When do Joint Ventures Create Value? - CiteSeerX

When do Joint Ventures Create Value? Partha Mohanram and Ashish Nanda1 February 1998

Abstract Analysis of a sample of 253 joint venture announcements suggests that joint ventures tend to be announced when the parent firms’ performance is deteriorating. However, the parent firms earn positive abnormal returns around the announcement date. There is considerable cross-sectional variation in the stock market reaction to joint venture announcements. We identify three drivers of stock market reaction – strategic considerations, managerial misalignment, and signaling. The stock market reacts positively to joint ventures that involve pooling of complementary resources. Joint ventures that are carried out by firms with high levels of free cash flow are received negatively, indicating that the stock market penalizes joint ventures that are susceptible to managerial misalignment. Small firms that enter into joint ventures with larger firms earn significant positive abnormal returns, because the joint ventures acts signals of the small firm’s value.

1

We are grateful to Linda Applegate, Richard Caves, Rosabeth Moss Kanter, Joshua Lerner, Chris Noe, Andre Perold, Philip Rosenzweig, HBS O&M unit members, and participants in the HBS joint ventures seminar, HBS young professors’ seminar, Harvard University Industrial Organization lunch seminar, and Academy of Management conference for their comments and suggestions. We thank Kathleen Ryan for her help with data collection and the HBS Division of Research for financial assistance. Please direct all correspondence to the authors at Harvard Business School, Boston MA 02163 [Ph: (617) 495-6506, Fax: (617) 496-4191, email: [email protected], [email protected]].

1.

INTRODUCTION

This paper looks at two issues. First, are joint ventures announced when firms are performing well or badly? Second, which joint ventures create value and which destroy value? In doing so, we go beyond prior research, which has focused on asking whether joint ventures create value in aggregate.

In order to conduct this study, we construct a data set of 253 joint venture announcements from 1986 to 1993. Unlike prior research, we focus on joint venture announcements rather than joint venture signings.

We take stock market reaction to the parent firms upon their announcing the joint venture to be the measure of value creation. Direct measures, such as market valuation of the joint venture or joint venture performance data, are not available, since joint ventures are legally reported as subsidiaries of their parent firms. It has also become increasingly clear that the traditional approach of equating longevity with value creation is flawed: successful joint ventures may be terminated early once their objectives are met, and conversely, unsuccessful joint ventures may persist for a long period of time (Gomes-Casseres, 1987; Nanda and Williamson, 1995). Researchers (e.g. McConnell and Nantell, 1985; Koh and Venkatraman, 1991; hereafter referred to as M-N and K-V respectively) have treated stock market reaction to joint venture announcement as a forward-looking measure of value creation by joint venturing. This approach has also been employed with considerable success in the related field of mergers & acquisitions. (Refer Jensen and Ruback, 1983 and Jarrell, Brickley, and Netter, 1988 for a survey of the vast literature in this area.) Research in the mergers and acquisitions literature has found evidence that ex-ante expected performance, as measured by stock market reaction, is positively related to ex-post realized performance (Healy, Palepu, and Ruback, 1992; K-V, 1991; Kaplan and Weisbach, 1992).

1

We find that joint ventures tend to be announced in times of deteriorating performance, when the parent firms are underperforming the stock market and their accounting performance is declining as well. The stock market reacts to joint venture announcements with significant positive abnormal returns.

There is wide dispersion of excess returns, influenced by three drivers: strategic considerations, signaling information, and managerial misalignment. The strategic driver refers to strategic considerations that underlie the formation of joint ventures.

Increased market power or access to complementary

capabilities can create value. The signaling driver refers to information about the parent firms that the joint venture conveys to the stock market. A joint venture between a large and a small firm may validate the worth of the small firm. Managerial misalignment refers to joint ventures that are driven by managerial motives misaligned with shareholders’ interests.

All three drivers significantly influence the abnormal returns that firms receive upon announcing joint ventures. There is a non-linear relationship between abnormal returns and the “distance” between the parent firms’ industries. Firms that are neither too close nor too distant from each other receive the highest abnormal returns, whereas firms that are very close or very distant from each other get lower abnormal returns. This is consistent with the hypothesis that the stock market values complementarity between the parent firms and inconsistent with the hypothesis that the stock market values joint ventures that increase market power.

There is a negative association between the level of free cash flow and the stock market reaction to joint venture announcements. Firms with high levels of free cash flow may be entering joint ventures because managers choose to use free cash flow in non value maximizing investments like joint ventures rather than disbursing it to shareholders. Such joint ventures are viewed negatively by the stock market.

The stock market reacts positively to small firms entering into joint ventures with larger partners. The joint venture announcement acts as a signal to the stock market of the smaller firm’s value.

2

The rest of the paper is structured as follows.

Section 2 discusses prior research on the

circumstances in which joint ventures are announced and the factors that can lead joint ventures to create or destroy economic value. Section 3 details how we arrive at our data set of joint venture announcements, describes the data set we have collected, and delineates the methodology used to derive stock market reaction to the announcements. Section 4 reports the performance of the parent companies in the time leading up to the joint venture announcement in order to understand the circumstances during which joint ventures tend to be announced. Section 5 describes the aggregate results of the event study. Section 6 discusses the findings of the cross sectional analysis to understand how the stock market’s reaction varies across different joint venture announcements. Section 7 concludes with a discussion of our results.

2

THEORY

2.1

When do firms enter into joint ventures?

Firms that are performing well may enter into joint ventures because their opportunities exceed their capacity to exploit all of them alone. Balakrishnan and Koza, 1993 (hereafter refered to as B-K), report that joint ventures occur when firms are performing well with respect to market expectations. However, it is also likely that underperforming firms use joint ventures as mechanisms to improve performance. For instance, Nanda and Williamson, 1995, point to several cases of firms that entered into joint ventures when their divisions were underperforming. It is an empirical matter as to which of these forces is predominant. In section 4, we address this issue by looking at the performance of firms entering into joint ventures in the period before the joint venture announcement.

2.2

Value creation through joint venturing

We categorize the factors influencing stock market reaction to joint venture announcements into three main ‘drivers’: strategic reasons, managerial misalignment, and signaling.

3

Strategic driver: Firms could enter joint ventures for two main strategic reasons: market power and complementarity.

Researchers have argued that entering joint ventures allow the partners to shape

competition by increasing their market power (Bresnahan and Salop, 1986; Reynolds and Snapp, 1986; Kwoka, 1992). Other researchers have stressed that joint ventures help firms access complementary assets in circumstances of market-failure (Hennart, 1988; B-K, 1993; Nanda and Williamson, 1995b).

If the stock market believes that market power is the main driver of value in a joint venture, we expect to see an inverse relation between stock market reaction and the “strategic distance” between the partners. The closer the parent firms are, the more effectively would pooling of operations help them exert market power in the product-market space they occupy.

If the stock market believes complementarity to be the main value driver, we expect to see an inverted-U relationship between partner distance and stock market reaction to the joint venture announcement. Firms that are very close simply duplicate their assets, whereas firms that are very far are too dissimilar to have complementary resources. Firms whose distance from each other is close enough to ensure complementarity but not so close as to cause duplication are likely to get the most positive reaction.

The distance measure we use to proxy for the strategic distance between the parent firms is as follows. If the parent firms have the same 4 digit SIC code, we assign them a distance of 0. If they have the same 3 digit SIC code but different 4 digit codes, we assign them a distance equal to the absolute

value of the difference of their fourth digit. To illustrate, the difference between 3678 and 3672 is 6. If they have the same 2 digit code but different 3 digit codes, we assign them a distance equal to the absolute value of the difference of their third digit times 10. Thus, the difference between 3678 and 3691 is 20 (i.e. the difference between 367x and 369y is 20 as x and y are irrelevant for the distance across 3 digit SIC codes). 4

If they have the same 1 digit code but different 2 digit codes, they are assigned a distance of the difference of their second digit times 100. Thus, the difference between 3678 and 3317 is 300 (the last two digits being irrelevant for a comparison at the 2 digit SIC code level). If their 1 digit codes are different, they are assigned a distance of 1000. The reason for this is that while there is

some ordering at the two digit level, the ordering of the 1 digit groups is arbitrary. It is difficult to argue that SIC 1000s are closer to SIC 2000s than they are to SIC 3000s. Hence we choose a uniform maximum number of 1000. The distance figure is then normalized through division by 1000, so as to lie between 0 and 1.

Although we recognize the limitations of the SIC code technique in accurately reflecting distances, we believe our approach has two main strengths. First, it allows us to use an objective, rather than a subjective, measure. Second, the distance measure recognizes that commonality at one SIC level implies an order of magnitude greater closeness than commonality at the next SIC level. Our measure takes advantage of the ordering within broad SIC classes while avoiding the pitfall of naïve distance estimation that simply takes the absolute value of the difference in SIC codes as the measure of distance.2

Managerial misalignment

Firms may enter non-value enhancing alliances if the interests of

the shareholders and the managers are not aligned. Managers may enter value-diminishing joint ventures if they expand the scope of managerial authority (empire building) or mitigate risk at the cost of lower returns. Kent (1991) reports that joint ventures perform poorer than non-joint ventures in the petroleum industry despite enjoying greater market power. Anand and Khanna (1995) report that joint ventures in the computer and telecommunications industries tend to be valued lower than other contractual forms like

2

To illustrate, a naïve approach would consider the distance between 3911 (precious metals jewelry) and 3961 (costume jewelry), to be the same as that between 3961 and 4011 (railroads). In our method, the first two would have a distance of 50/1000 and the second a distance of 1000/1000.

5

licensing and R&D agreements. Both studies suggest that a probable cause for the relatively weaker performance of joint ventures is the misalignment of the shareholders’ and the managers’ interests.

If the stock market can identify joint venture announcements that arise from managerial misalignment, we would expect it to respond by lowering its valuation for the parties announcing the joint ventures. Since the extent of free cash flow in a firm is a measure of the degree of managerial ‘slack’ in the firm, we expect a negative relation between a firm’s free cash flow level and the stock market reaction to its announcing a joint venture.

We use the excess cash measure developed by Barth and Kasznik (1996) to estimate a firm’s free cash flow. Free cash flow is defined as (cash from operations + cash from investments - cash from financing) deflated by total assets.

Signaling driver:

Apart from its intrinsic merits, a joint venture may also convey news to

the stock market about the partners themselves. The sheer fact that a small firm is able to attract the attention of a large partner into a joint venture could lead to an upward revision in the stock market’s valuation of the small firm. We use market ratio (the relative market capitalization of the partner firm to that of the firm) as a measure of the strength of the signal about the parent firm’s quality that the joint venture conveys to the stock market.

Apart from these three drivers, joint ventures might have differential value across different industries, based on the specific structure and conduct of those industries. Although M-N suggest that abnormal stock returns do not vary systematically across industries, other researchers have argued for industry specific differences. The biotechnology industry, for example, is characterized by dense networks

6

of interrelations and success in the industry requires establishing such alliances (Barley, Freeman, and Hybels, 1992; Powell and Brantley, 1992).

Conversely, research has found joint ventures in oil

exploration to be value diminishing (Kent, 1991).3 We control for industry variations by constructing industry dummies.

3

DATA

In this section, we identify our sample of joint venture announcements and describe the methodology used to derive excess returns to these announcements.

3.1

Data Set

We start with the joint venture announcements recorded by Securities Data Corporation (SDC) over the period from 1984 to 1994. SDC has a data set of alliances including joint ventures, marketing arrangements, and licensing agreements. We define a joint venture as the creation of new legal entity in which the parent firms have equity stakes. This excludes arrangements such as marketing or licensing agreements. From the chronological spread of the data, it is obvious that the SDC records are not exhaustive. SDC has been capturing a larger fraction of the universe of all joint venture announcements in recent years than in earlier years, and there is also a lag between the occurrence of an announcement and SDC’s record.

Our tests were also repeated using a simple method of assigning distance=0 for identical 4 digit SIC codes, distance=1 for same 3, but not 4 digit SIC codes, distance=2 for same 2, but not 3 digit SIC code, distance=3 for same 1 digit , but not 2 digit SIC code and diatcne=4 for totally unrelated SIC codes. The results are materially similar and not reported. (Contd.)

7

We focus on joint ventures announced during the time period from 1986 to 1993 that involve only two partners,4 both of which are U.S. corporations. We exclude joint ventures in unclassified industries (SIC codes 9000). We further narrow the data set to the 568 cases that pertain to joint ventures in which at least one of the partners is a publicly quoted company or subsidiary of a publicly quoted firm. In all these instances, we scan newswire reports to identify the first date when the joint venture was announced. Center for Research in Security Prices (CRSP) market value and returns data around the joint venture announcement date are available for 888 parent companies. Market value and returns data are available for both parents in 263 instances.

Table I.A classifies the joint ventures chronologically and Table I.B classifies them by industry. The data set is distributed over a wide range of industries: 46 2-digit SIC codes are represented, with the 6 largest industries (chemicals, computer equipment, communications, electric services, oil and gas extraction, and electrical equipment) constituting 55% of the sample.5 There are some interesting trends in joint venture formation: most parent firms in computer equipment industry use joint ventures to diversify out, and several non-service industry firms enter joint ventures classified as business services.

3

Quite apart from reasons of industry dynamics, Madhavan and Prescott (1995) suggest that industries systematically impose differential information processing loads on investment analysts. They find that joint ventures formed in industries that impose a moderate information processing load on analysts are likely to be greeted with lower abnormal returns. 4

In doing this, we differ from M-N and K-V who include multi-partner joint ventures in their studies. We follow this particular approach to exclude from our analysis multi-partner consortia which, we believe, have very different dynamics. 5

At the level of the 4-digit SIC codes of the parent firms, the four most commonly represented industries are crude petroleum and natural gas (1311 - 58 cases), electronic computers (3571 - 55 cases), pharmaceutical preparations (4813 - 29 cases), and electrical services (4911 - 29 cases); at the level of the 4-digit SIC codes of joint ventures that the firms enter into, the three most commonly represented industries are crude petroleum and gas (1311 - 47 cases), prepackaged software (7372 - 32 cases), and motor vehicle parts and accessories (3714 - 25 cases).

8

The 888 cases correspond to 468 parent firms, 162 of which appear more than once as joint venture partners.6

The 6 companies that appear most frequently in joint ventures are IBM (31

occurrences), du Pont (17), General Electric (15), Dow Chemical (13), General Motors (12), and Time Warner (10). The market values of the parent companies range from $2 million to $89 billion. The mean market value of a parent company is $9.2 billion and the median market value is $2.8 billion.7 The high median indicates that firms entering into joint ventures are large firms. The difference between the mean and median indicates that there are several firms with very high market value.

3.2

Event study approach

We use CRSP beta excess returns for the 730 firms for which they are available. The remaining firms are listed on NASDAQ. We are able to generate abnormal returns for 67 of these firms following the portfolio approach used by Fama and French (1992). We construct portfolios of firms, matched in size and risk, in order to generate excess returns. For each year, all public firms are classified into size deciles, and each size decile is divided into risk deciles based on their betas, using lagged annual returns. The excess return for each firm equals the return of that particular firm minus the mean return for its size and risk portfolio.

We conduct two event studies: (1) a long window study of monthly excess returns from 12 months before to 12 months after the joint venture announcement to observe long term trends in the stock market

6

The 263 joint ventures involve 253 parent companies, 60 of which appear more than once as joint venture partners. IBM appears as a partner in four of the ten largest joint ventures (in terms of combined parent firm size), General Electric appears in three. 7

For the 263 joint ventures, the mean and median market values of the parent firms are $10.7 billion and $3.6 billion, respectively. Firms in the joint venture sample are larger because of the requirement that data be available for both partners and the fact that data availability tends to be lower for smaller firms.

9

performance of the parent firms around joint venture announcements, and (2) a short window study of daily excess returns from 45 days before the joint venture announcement to 45 days after to observe stock market reaction to joint venture announcements.

To estimate stock market reactions at joint venture announcements, we compute daily excess returns over the event window of -45 to +45 days and then cumulate these excess returns for the periods 45 to -11, -10 to -2, -1 to +1 (the event itself), +2 to +10, and +11 to +45 for each corporation involved in the joint venture.

4

LONG WINDOW ANALYSIS : How are firms performing around the time they enter into joint ventures? An analysis of the industry and time adjusted Return on Equity and Return on Assets8 of the parent

firms in the three years leading up to the joint venture announcement shows a trend of declining performance (Figure 1).

Figure 2 plots the abnormal returns of the parent firms for the 24 months period around the joint venture announcement. Cumulative abnormal returns are negative in the months leading to and after the joint venture announcement. By month 0, the cumulative abnormal return is -8.2%, by month +12, it is 13.0%.9

8

Industry and time adjustment is done by subtracting from the unadjusted figures the median values of these variables for all firms that belong to in the same 4 digit SIC code in that year. 9

We do not allow firms to repeat in the panel within a 3 year period. If repetitions are allowed, the downward drift in returns gets even more accentuated. By month 0, the cumulative abnormal return is -9.8%; by month +12, it is -16.3%.

10

Joint ventures are akin to partial mergers, and our finding resonates with the findings of merger studies as summarized by Scherer and Ross (1990, pp. 168-170). They report that stock returns for target firms tend to drift downward relative to the stock market over a period between two years and six months before merger announcement while the stock returns for acquiring firms do not change in a systematic manner.

Our result is at variance with B-K’s findings (1993, Table 2, p. 109) that joint ventures are announced at a time when the partner firms are outperforming expectations. They report cumulative abnormal return of +1.28% between months -12 and month -6, which increases to +2.72% by month -1, and then +3.90% by month 0. These different findings suggest that, while in the mid-1970s (the B-K sample covers 1974 to 1977) joint ventures were announced when firms were doing well, in the late 1980s-early 1990s, whereas firms entered joint ventures while they were doing poorly.10

5

EVENT STUDY RESULTS

In this section, we report the results of the event study. We report abnormal returns at firm level and joint venture level, and then address whether economic value is created in aggregate.

10

Kothari and Warner (1997) point out that long window event study results should be interpreted with caution. They argue that persistent drift in cumulative abnormal returns can occur both in the same direction as the announcement effect as well as in the opposite direction in long-window event studies of securities, because the market model does not adequately control for size and book-to-market effects (Fama and French, 1993). In order to ensure that the negative drift we observe is driven by economic reasons rather than because of weaknesses in the model used by CRSP to estimate abnormal returns, we also conduct the long window test by calculating abnormal returns by the portfolio approach discussed in Section 3.2 above. The downward drift persists, although it is attenuated. Cumulative abnormal returns are -4.50% by month 0 and -6.47% by month +12.

11

5.1

Firm level results

Panel 1 of Table II presents the results for the 797 parent firms for which we have abnormal returns. Figure 3.A plots the daily abnormal returns and Figure 3.B plots the cumulative abnormal returns over the period day -45 to day +45. Firms experience significant positive abnormal return of +0.49% during the event period (defined as the actual announcement date and the two days around it).

In order to ensure independence of returns within the panel of available data, we also conduct the event study for a restricted sample. For parent firms that appear more than once during a three year period, we retain only the earliest occurrence of the firms in that three year period. The results for the panel of 431 parent firms, reported in Panel 2 of Table II, are similar. The abnormal return around the event date is +0.40%. Its significance drops, however, owing to the smaller sample size.

Our result is in the same direction but weaker in magnitude, compared with the results of earlier studies: M-N (1985) report an excess return of 0.73% for the two day window of -2 to 0, whereas K-V (1991) report an abnormal return of 0.87%. These numbers may appear small, but one must take into account the fact that these are returns over very small intervals: 0.49% over three days corresponds to an annualized return of 81%.11

Our finding of a lower magnitude of stock market reaction to joint venture announcements, compared to prior papers, may be because of two reasons. First, our sample firms are larger compared to the sample firms of the earlier studies. As we see in section 6.1 below, abnormal returns are negatively correlated with firm size.. Second, our data set is more recent (1986-93) compared to M-N’s (1972-79)

11

Compared to the excess returns that greet merger announcements, the excess returns to joint venture announcements are indeed small. Asquith (1983), for instance, reports 2-day excess returns of +6.5% for the acquired firms and +0.3% for the acquirers. Since joint ventures are akin to merger of only fractions of the parent firms, the difference between stock market reaction to joint ventures and mergers is consistent with expectations.

12

and K-V’s (1972-86) data sets. The magnitude of the stock market’s reaction may have differed during these time spans.

As Table II indicates, the positive returns during the event window appear amidst a context of negative abnormal returns. Joint ventures are viewed, in general, as positive events announced during a time frame time when firms are losing value with respect to the market. The joint venture announcements only temporarily arrest the slide the firms are in.

Our finding, that joint ventures are announced at a time when the parent firms are underperforming on a market adjusted basis, is in contrast to M-N’s finding that joint ventures occur at a time when the firms are performing well. M-N report CAR of +0.94% by day -10, +2.15% by day 0, and +2.89% by day +10 (Table III, p. 526). Their finding of a positive drift in abnormal returns over a short window parallels B-K’s finding over a long window.

The data samples of the two studies cover

overlapping time-spans (1972-79 for M-N’s sample and 1974-77 for B-K’s sample). It appears that, whereas in the 1970s joint ventures may have been announced when firms were performing well, this trend has been reversed during the late 1980s-early 1990s.

5.2

Joint venture level results

In order to understand how the market values the joint venture as a whole, we aggregate the abnormal returns of both parties involved in the joint venture. For each joint venture, we derive an excess return figure by weighting the excess returns of the partners by their market value, and then calculating the arithmetic mean across all joint ventures. 12 For the 253 joint ventures in the sample, the abnormal returns

12

Equally weighting the returns of both partners in the joint venture would bias the joint venture returns in favor of the returns experienced by the smaller partners. For example, if the two partners in a joint venture have market values of $100 million and $10 million, and experience excess returns of -0.1% and +1% respectively, then the market has placed an aggregate (Contd.)

13

are reported in Panel 3 of Table II and plotted in Figures 3.A and 3.B. The joint ventures receive positive but non-significant abnormal return of 0.19% around the event date.

If we remove joint ventures from the sample that have one or both parents appearing in another joint venture up to 3 years before, in order to ensure independence of data points within the panel, the sample size shrinks to 82 joint ventures. The excess returns for this panel are also reported in Panel 4 of Table II. The joint ventures receive positive but non-significant abnormal return of +0.40% around the event date.

5.3

Value creation through joint venture announcement

In dollar terms, the average excess return during the event period is $17.8 million. This finding is in consonance with M-N’s finding of an average value gain of $4.8 million (p. 527) and K-V’s finding of an average value gain of $12.6 million (p. 881).

Although the dollar abnormal returns are not significant at the aggregate level, the range of value creation is very broad. Abnormal returns over the three day event window range from negative $2.97 billion to positive $7.0 billion. The stock market views joint ventures as significant events, though not as unequivocally good or bad events. 13

valuation of $0 on the two partners taken together. Equally weighting the partners would yield an abnormal return figure of +0.45% for the joint venture. Since M-N (1985) create equally weighted portfolios, we believe their excess return results get biased in the direction of the smaller firms.

14

6

CROSS SECTIONAL ANALYSIS OF EXCESS RETURNS

In this section, we attempt to identify factors that distinguish joint venture announcements that receive negative stock market reaction from announcements that receive positive reaction by testing the model developed in Section 2.2.

( EXRETi ) j = β0 + β1 DISTij + β2 ( DISTij )

2

(

)

+ β3 FCFi + β4 MKRATIOij + β5 INDij + εij

(EXRETi)j is the abnormal return experienced by the firm i around the date of the announcement of a joint venture with firm j. DISTij is the “strategic distance” between the two parent firms, FCFi is the free cash flow ratio of firm i, MKTRATIOij = MV j / MVi measures the ratio of the market capitalization of partner firm j to that of firm i, INDij is the industry dummy, and εij is random noise.

If market power consideration drives abnormal returns, we expect the coefficient β1 to be negative and have no apriori expectation about the coefficient β2 .

If complementarity considerations drives

abnormal returns, we expect β1 to be positive and β2 to be negative. The managerial misalignment reasoning suggests that the coefficient β3 is negative, and the signaling argument suggests that the coefficient β4 is positive.

In order to ensure independence of observations, we remove from the sample those firms that have appeared earlier in the sample as parents of other joint ventures three years before or less. Out of this panel of 431 firms, accounting data, from which the free cash flow metric is constructed, is available for only 213 firms. The regression analysis is restricted to these 213 firms.

13

We construct a standardized squared abnormal returns statistic as in Warner, Watts, and Wruck (1988) to test whether the abnormal returns are significantly distributed around the mean. We find that the squared returns statistic is highly (Contd.)

15

The 213 firms announced joint ventures in 94 industries. We construct specific industry dummies for the 9 industries in which 5 or more joint ventures were announced; all other industries are grouped together in one category.

6.1

Summary statistics

Table III-A presents univariate statistics for the variables.

The medians for EXRET and

MKTRATIO are considerably less than the means, indicating that the distributions of both variables are right-skewed. Table III-B presents the correlations between the variables. As expected, EXRET is negatively correlated with FCF. MKTRATIO and FCF are also negatively correlated, indicating that small firms tend to be capital constrained and large firms tend to have proportionately more free cash.

On partitioning the sample of firms into three groups based on their market capitalization – small (the smallest 25%), medium (the middle 50%) and large (the largest 25%) – we find that stock market reaction to joint venture announcement is highest for the small firms, middle sized firms earn negative abnormal returns, and large firms earn positive but small abnormal returns (Table III-C). Overall, there is a significant negative correlation between abnormal returns and firm size.14

This result is consistent with M-N’s finding that medium-sized firms earn the lowest abnormal returns. K-V also find that small firms earn the highest abnormal returns; however, they find a strictly declining relation between abnormal returns and firm size. (See Table III-D for a comparison of these findings.)

significant, suggesting that excess returns are indeed widely dispersed. 14

This explains why average abnormal returns in reaction to the joint venture announcement decline as we move from firm-level analysis to joint venture level analysis. Firm level results weight all abnormal returns equally. Joint venture level results market value weight the partners’ excess returns, thereby giving greater weightage to larger firms. Hence, the average abnormal return figure is lowered.

16

6.2

Regression analysis Table IV reports the regression results. In Model 1, the independent variables are DIST, DIST2,

MKTRATIO, and FCF. The regression has an R 2 of 5.94%. DIST has a significant positive coefficient and DIST2 has a significant negative coefficient. The stock market values complementarity considerations rather than market power considerations in evaluating the strategic worth of a joint venture. MKTRATIO has a significant positive coefficient, supporting the signaling hypothesis.

The stock market reacts

positively to announcements by small firms that they are entering joint ventures with larger partners. FCF has a significant negative coefficient, supporting the managerial misalignment hypothesis. The stock market reacts negatively to joint venture announcements by firms that have a large amount of free cash.

Introducing the industry dummies does not improve the predictive power of the regression. The regression including industry dummies has an R 2 of 5.68% (Model 2). The only industry dummy with a significant coefficient is IND5045, suggesting that the stock market values joint venture announcements in the computer and software wholesale industry more than joint venture announcements by similar parent firms in other industries. Controlling for industries, the coefficients for DIST, DIST2, and FCF remain significant and change only slightly from Model 1. Although the coefficient for MKTRATIO is of the same magnitude as in Model 1, it is no longer significant in Model 2.

In order to better understand the effects of “distance”, the regression is also rerun using dummies for the following four categories of variables •

SAME4 : Firms with identical 4 digit SIC codes

•

SAME3 : Firms with the same 3 digit SIC code, but different 4th digit

•

SAME2 : Firms with the same 2 digit SIC code, but different 3rd digit

17

•

SAME1 : Firms with same 1st digit SIC code, but different 2nd digit SIC code

This method has two advantages over using the distance proxy measure. First, it is less prone to specification error. In particular, it considers distance to be an ordinal measure and does not cardinalize distance as DIST does. Second, it allows one to observe the categories separately to see when and if returns start to decline. The results are presented in Model 3. The intercept refers to the baseline return for firms in totally different SIC codes (0.4%). Firms with the same 4 digit SIC codes receive a significant negative return of -1.5% lower than the baseline case. Firms with the same 3 digit (but not 4 digit) SIC codes receive significant positive excess returns of 1.1% over the baseline case. The returns drop off to 0.8% for same 2 digit, and 0.6% for the same 1 digit. These results confirm the results using the distance variable. The best received joint ventures are those for which the parent firms are at an optimal distance to ensure complementarity while preventing duplication. The results indicate that this optimal distance is similarity at the 3 digit SIC code level. The negative returns for identical SIC codes also indicates that market power considerations are not important. The results for MKTRATIO and FCF are unchanged from before. In order to estimate the relative explanatory power of the different drivers, we conduct a test akin to the one undertaken by Schmalensee (1985). We run a set of regressions, taking different combinations of the three drivers – strategic, signaling, and managerial misalignment – with abnormal returns as the dependent variable. Table V summarizes the results. All three drivers contribute significantly to the regressions, suggesting that the stock market takes all three drivers into consideration in valuing a joint venture announcement. In order to estimate the relative importance of each of the drivers, we calculate the

18

Shapley Value15 of the incremental contribution of each driver to the explanatory power of the regression. The strategic driver is the strongest determinant of stock market reaction; the managerial misalignment and signaling drivers each contribute about two-third as much as the strategic driver. If we control for industry, the strategic and managerial misalignment drivers are almost equally important determinants of stock market reaction, and the signaling driver contributes about half as much.

7

DISCUSSION

What does this paper add to the vast literature in the area of joint ventures? We show that joint ventures tend to be announced in times of deteriorating performance, when firms are underperforming the stock market and their accounting performance is declining as well. Firms entering joint ventures receive positive stock market reaction in aggregate, but there is considerable cross-sectional variation. We find support for the hypotheses that three drivers – strategic considerations, managerial misalignment, and signaling rationale – significantly influence stock market reaction.

This paper has normative implications for managers of firms looking at potential joint ventures. It suggests that whereas some kinds of joint ventures are value creating, others can be value diminishing. Joint ventures create value if they help firms to overcome market failure in accessing complementary assets. Signaling also plays a key role in influencing the stock market reaction. Small firms are benefited by joint venturing with larger partners because, quite apart from the worth of the joint venture itself, the fact that a large company has chosen the small company as a partner raises the value of the small firm in the eyes of the stock market. On the other hand, managerial misalignment destroys value. Managers of firms with lots of free cash flow should be circumspect about entering joint ventures.

15

The Shapley value is a measure used in cooperative game theory to gauge the contribution of a player to a coalition.

19

Our study also challenges some conventional wisdom in the field. Some managers perceive joint ventures as value-creating options, allowing participation in the potential upside without having to invest wholly in a project. We find that the stock market reacts unfavorably to joint venture announcements by firms that have the free cash flow to undertake the projects on their own. Core competency, “stick to your knitting,” and market power arguments would suggest that joint ventures among closely related parent firms should create value. Instead, we find that there is an optimal “distance” between the parent firms, to ensure that both parents bring capabilities to the table that are valuable, different, and mutually complementary.

We plan to extend our study along two dimensions. First, we plan to examine whether and how administrative considerations influence value creation.

This paper hasn’t considered issues of

administrative fit, cultural match, alignment of management philosophies, and past experience with joint ventures. Stock market reaction to joint venture announcements displaying firm specific effects would be an indication that administrative considerations do influence the stock market. The current study hints that such may be the case. Of the sample of 253 joint ventures, 21 have IBM as one of the parent firms whereas 7 have AT&T. IBM received positive stock market reaction on 14 of the 21 joint ventures, averaging an abnormal return of +1.04% across the 21 joint ventures; AT&T received negative stock market reactions on 6 of the 7 joint ventures, averaging an abnormal return of -1.01% across the 7 joint ventures. Could such differences exist because the stock market differentially values firms’ ability to manage joint ventures? We plan to repeat our test after including in the set of independent variables proxies for the administrative driver. Second, the current study focuses on stock market’s valuation of a joint venture at the time it is announced. It would be intriguing to study how joint venture performance measures against this ex-ante valuation. We plan to examine ex-post performance of the announced joint ventures and relate this to ex-ante stock market predictions.

20

REFERENCES •

Anand, B., and Khanna, T., 1995, “On the Market Valuation of Inter-Firm Agreements : The Computer and Telecommunication industries, 1990-1993", Working Paper, July 20, 1995.

•

Asquith, P., “Merger Bids, Uncertainty, and Stockholder Returns,” Journal of Financial Economics, 11(1-4), p. 51-83, 1983 April.

•

Balakrishnan, S., and Koza, M.A., 1993, "Information Asymmetry, Adverse Selection, and JointVentures," Journal of Economic Behavior and Organization, January, v, pp. 99-117.

•

Barley, S.R., Freeman, J., and Hybels, R.C., 1992, "Strategic Alliances in Commercial Biotechnology," in Nohria, N., and Eccles, R.G., eds., Networks and Organizations: Structure, Form, and Action, HBS Press: Boston.

•

Barth, M., and Kaznik, R., “Share Repurchases : Accounting Deficiencies and Information Effects,” Stanford University Working Paper, 1996.

•

Bresnahan, T.F., and Salop, S.C., 1986, “Quantifying the Competitive Effects of Production Joint Ventures,” International Journal of Industrial Organization, 4, pp. 155-175.

•

Fama, E.F., and French, K.R., 1992, “The Cross-Section of Expected Stock Returns,” Journal of Finance, 17, pp. 427-465.

•

Fama, E.F., and French, K.R., 1993, “Common Risk Factors in the Returns on Stocks and Bonds,” Journal of Financial Economics, 33, pp. 3-56.

•

Gomes-Casseres, B., 1987, " Joint Venture Instability: Is It a Problem?" Columbia Journal of World Business, Summer, v22n2, pp. 97-102.

•

Healy, P., Palepu, K., and Ruback, R., 1992, "Does Corporate Performance Improve After Mergers?", Journal of Financial Economics,v32, pp 135-175.

•

Hennart, Jean-Francois, 1988, "A Transaction Costs Theory of Equity Joint Ventures," Strategic Management Journal, July/August, v9n4, pp. 361-374.

•

Jarrell, G.A., Brickley, J.A., and Netter, J.M., 1988, “The Market for Corporate Control: The Empirical Evidence Since 1980,” Journal of Economic Perspectives, v2, Winter, pp. 49-68.

•

Jensen, M.C., and Ruback, R.S., 1983, “The Market for Corporate Control: The Scientific Evidence,” Journal of Financial Economics, v11n1, pp. 5-50.

•

Kaplan, S.N., and Weisbach, M.S., 1992, “The Success of Acquisitions: Evidence from Divestitures,” Journal of Finance, v47n1, March, pp. 107-138.

•

Kent, D.H., 1991, " Joint Ventures vs. Non- Joint Ventures: An Empirical Investigation," Strategic Management Journal, July, v12n5, pp. 387-393.

21

•

Koh, J., and Venkatraman, N., 1991, “ Joint Venture Formations and Stock Market Reactions: An Assessment in the Information Technology Sector,” Academy of Management Journal, 34:4,pp. 869892.

•

Kothari, S.P., and Warner, J.B., 1997, “Measuring Long-Horizon Security Price Performance,” Journal of Financial Economics, forthcoming.

•

Kwoka, J.E., Jr., 1992, "The Output and Profit Effects of Horizontal Joint Ventures," Journal of Industrial Economics, September, v40n3, pp. 325-338.

•

Madhavan, R., and Prescott, J.E., 1995, “Market Value Impact of Joint Ventures: The Effect of Industry Information-Processing Load,” Academy of Management Journal, v38n3, pp. 900-915.

•

McConnell, J.J., and Nantell, T.J., 1985, “Corporate Combinations and Common Stock Returns: The Case of Joint Ventures,” Journal of Finance, XL:2, June, pp. 519-536.

•

Nanda, A., and Williamson, P.J., 1995, Use Joint Ventures to Ease the Pain of Restructuring,” Harvard Business Review, v73n6, Nov-Dec, pp. 119-128.

•

Nanda, A., and Williamson, P.J., 1995b, “Corporate Restructuring Through Transitory Ventures,” HBS Working Paper 94-062.

•

Powell, W.W., and Brantley, P., 1992, "Competitive Cooperation in Biotechnology: Learning through Networks?" in Nohria, N., and Eccles, R.G., eds., Networks and Organizations:

Joint

Structure, Form, and Action, HBS Press: Boston. •

Reynolds, R., and Snapp, B., 1986, "The Economic Effects of Partial Equity Interests and Joint Ventures," International Journal of Industrial Organization, 4, pp. 141-153.

•

Scherer, F.M., and Ross, D., Industrial Market Structure and Economic Performance, Houghton Mifflin: Boston, 1990.

•

Schmalensee, R., 1985, “Do Markets Differ Much?” American Economic Review, v75n3, pp. 341-351.

•

Warner, J.B., Watts, R.L., and Wruck, K.H., 1988, “Stock Prices and Top Management Changes,” Journal of Financial Economics, v20n1/2, January-March, pp. 461-492.

22

TABLE I: DATA DESCRIPTION TABLE I.A: CHRONOLOGICAL DISTRIBUTION OF THE SAMPLE Year

Joint Ventures

Firms

No.

%

No.

%

1986

20

8

70

8

1987

15

6

41

5

1988

20

8

71

8

1989

19

7

61

7

1990

52

20

166

19

1991

73

28

253

28

1992

31

12

109

12

1993

33

12

117

13

Total

263

100

888

100

TABLE I.B: DISTRIBUTION OF SAMPLE ACROSS INDUSTRIES 2-digit SIC Code

Industry

Joint Ventures

Firms

Partner SIC

JV SIC

28

Chemicals and allied products

29

111

90

35

Ind Comml Machinery & Computer Eqp

12

102

52

48

Communications

21

76

59

49

Electric, Gas, and Sanitary Services

16

68

52

13

Oil and Gas Extraction

17

66

60

36

Elec & Elec Eqpmnt except Computers

29

64

76

37

Transportation equipment

27

55

70

38

Measuring Instr, Photo Goods, & Watches

10

40

28

27

Printing, Publishing, & Allied

4

40

18

73

Business Services

27

19

83

51

Nondurable Goods Wholesale

10

5

33

Others

61

242

267

Total

263

888

888

23

FIGURE 1: INDUSTRY AND TIME ADJUSTED ACCOUNTING PERFORMANCE OF PARENT FIRMS Year Before Joint Venture -3

-2

-1

0

%

0.0%

-1.0%

-2.0% Return on Equity

Return on Assets

FIGURE 2: CUMULATIVE ABNORMAL RETURN (LONG WINDOW) 12

10

8

6

4

2

0

-2

-4

-6

-8

-10

-12

0.0%

-2.0%

-4.0%

%

-6.0%

-8.0%

-10.0%

-12.0%

-14.0% Month

Months -12 and +12 receive negative abnormal returns at 1% level of significance. Months -6, -5, -2, +2, and +6 receive negative abnormal returns at 5% level of significance. Month -8 receives negative abnormal returns at 10% level of significance.

24

TABLE II : EVENT STUDY RESULTS

FIRM LEVEL RESULTS

PANEL 1 TIME PERIOD

REPETITIONS

PANEL 2

(797 observations)

NO REPETITIONS (431 observations)

EXCESS RETURN (%)

(T STAT)

EXCESS RETURN (%)

(T STAT)

-45 TO -11

-0.74

(-1.55)

-0.87

(-1.29)

-10 TO -2

+0.17

(+0.68)

-0.02

(-0.07)

b

(+2.23)

+0.40

(+1.27)

(-0.98)

-0.08

(-0.19)

(-3.41)

c

(-1.82)

-1 TO +1

+0.49

+2 TO +10

-0.29

+11 TO +45

a

-1.69

-1.27

JOINT VENTURE LEVEL RESULTS PANEL 3 TIME PERIOD

a b c

REPETITIONS

PANEL 4

(253 observations)

NO REPETITIONS (82 observations)

EXCESS RETURN (%)

(T STAT)

EXCESS RETURN (%)

(T STAT)

-45 TO -11

-0.12

(-0.18)

+0.93

(+0.69)

-10 TO -2

+0.39

(+1.26)

+0.01

(+0.02)

-1 TO +1

+0.19

(+1.12)

+0.40

(+1.32)

+2 TO +10

-0.47

(-1.69)

+0.05

(+0.11)

+11 TO +45

a

(-2.73)

+0.34

(+0.37)

-1.46

Different from 0 at 1% level of significance Different from 0 at 5% level of significance Different from 0 at 10% level of significance

Excess returns at the joint venture level are value weighted. The repetitions column includes all announced joint ventures for which we have abnormal returns. In the no repetitions column, for parent firms that appear more than once during a three year period, we retain only the earliest occurrence of the firms in that three year period.

25

FIGURE 3A: DAILY ABNORMAL RETURNS 0.20%

FIRM LEVEL 0.10% %

FIRM LEVEL

0.00% -60

-40

-20

0

20

40

60

JV LEVEL

-0.10% DayJV

LEVEL

FIGURE 3B: CUMULATIVE ABNORMAL RETURNS 1.00%

0.00% -60

-40

-20

0

20

40

60

%

JV LEVEL -1.00%

-2.00%

FIRM LEVEL -3.00% Day

26

SUMMARY STATISTICS TABLE III-A: UNIVARIATE STATISTICS Variable EXRET DIST MKTRATIO FCF

Mean 5.59% 0.526 45.45 0.023

Std. Dev. 4.01% 0.459 256.69 0.239

Median 0.18% 0.550 1.66 0.036

TABLE III-B: BIVARIATE STATISTICS

EXRET DIST MKTRATIO FCF

EXRET 1 0.08 0.04 -0.04

DIST 0.11 1 0.08 -0.04

MKTRATIO 0.35 0.12 1 -0.14a

FCF -0.15a -0.06 -0.16a 1

Pearson correlation coefficients are reported above the diagonal; Spearman correlation coefficients are reported below the diagonal. a

Different from 0 at 5% level of significance.

TABLE III-C: SIZE PARTITIONS Category Small (bottom 25% by market value) Medium (middle 25% by market value) Large (top 25% by market value)

N 118 234 117

Abnormal Return +1.62% -0.20% +0.31%

TABLE III-D : COMPARISON OF SIZE FINDINGS WITH OTHER PAPERS Source

SMALL Avg. firm Abn. ret. size ($b) (%)

This Paper M-N

a

K-Vb a

Table V, p. 532

MEDIUM Avg. firm Abn. Ret. size ($b) (%)

LARGE Avg. firm Abn. ret. size ($b) (%)

0.16

+1.62

2.09

-0.20

17.20

+0.31

0.63

+1.10

1.01

+0.57

3.41

+0.63

1.43

+1.13

6.49

+0.94

10.01

+0.44

b

Table 7, p. 887

27

TABLE IV : REGRESSION FOR EXRET The dependent variable is EXRET (abnormal return). Figures in parentheses represent t-statistics. Independent Variable INTERCEPT DIST DIST2

Model 1 -0.004 (-0.83) 0.099 b (2.54) -0.089 b (-2.37)

Model 2d -0.002 (-0.43) 0.097 b (2.38) -0.089 b (-2.24)

Model 3 0.004 (0.94)

-0.015c (1.71) 0.011c (1.81) 0.008 (0.99) 0.006 (0.72) 2.72*10-5 b (2.47) -0.020c (1.81)

SAME4 SAME3 SAME2 SAME1 MKTRATIO FCF

2.23*10-5 b (2.14) -0.024 b (-2.15)

5.94%

1.71*10-5 (1.23) -0.026 b (-2.01) 0.009 (0.39) 0.003 (0.22) -0.011 (-0.86) -0.024 (-0.92) -0.005 (-0.57) -0.011 (-0.56) 0.034 a (2.62) -0.010 (-0.55) 0.004 (0.35) 5.68%

4.36 213

1.99 213

IND1311 IND3674 IND3714 IND3721 IND4911 IND4922 IND5045 IND7372 IND7373

R2 F-value Number of Observations a b c

4.20% 2.55 213

Different from 0 at 1% level of significance. Different from 0 at 5% level of significance. Model 2 is heteroscedastic at 5% level of confidence. The t-statistics have been corrected for heteroscedasticity.

28

TABLE V : RELATIVE CONTRIBUTION OF THE THREE DRIVERS DEPENDENT VARIABLE:

EXRET

STRATEGIC, MANAGERIAL, & SIGNALING R 2 =5.94 4.60

3.65

4.62 STRATEGIC & MANAGERIAL

STRATEGIC & SIGNALING

MANAGERIAL & SIGNALING

R = 4.33

R = 4.32

R 2 = 358 .

2

4.21

2

4.19

3.71

4.10

4.73

5.52

STRATEGIC

MANAGERIAL

SIGNALING

R = 2.88

R =189 .

R 2 =152 .

2

2

5.09 4.30

4.15 NULL MODEL

N = 213

INDEPENDENT VARIABLES STRATEGIC: DIST, DIST2 MANAGERIAL: FCF SIGNALING: MKTRATIO

Figures on the arrows give the statistic at which an F-test rejects the null hypothesis that the incremental driver added in the box to which the arrowhead points does not contribute added explanatory power to the regression. The null hypothesis is rejected at the 5% significance level for all the F-tests.

CONTRIBUTIONS OF THE THREE DRIVERS TO R 2 VALUES OF REGRESSIONSa Driver STRATEGIC MANAGERIAL SIGNALING

Shapley value of contribution 2.62% 1.76% 1.66%

a

Relative contribution to explanatory power 43% 29% 27%

If the analysis is conducted controlling for industry, the three drivers still contribute significantly to the regressions. The relative contributions to the explanatory power of the strategic, managerial, and signaling drivers become 38%, 41%, and 21%, respectively.

29