzed U.K. firms and found that if the average "q" ratio is more than one, the takeover (i.e., investment) is desirable. If it is less than one, it is not.
Lang, Stulz and Walking (1989) studied tender offers and their relationship to Tobin's q. Under the assumption that the financial market rewards well-managed firms, Lang et. el. interpret a "q" greater than one as a measure of good management. Conversely, a ratio less than one is viewed as evidence of poor management. They conclude that, to the extent that Tobin's q measures managerial performance, the results of their research may be interpreted as follows: (a) Well-managed bidders benefit substantially from tender offers, but more so when they take over poorly managed targets; (b) Well-managed targets benefit less from tender offers than poorly managed targets; (c) The total takeover gain is highest for tender offers by well managed bidders which acquire poorly managed targets.(3) This is consistent with Jensen's (1988) contention that highly valued firms are likely to be bidders in the market for corporate control. Therefore, we test the following hypothesis, which we called the Undervaluation-Bidder Hypothesis:
H2: The likelihood of a foreign firm bidding for a U.S. company increases when the firm is perceived as being overvalued.
We proxy this strength component by using the ratio of market value to replacement cost of the overseas company, i.e., we imply that the ratio would be greater than one. Since this hypothesis expands the research into the international marketplace, the role of the exchange rate must be accounted for. Several alternative (and not mutually exclusive) explanations for the importance attributed to the exchange rate have been offered in previous studies. For example, Vasconcellos, Madura, and Kish (1990) and Vasconcellos and Kish (1993), examining the difference between the number of U.S. acquisitions of foreign companies and the number of foreign acquisitions of U.S. firms, found that the exchange rate could affect the timing of the acquisition more than the acquisition decision itself. In order to control for exchange rate variations, we follow the approach of Harris and Ravenscraft (1991) and proxy exchange rate effects with the quotient of the exchange rate differences (current exchange rate less the three-year moving average) and the three-year moving average exchange rate.
Prior studies have also documented the gains made from taking over an undervalued company.(4) This undervaluation could be observed in a company whose stock price does not reflect the replacement cost of the company's assets or, in a related fashion, it could be due to inefficient management not operating the company to its true potential. Therefore as a complement to our set of undervaluation hypotheses, we also tested the Management Inefficiency Hypothesis. Lang et. al. (1989) have argued that undervalued companies (those with Tobin's q [less than] 1) are an indication of management inefficiency. This interpretation is based on the premise that management fails to use the resources of the company up to their full potential. Thus, our Management Inefficiency Hypothesis can be stated as:
H3: The more inefficient a firm's management, the greater the probability of the firm becoming a target.
Examples of variables that have been used before (in addition to the Tobin's q) to gauge management efficiency are the return on equity and sales growth. If management is inefficient, then we might expect both variables to be negatively related with the probability of an acquisition.
II. THE DATA
The compilation of the data started by identifying U.S. firms that have been acquired by foreign companies during the 1981-1990 period. This information was gathered from two sources: Mergers and Acquisitions and Mergerstat Review. At the same time, the name and country of origin of the foreign bidder was also gathered. Relevant information for each target was extracted from COMPUSTAT (Research and Industrial Files). Information regarding the foreign bidders was obtained from COMPUSTAT GLOBAL VANTAGE, an international data base compiled by Standard and Poor's.
Moody's Industrial Manuals were used to obtain the information on yields (bonds and preferred stock) in the United States. Information on macroeconomic aggregates was gathered from the Survey of Current Business. The sources for the relevant data on yields and macroeconomic aggregates of foreign countries included the following publications: the Monthly Bulletin of Statistics published by the United Nations, International Financial Statistics prepared by the International Monetary Fund, and Financial Statistics issued by the Statistical Office of the European Communities.
In order to include a firm in our sample, the following criteria had to be met: the merger or acquisition took place in the 1981-1990 period; the merger or acquisition transaction was reported by Mergers and Acquisitions; and the necessary company data for testing the hypotheses of interest was available from COMPUSTAT. A total of 242 U.S. companies that were acquired by foreign companies met the necessary requirements to be included in the sample. A total of 216 U.S. bidders meeting the same set of requirements were also included in the sample. In the case of foreign bidders for U.S. firms, a total of 76 companies were sampled.
In order to test the Undervaluation-Target and Management Inefficiency Hypotheses pertaining to U.S. companies, a control sample of 2000 companies was used. The targets and bidders were classified by their SIC industrial code. Within each SIC code, companies in the same sub-sector were ordered alphabetically. Each company was then matched with a randomly selected firm from within the control group, using SIC codes up to a 4 digit level. A total of 700 foreign companies formed the control sample for testing of the Undervaluation-Bidder Hypothesis. Matching within this sector was also controlled by the country of origin of the bidder.(5)
The incidence of U.S. companies as targets was observed throughout the period of analysis. The U.S. industrial sector most actively involved in the international cross-border merger and acquisition activity was Retail, with 9% of the targets. Other significant sectors included Conglomerates (Mergerstat defines a conglomerate as a company diversified in 3 or more areas) accounting for 6% of the U.S. companies acquired by foreign companies, Wholesale/Distribution and Oil/Gas both with 5.6%, and Drugs/Medical Supplies with 5%.
In the case of U.S. companies bidding for foreign companies, 6.5% were from Conglomerates and Instruments/Photographic Equipment. The other leading sectors were Aerospace/Aircraft, Industrial/Farm Equipment, Insurance, and Office Equipment/Computer Hardware (5.5% each). From the 76 foreign bidders sampled, 32% were Conglomerates, 12% were classified as Oil/Gas, and 8% were from both the Banking/Finance and Retail sectors.Although both the U.S. target firms and the U.S. bidders range across a number of similar industries, we do not observe the same proportions. The foreign bidders cover a more limited range of industries than their American counterparts. Table 2 shows a summary of successful foreign bidders, desegregated by country of origin. Companies from Great Britain were the most active bidders for U.S. companies over the period 1982 through 1990 (53%). The second most active country was Japan with 17% of the companies involved as buyers of American enterprises.
In order to test the hypotheses using the ratio of market value to replacement cost of assets (Tobin's q ratio), we employed the proxy used by Lindenberg and Ross (1981) and by Smirlock, Gilligan and Marshall (1984). We define the market value of a company as the sum of the market value of common stock, the [TABULAR DATA FOR TABLE 2 OMITTED] market value of the preferred stock, and the market value of the debt. The market value of the common stock is the product of the year-end price times the number of outstanding shares at the end of the year prior to the acquisition.(6) The preferred stock was assumed to be a perpetuity valued at the average yield reported by Moody's for the year preceding the acquisition.(7)
Due to the difficulty of readily observing the market value of the components of debt, we follow the proxy methodology employed by Smirlock et al. (1984) and Fabozzi (1990). Thus, the market value of the debt was computed by discounting the balances of the debt maturing in two, three, four, and five years, assuming that the debt was originally issued with a maturity of 20 years, using the Moody's Composite Average yield on industrial bonds for the year of the issue.(8) The use of the debt maturing in these periods is ad-hoc but the lack of a better approximation to the actual distribution of long-term debt and the availability of the data (i.e., COMPUSTAT reports these items) make this proxy a reasonable one. For the case of foreign countries, we could not obtain the distribution of the maturities of the long-term debt. Thus, we employed the book value of the long term debt. The probable impact of using the book value is to underestimate the market value of debt, making it easier to disprove the Undervaluation-Bidder Hypothesis.(9) Unlike the case of long-term debt, we assume that the information given in the financial statements regarding short-term debt reflects the market value of this debt. Deferred Taxes were subtracted from the value of the debt under the assumption that equity investors never expect to pay these non-interest bearing amounts.(10)
Given the significance and valuation problems for Property, Plant and Equipment, and Inventory, special computations were necessary for these items in order to calculate their replacement cost. In other words, in order to compute the replacement cost of assets we had to use a proxy for Property, Plant and Equipment and Inventory, as shown in Appendix I. All the other assets were assumed to have a market value similar to their book value.
In order to test our hypotheses, we use a logit model whose parameters were estimated using the Weighted Maximum Likelihood Method (WMLM) under choice-based sampling. The logit model is defined as:(11)
p(i,t) = 1/[1 + [e.sup.-[Beta]x(i,t)]] (1)
where, p(i,t) is the probability that firm i will be acquired in period t, x(i,t) is the vector of measured attributes, and [Beta] is the vector of unknown parameters to be estimated.
Thus, the bias caused by the characteristics of the sampling procedure can be eliminated by modifying the simple Maximum Likelihood Estimator (MLE). In our analysis we use WMLE as computed by the econometrics package LIMDEP. This estimation procedure was complemented by performing paired t-tests (see Appendix II) on the different variables for the targets and non-targets and for the bidders and non-bidders.
IV. RESULTS AND ANALYSIS
Our hypothesis testing is partitioned into three subsets: the Undervaluation-Target Hypothesis, the Undervaluation-Bidder Hypothesis, and Management Inefficiency Hypothesis.(12)
A. The Undervaluation-Target Hypothesis
The Undervaluation-Target Hypothesis states that the likelihood of a firm being a target increases when the firm is perceived as being undervalued, which we proxy by the ratio of the market value to replacement cost of its assets being less than one. Thus, the Undervaluation-Target Hypothesis can be expressed as:
Table 3. Undervaluation-Target Hypothesis H(0): The ratio of market value to replacement cost of a U.S. firm's assets has no effect on the likelihood of a firm becoming a target of foreign firms. Y = a + b Q where: Y = 1 (acquisition) = 0 (no acquisition) Q = Tobin's q Variable Coefficient t-ratio Constant -1.51 -7.812(a) Q -1.63 -8.434(a) Likelihood ratio index: 0.72(b) Chi-squared: 75.249(c) Significance level: 0.0000032 Notes: a. Significant at a 0.05 level two-tailed test - supports the Undervaluation Target Hypothesis. b. The log likelihood ratio index is defined as 1 - (log likelihood at convergence/log likelihood at zero). It plays the same role as the [R.sup.2] in regression analysis, providing an indication of the model's explanatory power. c. Tests the hypothesis that all the parameters in the model are simultaneously equal to zero.
H4: The ratio of market value to replacement cost (Q) of a U.S. firm's assets has no affect on the likelihood of the firm becoming a target of foreign firms.
H5: The likelihood of a U.S. firm becoming a target increases when the ratio of market value to replacement cost of its assets is less than one.
The Undervaluation-Target Hypothesis implies that there is an inverse relationship between the probability of a U.S. company being acquired and the Tobin's q. Table 3 shows the results of the Logit model that has as the dependent variable acquisition (Y = 1) or no acquisition (Y = 0) and as the independent variable Tobin's q (Q). The coefficient for q is statistically significant and supports the inverse relationship between Tobin's q and the probability of a U.S. company being acquired by a foreign firm. Thus, at least for the decade under analysis, the data provide support for the Undervaluation-Target Hypothesis.(13)
B. The Undervaluation-Bidder Hypothesis
The Undervaluation-Bidder Hypothesis states that the likelihood of a foreign firm bidding for a U.S. company increases when the firm is perceived as being overvalued. Therefore, the relationship between the ratio of market value to replacement cost of assets of foreign firms (QF) to the likelihood of these companies acquiring U.S. companies is analyzed. Within the context of crossborder merger and acquisition activity, our Undervaluation-Bidder Hypothesis can be stated as:
H6: The ratio of market value to replacement cost of a foreign firm's assets (QF) does not affect the probability of this firm becoming a bidder for a U.S. firm.
H7: There is a positive relationship between the likelihood of a foreign firm bidding for a U.S. company and the ratio of market value to replacement cost of the foreign firm.
The independent variables in this case include Tobin's q of foreign bidders (QF) and a control variable for the exchange rate (EXRA). In this case, Y = 1 if the foreign firm bid for a U.S. company and Y = 0 if the overseas company did not. In order to construct the exchange rate variable based on the home country of each buyer, we took the currency's three-year moving average exchange rate for the sample period 1981-1990 and subtracted the currency's exchange rate for the year of the takeover. The exchange rate variable (EXRA) was defined as this difference divided by the three-year moving average exchange rate. As a result, positive (negative) values will indicate the currency is strong (weak) relative to the U.S. dollar.(14) For example, if the $/[pounds] exchange rate was on average 2.0 for the sample period and the 1988 $/[pounds] exchange rate was 1.5, then the exchange rate figure is -0.25 (i.e., (1.5 - 2)/2).
Table 4. Undervaluation-Bidder Hypothesis H(0): The ratio of market value to replacement cost of a foreign firm's assets has no effect on the probability of this firm becoming a bidder of a U.& firm. Y = a + [b.sub.1] QF + [b.sub.2] EXRA where: Y = 1 (acquisition) = 0 (no acquisition) QF = Tobin's q for foreign firm EXRA = Exchange rate variable Variable Coefficient t-ratio Constant -1.632 -3.52(a) Q -1.932 -5.55(a) EXRA +0.380 +0.54 LIKELIHOOD ratio index: 0.38 Chi-squared: 38.03 Significance level: 0.000001387 Notes: a. Significant at a 0.05 level two-tailed test - supports the Undervaluation Bidder Hypothesis.
The results show that there was a direct relation between the possibility of a foreign firm bidding for a U.S. firm and the Tobin's q of the overseas firm. The coefficient of our "Tobin's q" (QF) variable is 1.93 and it is significant at a 5% level, using a two-tailed test. On the other hand, the exchange rate variable, EXRA, although with the expected sign (+), is not significantly different from zero. The model appears to explain well the relationship between the independent variables and the dependent variable, as shown by the Chi-square test of the overall model.(15)
C. The Management Inefficiency Hypothesis
As a complement to the Undervaluation Hypotheses, we test the Management Inefficiency Hypothesis. Previous studies (see for example Lang et. al. 1989) argue that undervalued companies (those with Tobin's q [less than] 1) are an indication of management inefficiency. This interpretation is based on the premise that management fails to use the resources of the company up to their full potential. Examples of variables that have been used before (in addition to the Tobin's q) to gauge management efficiency are the return on equity (ROE) and sales growth (GROWTH). If management is inefficient, then we might expect both variables to be negatively related with the probability of an acquisition. Thus, our Management Inefficiency Hypothesis, which tests whether more inefficient management increases the probability of that firm becoming a target, is stated for testing purposes as:
Table 5. Management Inefficiency Hypothesis H(0): Return on Equity and Growth of a U.S. firm has no impact on the probability of this firm becoming a target of a foreign acquirer. Y = a + [b.sub.1] ROE +[b.sub.2] GROWTH where: Y = 1 (acquisition) = 0 (no acquisition) ROE = Return on equity GROWTH = Sales growth Variable Coefficient t-ratio Constant +0.66 +2.28(a) ROE -3.07 -5.93(a) GROWTH -2.66 -5.33(a) Likelihood ratio index: 0.50 Chi-squared: 134.00 Significance level: 0.00000031 Notes: a. Significant at a 0.05 level two-tailed test - supports the Management Inefficiency Hypothesis.
H8: The ratio Return on Equity (ROE) and the Growth (GROWTH) of a U.S. firm has no impact on the probability of this firm becoming a target of a foreign acquirer.
H9: There exists a inverse relationship between the ratio Return on Equity and Growth and the probability of the U.S. company becoming a target of foreign firm.
Therefore, we hypothesize that low ROE and/or GROWTH are manifestations of low quality management. Table 5 presents the results of a Logit model where both variables (ROE and GROWTH) are employed as predictors of the probability of U.S. companies becoming targets of foreign firms.
Both the ROE and GROWTH coefficients are negative and significant at a 5% level, implying that the probability that a U.S. company will be taken over by a foreign firm is higher the more inefficient the management of the domestic company. Paired t-tests, not shown, are also significant at a 5% level, offering additional support. These results are consistent with the management inefficiency interpretation of the Tobin's q.
The increasing importance of foreign firms in the U.S. market for corporate control motivated us to take a closer look at the financial characteristics of the companies involved in international mergers and acquisitions. The financial characteristics of a total of 533 foreign and U.S. firms involved in cross-border mergers and acquisitions are analyzed.
This study empirically validates the Undervaluation Hypothesis within the international setting using a Logit analysis. The results support the existence of an inverse relationship between the probability of a U.S. firm becoming a target of a foreign company and the Tobin's q ratio. In other words, undervalued U.S. companies are more likely to be targets of foreign companies. This result is consistent with previous studies applied to the domestic market for corporate control (see for example Chappel and Cheng, 1984), but never tested empirically within the international merger and acquisition market.
It is also argued that undervalued companies are the result of a lack of managerial capabilities. Using a Tobin's q [less than] 1 as a proxy for management inefficiency, we test this interpretation for the case of U.S. targets. Our findings show that firms with low return on equity and low growth are more likely to be acquired by foreign companies.
Using the financial characteristics of 76 foreign bidders from 7 industrialized nations, we find that the exchange rate does not have a strong impact on the probability of acquisition of a U.S. company when measured with the valuation of the firm. We observed a very strong dollar during the first half of the 1980's and a weaker dollar in the second half; however, the number of U.S. companies acquired is on average the same. We also find that the foreign firms have a relatively high return on equity (ROE) when compared to the industry average. Since we employ ROE as a proxy for management efficiency, we may conclude that foreign companies with above average efficiency in their countries have a higher likelihood of acquiring U.S. firms.
If we relate these findings to Lang's et al. (1989) conclusions from the domestic marketplace, then one should observe positive abnormal returns for foreign companies upon the announcement of the foreign firms taking over poorly managed U.S. firms. A firm's overvaluation is proxied by a Tobin's q [greater than] 1. Recall that Lang et al. (1989) found positive abnormal returns when a firm with a Tobin's q [greater than] 1 (well-managed firm) acquired an undervalued company. If this is the case then this study extends Lang's et al. (1989) conclusions to the takeover market across countries. We also found that the foreign acquirers and U.S. targets belong to the same industrial sectors. This can be interpreted as foreign companies reducing acquisition costs by acquiring undervalued firms or, as previously said, as foreign firms trying to use their business know-how to enhance the efficiency of the U.S. targets.
As pointed out within the introduction, there are many factors that influence a management's decision under both international and domestic mergers. Therefore, future research endeavors should not only isolate the impact of the various merger factors (e.g., industry characteristics, competitiveness, timing of capitalization, etc.), but also try to capture within a single model or a multi-staged model the total merger decision process. Of course, as with all research endeavors, the researcher has to overcome the data gathering problem, which in the international setting usually is more formidable.
VALUING NET PROPERTY PLANT & EQUIPMENT AND INVENTORY
The replacement cost of Net Property, Plant and Equipment can vary over time due to price level changes, technological change, real economic depreciation, and investment in new plant. Following Lindenberg and Ross (1981) we compute the Replacement cost for Net Property, Plant and Equipment at time t([RNP.sub.t]) as:
[Mathematical Expression Omitted] (A1)
where [RNP.sub.t-1] for the reporting period t = 0 is the same as the book value of the net plant in year 0. Year 0 is defined as the first year of a 10 year period before the acquisition. It represents the investment in new plant equipment, [[Phi].sub.t] is the rate of growth of capital goods prices, [[Delta].sub.t] is the rate of (real) depreciation, and [[Theta].sub.t] is the rate of cost-reducing technical progress.
In order to obtain an estimate of the rate of growth of capital goods prices ([[Phi].sub.t]), we also follow the methodology outlined by Lindenberg and Ross (1981) by using the GNP deflator for nonresidential fixed investment. The real depreciation rate ([[Delta].sub.t]) was computed using equation (A2)
[[Delta].sub.t] = [DEP.sub.t]/[HNP.sub.t-1] (A2)
where [DEP.sub.t] is book depreciation and [HNP.sub.t] is the historical value of net property, plant, and equipment. The rate of cost-reducing technical progress ([[Theta].sub.t]) was computed by the ratio of the firm's annual growth of Research and Development Costs to Total Assets.
In the case of Inventory, its book value was adjusted according to the valuation method employed by the firm. Under LIFO (Last in-First out), ending inventory is valued at old prices. Thus, in inflationary times the book value of inventory will be underestimated. This adjustment gives more weight and accounts for large undervaluations of old inventory (first term) and smaller weight to recent increments (second term). It was assumed that this iterative computation begins at year 0 with the Replacement Value of Inventory ([RINV.sub.t]) equal to the Historical (Book) Value of Inventory ([HINV.sub.t])
[RINV.sub.t] = [RINV.sub.t-1] [multiplied by] [P.sub.t]/[P.sub.t-1] + ([HINV.sub.t] - [HINV.sub.t-1]) [multiplied by] 0.5 [multiplied by] ([P.sub.t] + [P.sub.t-1]) / [P.sub.t-1] (A3)
where [P.sub.t] equals the Inventory Price Index and the subscript is used to designate the year.
Adjustments are also undertaken for the other inventory methods. Under FIFO (First in-First out), inventories at the end of the year reflect current costs:
[RINV.sub.t] = [HINV.sub.t] (A4)
Under the AVERAGE COST METHOD, inventory is reported at time t at roughly an average of the prices at t- 1 and t. In equation form, the Average Cost Method is:
[RINV.sub.t] = [HINV.sub.t] [multiplied by] 2[P.sub.t] / ([P.sub.t] + [P.sub.t-1]) (A5)
Finally under the RETAIL COST METHOD, inventory quantities are priced at the expected retail prices. Since producers normally sell at wholesale it is necessary to do the following adjustment,
[RINV.sub.t] = [HINV.sub.t] [multiplied by] WPI/RPI (A6)
where WPI equals the Wholesale Price Index and RPI equals the Retail Price Index. [RINV.sub.t] and [HINV.sub.t] are as defined previously.
LOGIT, CHOICE-based SAMPLING, WMLE, AND PAIRED T-TESTING
The logit model is derived from assuming that the random error [u.sub.i] follows a Logistic distribution. The Logistic distribution is the cumulative distribution of the hyperbolic-secant-square ([sech.sup.2] distribution whose density function is given by:
f(u) = [e.sup.u] / [(1 + [e.sup.u]).sup.2] du - [infinity] [less than] u [less than] = [infinity] (A7)
where u is an underlying random variable, and e is the exponential function. The cumulative distribution is
F(Z) = [e.sup.Z] / 1 + [e.sup.Z] (A8)
where Z is the variable of interest. One of the advantages of this distribution is that it has a closed form solution.
In the case of the logit model, the realizations of the dependent variable y (0,1) will be the realization of a binomial process with a likelihood function given by:
L = [[Pi].sub.[y.sub.i] = 0] F(-[Beta][prime] [x.sub.i]) [[Pi].sub.[y.sub.i] = 1] [1 - (F - [Beta][prime] [x.sub.i])] (A9)
where F(-[Beta][prime][x.sub.i]) is the cumulative distribution function for u. In the logit model, the x's represent the attributes of the target firm and the bidder (i.e., "q" ratios, financial ratios, etc.) that influence the probability of being acquired. We assume that these attributes are quantitatively measured. However, we are not able to observe or quantify all the characteristics that could play a role in the takeover decision. Accordingly, we assume that these unobservable characteristics are random and follow a Type I extreme value distribution. This is consistent with Palepu (1986).
The cumulative density function of the standard Type I extreme value distribution has the form: 1 - exp(-(exp(x)). The probability density function of a standard Type I extreme value distribution is very close to that of a log-normal distribution. This approach is similar to McFadden's (1974) consumer utility analysis within the context of conditional logit analysis. McFadden (1974) argues that the consumer utility function can be written in the form
U = V(s,x) + [Epsilon](s,x) (A10)
where V is nonstochastic and reflects the "representative" tastes of the population, and [Epsilon] is stochastic and reflects the idiosyncracies of this individual in tastes for the alternative with attributes x. The individual will choose the alternative that maximizes utility.
The sampling procedure used is known as choice-based sampling (other names are hold-out sampling or state-based sampling). Since the number of targets and bidders in the population is limited, the use of a random sampling procedure could exclude many of the companies whose attributes are of interest. Hence, in order to gain efficiency we have to oversample both the targets and the bidders. If the sample was random, the information provided by this sample would be very small since the majority of the firms would be non-targets or non-bidders.
The practice in previous studies dealing with prediction of targets and/or bidders has been to choose a target or a bidder and a non-target or a non-bidder from a control sample. Imbens (1992) concludes that the equal share sample is significantly better than random sampling to the extent that controlling with an equal share sample gives more relevant information. Coslett (1981) argues that the choice-based sample of equal proportions is usually a close-to-optimum design. However, if we use estimators that assume random sampling, the estimates of the model parameters could be inconsistent and asymptotically biased. The bias caused by the use of simple maximum likelihood (MLE) procedure that assumes random sampling is described by Palepu (1986) as follows:
Consider a firm i in the population with a probability p of being a target. Let p' be the probability that the firm i in the sample is a target. Using Bayes' formula for conditional probability,(16)
p[prime] = probability (i is target/i is sampled) (A11)
In the case of random sampling, the probability of the firm being sampled is the same whether it is a target or not. Hence, the above expression is equal to p. However, under choice-based sampling if [N.sub.1] and [N.sub.2] are the number of targets and non-targets in the population and [n.sub.1] and [n.sub.2] are the corresponding numbers in the sample, then
p[prime] = p([n.sub.1]/[N.sub.i]) / p([n.sub.1]/N.sub.1]) + (1 - p)([n.sub.2]/[N.sub.2] [not equal to] p (A12)
Thus, the bias is:
p[prime]-p = ([n.sub.1]/[N.sub.1])p(1 - p) / p([n.sub.1]/[N.sub.1]) + (1 - p)([n.sub.2]/[N.sub.2]) (A13)
since in most of the cases [N.sub.1] [less than] [N.sub.2] and [n.sub.1] [less than] [n.sub.2], (p[prime] - p) [greater than] 0.
Imbens (1992) presents the following example to describe the Weighted Maximum Likelihood estimator. Consider a model with two choices, i = 1,2 and two strata, s = 1,2. With probability [H.sub.1] = h an observation is drawn from strata 1, T(1) = 1}, and with probability [H.sub.2] = 1-h it is drawn from T(2) = 2}. The population probability of choice 1 is [Q.sub.1] = q, and that of choice 2 is [Q.sub.2] = 1 - q. The joint density function of (s,i,x) is
[Mathematical Expression Omitted] (A14)
where r(x) is a unknown function; compared to
P[(1/x, [Theta]).sup.I[i = 1]] [multiplied by] [(1-P(1/x, [Theta])).sup.I[i = 2]] [multiplied by] r(x) (A15)
when the sampling is random.
A proper analysis of paired data could be supplemented by taking into account the absolute value of the difference of the observations (Y = 1 and Y = 0) for each variable, to test the null hypothesis that the mean difference, [[Mu].sub.d] is [D.sub.0]. This hypothesis is equivalent to [H.sub.0] : [[Mu].sub.1] - [[Mu].sub.2] = [D.sub.0]. In this study [H.sub.o] is [D.sub.0] = 0, and the test statistic is:
[Mathematical Expression Omitted] (A16)
where d and [s.sub.d] are the sample mean and standard deviation of the n differences with n-1 degrees of freedom. We divided the units into two groups for each hypothesis (U.S. targets - U.S. nontargets, foreign bidders - foreign nonbidders, and U.S. bidders - U.S. nonbidders) based on the S.I.C. codes.
According to the hypothesis under test, corresponding observations are assumed to have approximately the same value apart from random variations, from which it follows that the differences all have the true value of 0. Further, the differences are assumed to be normally distributed with variance [Mathematical Expression Omitted], so that the mean difference,
[Mathematical Expression Omitted] (A17)
is normally distributed with parameters [Mathematical Expression Omitted] and
[Mathematical Expression Omitted] (A18)
is normally distributed with parameters (0,1). As an estimate of [Mathematical Expression Omitted] we compute the sample variance as
[Mathematical Expression Omitted] (A19)
Hence, it follows that
[Mathematical Expression Omitted] (A20)
has a t-distribution with n-1 degrees of freedom.
By choosing the pairs so that the properties of the units within each pair are similar and the properties of the units differ widely from one pair to another, we may investigate the differences or similarities of the firms involved in cross-border mergers and acquisitions. Another good property of this test is that it is not affected by measurement errors. In this sense it represents a robust test statistic. Furthermore, the variables between the pairs will not influence the variance of the mean difference, because this variance only depends on the variations of the differences between the units within the pairs.
Acknowledgment: This paper is based in part on Gonzalez's doctoral dissertation at Lehigh University. We thank the participants of the doctoral workshops at Lehigh for their comments and suggestions. Special thanks go to the two anonymous referees and the editor, Joe Finnerty, for their helpful feedback. Remaining errors are our own.
* Direct all correspondence to: Pedro Gonzalez, School of Business Administration, University of Puerto Rico at Rio Piedras, P.O. Box 21869, San Juan, Puerto Rico 00931-1869.
1. It would be simplistic to believe that the primary motivation for M&A activity internationally is simply the random search for undervalued assets. For instance, Doukas and Travlos (1988: pp 1161-1162) point out the underlying valuation of international assets stems from "(a) the firm's ability to arbitrage institutional restrictions (e.g., tax codes, antitrust provisions, and financial limitations), (b) the informational externalities captured by the firm in the conduct of international business (e.g., learning cost externalities), and (c) the cost saving gained by joint production in marketing and in manufacturing." We thank an anonymous referee for bringing this to our attention.
2. "Tobin's q" is defined as the market value of assets divided by the replacement cost of assets.
3. If we identify a high "q" ratio with good managerial performance, this conclusion is consistent with the free cash flow theory advanced by Jensen (1988). According to this theory, firms with free cash flow (i.e., a high "q" ratio) may waste resources on unprofitable investments rather than make higher cash payments to shareholders; acquisitions that force a bidder to make better use of these cash flows benefit both target and bidder shareholders.
4. See for example Lang et al. (1989), Holly and Longbottom (1988), Chappell and Cheng (1984).
5. Mergerstat classifies the companies in 50 major industrial groups based on SIC codes.
6. We recognize the possibility of the existence of the anomalies known as the "January Effect" or "Small Firm" effect. This would mean that the end of the year prices could be depressed, not reflecting the price of the stock for the rest of the year. However, we do not know a priori how large are the U.S. targets.
7. For example, if the firm pays x dollars of preferred stock dividends, and the average yield for preferred stock was 10%, the market value would be x divided by 10%. In the case of the foreign companies we will look into the possibility of using country specific preferred stock yields.
8. Here again, we will be using each country's average yield for industrial bonds. For instance, let's assume we are computing the market value of the debt for 1990. In this case, the debt maturating in 5 years is supposed to have been issued in 1975. Thus, the balance that appears in the financial statement will be discounted using the country's average yield for industrial bonds of 1975.
9. Since the book value of debt will almost always be greater than the market value of debt in real terms (i.e., after adjusting for changes in purchasing power over the life of the debt contract), the numerator of the Tobin's q ratio will be overstated in real terms making it harder to reject the null hypothesis. Thus, a significant QF builds a better case for strong firms becoming bidders in the cross-border M&A market.
10. Deferred Tax Liabilities are recorded when there is a timing difference that causes a difference between the Tax Expense and the Tax Liability. Thus, it is an accounting treatment that does not affect the equity investors. In the worst case, if the company goes bankrupt, the timing differences never will be reversed.
11. See G.S. Maddala, 1983, Limited-Dependent and Qualitative Variables in Econometrics, Econometric Society Monographs.
12. Our test results would be strengthened by a test on unsuccessful attempts and unsuccessful bidders. Unfortunately, to the best of our knowledge, this data is unavailable. Also since our data set of foreign firms as bidders is limited, we did not think that segmenting it into new acquisitions and follow-on acquisitions would have an adequate critical mass. We thank an anonymous referee for making this point.
13. A paired t-test was also performed, and there was significant difference between the Tobin's q of the targets and non-targets at a .05 level.
14. For example: if the $/DM exchange rate was on average 0.5 (i.e., DM 2.0/$) for the 1980-1990 period, and the 1988 $/DM exchange rate was 0.67 (i.e., DM 1.5/$), then the exchange rate figure to be used will be (0.67-0.5)/0.5 = .33.
15. The paired t-test shows that there was a statistically significant difference between the QF of bidders and non-bidders.
16. p' is further defined as
P(i:target) [multiplied by] P(i: sampled| i:target) / P(i:target) [multiplied by] P(i:sampled|i:target) + P(i:non-target) [multiplied by] P(i:sampled|i:non-target)
Chappell, H., and D. Cheng. 1984. "Firms' Acquisition Decisions and Tobin's q Ratio," Journal of Economics and Business, 36: 29-42.
Chirinko, R.S. 1987. "Tobin's q and Financial Policy," Journal of Monetary Economics, 19(1): 69-87.
Conn, R.I. and F. Connell. 1990. "International Mergers: Returns to U.S. and British Firms," Journal of Business, Finance, and Accounting, 17(5): 689-710.
Coslett, S.R. 1981. "Maximum Likelihood Estimation for Choice-based samples," Econometrica, 49(5): 1289-1316.
Dietrich, J.K. and E. Sorensen. 1984. "An Application of Logit Analysis to Prediction of Merger Targets," Journal of Business Research, 12: 392-402.
Doukas, J. and N. G. Travlos. 1988. "The Effects of Corporate Multinationalism on Shareholders' Wealth: Evidence from International Acquisitions," Journal of Finance, 43(5): 1161-1175.
Fabozzi, F. 1990. "Corporate Bonds." Pp. 253-287 in The Handbook of Fixed Income Securities, edited by Frank Fabozzi. Homewood, IL: Business One Irwin.
Harris, R.S. and D. Ravenscraft. 1991. "The Role of Acquisitions in Foreign Direct Investment: Evidence from the U.S. Stock Market," The Journal of Finance, 46(3): 825-844.
Holly, S. and A. Longbottom. 1988. "Company Acquisitions, Investment and Tobin's Q: Evidence from the U.K.," Journal of Economics and Business, 40(2): 103-115.
Imbens, W. 1992. "An Efficient Method of Moments Estimator for Discrete Choice Models with Choice-based Sampling," Econometrica, 60(5): 1187-1214.
Jensen, M. 1988. "Takeovers: Their Causes and Consequences," Journal of Economics Perspectives, 2(1): 21-48.
Lindenberg, E.B. and S.A. Ross. 1981. "Tobin's q Ratio and Industrial Organization," Journal of Business, 54(1): 1-32.
Lang, L.H., R.M. Stulz, and R.A. Walking. 1989. "Managerial Performance, Tobin's Q, and the Gain from Successful tender Offers," Journal of Financial Economics, 24(1): 137-154.
Maddala, G.S. 1983. Limited-dependent and Qualitative Variables in Econometrics. Cambridge, MA: Cambridge University Press.
Manski, C.F. and D. McFadden. 1981. "Alternative Estimators and Sample Designs for Discrete Choice Analysis." Pp. 51-111 in Structural Analysis of Discrete Data with Econometric Applications, edited by C.F. Manski and D. McFadden. Cambridge, MA: MIT Press.
McFadden, D., 1974, "Conditional Logit Analysis of Qualitative Choice Behavior." Pp. 113-118 in Frontiers in Econometrics, edited by Paul Zarembka. New York: Academic Press.
Palepu, K. G. 1986. "Predicting takeover targets: A Methodological and Empirical Analysis," Journal of Accounting and Economics, 8(1): 3-36.
Servaes, H. 1991. "Tobin's q and the Gains from Takeovers," The Journal of Finance, 46(1): 409-419.
Smirlock, M., T. Gilligan, and W. Marshall. 1984. "Tobin's q and the Structure-Performance Relationship," The American Economic Review, 74 (5): 1051-1060.
Tobin, J. 1969. "A General Equilibrium Approach to Monetary Theory," Journal of Money, Credit and Banking, 1: 15-29.
Tobin, J. and W. Brainard. 1977. "Asset Markets and the Cost of Capital." Pp. 235-262 in Economic Progress, Private Values and Public Policies: Essays in Honor of William Fellner, edited by B. Belassa and R. Nelson. Amsterdam: North-Holland.
Vasconcellos, G., J. Madura, and R. Kish. 1990. "An Empirical Investigation of Factors Affecting Cross-Border Acquisitions: the United States vs. United Kingdom Experience," Global Finance Journal, 1 (3): 173-189.
Vasconcellos, G.M. and R.J. Kish. 1993. "Cross-Border Mergers and Acquisitions and International Capital Flows: U.S.-Japan." Pp. 85-120 in Competitiveness and American Society, edited by Steven L. Goldman. Bethlehem, PA: Lehigh University Press.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||includes appendices|
|Author:||Gonzales, Pedro; Vasconcellos, Geraldo M.; Kish, Richard J.|
|Publication:||Quarterly Review of Economics and Finance|
|Date:||Mar 22, 1998|
|Previous Article:||On the welfare effects of mergers: short run vs. long run.|
|Next Article:||Direct foreign investment and expropriation incentives: a mitigating role for match-specific capital.|