DRAFT
Efficient Alignment and Survival in the U.S. Automobile Industry
Lyda S. Bigelow*
Olin School of Business
Campus Box 1133
Washington University
St. Louis, MO 63130-4899
(314) 935-6318
January 2000
This paper prepared for presentation at the Organization Science Winter Conference, to be held in Breckenridge, CO, February, 2000
* I would like to thank Glenn Carroll, John Freeman, Patrick Moreton, Jack Nickerson, Michael Thayer, and Oliver Williamson for helpful comments on this draft.
Efficient Alignment and Survival in the U.S. Automobile Industry
Abstract
Various authors have criticized transaction cost economics for its implied assertion that getting the governance right, i.e. organizing transactions so as to economize on transaction costs, is of paramount concern to organizations. They argue that if economizing on transactions is of such importance, then there should be some way of measuring the effect of getting the alignment right, from a TCE perspective, on firm performance. To date, there have been only a handful of empirical papers which test this link. This paper offers evidence that, yes, getting the alignment right does have profound effects on firm performance. Misalignment has a positive effect on the likelihood of firm failure. Perhaps of even greater interest, this paper also demonstrates that the effect of misalignment varies with firm age. Timing is everything. Getting the alignment right does matter, but depending on the age of the firm, getting to market may matter as well.
I. Introduction:
Understanding the trade-offs involved in relying on internal or external sourcing of components and/or services has been a central focus of the transaction cost economics research program. And indeed, many of the hazards associated with governing highly asset specific transactions by relying on an external market rather than vertical integration have been theoretically elaborated (e.g. Williamson, 1991, 1996). A recent review of the empirical transaction cost literature lists nearly 200 studies which support the central prediction that highly asset specific transactions are more likely to be vertically integrated than outsourced (Klein and Shelanski, 1995). However, in that same review, there were only a handful of studies which connected this governance decision with firm performance. In order to demonstrate the usefulness of the transaction cost lens to other organizational and strategy scholars, it is necessary to demonstrate this connection.
Of this handful of studies which investigate the relationship between transaction alignment and firm performance, most rely on survey measures of outcomes and a few look at comparative costs (e.g. Masten et al. 1991; Walker and Poppo, 1991). Very few have analyzed the link between alignment and financial performance, though early studies tested whether the M-form of organization was more profitable for diversified firms (e.g. Armour and Teece 1978). To date this author knows of only one study which has attempted to link alignment with survival (Silverman et al.,1997) and the results were inconclusive.[1]
Why this lack of empirical work in an area that many have recognized as a serious gap in the literature (e.g. Gulati, 1999, Winter, 1990)? First, there is the theoretical issue of how to link events at the transaction level of analysis with outcomes at the firm level of analysis. (See Masten, 1991 for a good summary of these concerns.) Even the simplest firm must organize multiple transactions simultaneously. How do we know which transaction is most likely to effect firm performance? In its strictest interpretation, transaction cost reasoning would suggest that each transaction could be assessed in isolation for its ability to economize on transaction costs. But recent work (e.g. Argyris and Liebeskind,1998?) suggests that there may be bundling problems (some transactions are inherently tied to related transactions) or there may be path dependent problems, i.e how transactions were organized in the past may constrain how future transactions can be organized. For the moment, carefully choosing a focal transaction is one way of controlling for these level of analysis problems. That is the argument that will be developed in this paper, though intriguing theoretical issues remain to be sorted out in future work.
Another serious impediment to undertaking empirical work in this area is the substantial data required. Gathering data at both the transaction and firm level of analysis may not always be feasible. Building in the additional requirement of creating a complete, longitudinal time series becomes a non-trivial task. Yet, it is imperative if we are to empirically test the effects of transaction alignment on firm performance.[2]
This paper uses survival as the outcome measure for two reasons.[3] The first has to do with assessing the fundamental nature of transaction alignment. If governance really matters, we ought to be able to empirically demonstrate this in a fundamental way. Survival is a significant performance measure. Though not as fine-grained as financial measures, it nonetheless represents the cumulative success of the firms operations and decisions over time in an unambiguous way.[4] Second, in the course of attempting to understand the effects of misalignment, it is difficult to ignore the case of firms that attempt to remedy their misalignment by undergoing change. This paper takes a perspective on change from organizational ecology theory’s structural inertia hypothesis. As will be argued in a following section, firms may face competing risks from either remaining misaligned or undergoing core structural change. Testing for these effects in a way that compliments the existing change literature requires using the hazard rate of failure as the dependent variable.
The primary goal of this paper is to demonstrate that getting the alignment right does indeed matter. Over time, those firms that are misaligned are more likely to fail than those which are aligned more efficiently. Therefore, although the analysis which follows is preliminary it has the potential to make a contribution simply because there are no studies to date which have empirically established this assertion.[5] Further, this paper develops the argument that integrating ideas from structural inertia perspectives on change may influence the way we perceive the viability of and the constraints on the realignment process, another area of transaction cost research which has yet to be explored.
The paper is organized as follows: The next section reviews the theoretical basis for predicting vertical integration decisions and briefly reviews the extant empirical literature on vertical integration decisions in the U.S. automobile industry. Section III describes the basis for predicting the effect of misalignment on survival and compares this to the predicted effect of core structural change. Section IV describes the industry, section V describes the data used in this analysis, section VI presents results, and section VII concludes.
II. Determining choice of organizational form using transaction cost logic
Clearly it is important to describe the logic of transaction cost reasoning which drives the determination of efficient alignment in this study. This section, then, presents a straightforward summary of transaction cost predictions on vertical integration decisions with particular emphasis on the automobile industry.[6] Transaction cost economics has emerged as the dominant theoretical framework for addressing various questions related to the make-or-buy decision, the degree of vertical integration, and specifically vertical integration within the automobile industry (Langlois and Robertson, 1989; Masten et al., 1989; Helper, 1991). Indeed, the paradigmatic case of investigating the impact of transaction characteristics on governance structure is that of vertical integration in the automobile industry. Below we provide a brief review of empirical research to date as well as relevant theoretical analyses and case studies. Contributions and gaps in the literature are summarized.
Theory and Case Studies
There have been several case studies of vertical integration in the US automobile industry, many of which focus on the 1920’s and 1930’s. Klein et al. (1978) provide the most well-known case study, that of GM’s decision to acquire Fisher Body in 1924. Others include Langlois and Robertson (1989), Helper (1991), Katz, (1977) and Raff (1991). One advantage of the case study approach is that several hypotheses concerning the observed pattern of vertical integration have been created. As each author revisits the phenomena, cumulativity reveals the importance of subtle distinctions in the emphasis each author puts on certain features or hazards of the transaction.
Helper (1991) offers an illustrative example. She argues that integration was the preferred mode of organization not because of asset specificity alone, but because it fostered asset specific learning. She then differentiates her argument from Chandler and Salisbury (1971) who disregard asset specificity but emphasize the acquisition of managerial and production capabilities (learning) and Langlois and Robertson (1989) who emphasize the specificity of the knowledge (learning) but suggest that it is the geographic proximity of integration which supports this process.
Edmonds (1923) is the first of many to use the case study approach to study the pattern of vertical integration in the US auto industry. His work identifies three firms GM, Ford, and Durant Motors as large enough to support extensive integration, but suggests that integration decisions within the over one hundred independents (e.g. Studebaker, Dodge, Willys, Hudson) varied considerably. He notes that this observed heterogeneity in form suggests that more than economies of scale are driving the integration decision. His work emphasizes hold-up hazards (though no mention of asset specificity) as a primary motivation but hypothesizes that quality control (monitoring hazard) and marketing benefits (reputation effects) may have had as much to do with the decision to integrate. He also argues that market segment has an effect as well.
Empirical Research
Several studies have empirically tested transaction cost predictions on vertical integration in the auto industry. In general, measures of asset specificity, uncertainty, frequency, complexity are modeled to assess their predicted effect on the dependent variable, often a dichotomous measure of integration, occasionally a continuous measure. Overall, the results from these studies support the theory. Asset specificity, particularly specialized technical know-how, uncertainty, as well as the size of appropriable quasi rents, all have positive effects on the integration decision.
Monteverde and Teece (1982a) collected information on 133 components to reveal a statistically significant positive relationship between the engineering effort (proxy for size of appropriable quasi rents), the specificity of a component and the likelihood that the component is produced in-house as opposed to procured from an external supplier.
The dependent variable was a dichotomous dummy variable which measured whether a component was sourced internally or externally. Since GM, Ford, and other auto producers often use both sourcing options for a given component, the authors chose a cutoff point of 80% to determine the value of the dependent variable. In other words, if 80% or more of a given component were manufactured in-house, then that component was considered to be sourced internally.[7]
Independent variables were devised as follows. To measure engineering effort, the authors surveyed two experienced design engineers asking them to rate the development cost of each component on a 10 point scale from 1 = ‘none’ to 10 = ‘a lot’. They then measured specificity by asking officials of a replacement parts wholesaler to categorize each of the components as either common across many manufacturers or relatively unique[8]. Roughly 75% of the components were considered firm-specific. The authors also argue that the complexity and degree of mechanical ties between components, i.e. systems effects, should also have a positive effect on integration due to greater coordination capability of hierarchy over market transactions. Unable to construct such a measure, the authors simply control for the type of sub-system to which each component belongs instead.
Engineering effort, firm-specificity, and company identity (in this case, GM) had a positive effect on the probability of vertical integration. Subsystem identity had virtually no significant effect. From these results the authors argue that as engineering know-how increases, firms are more likely to take production in-house because of the hazard of potential high supplier switching costs.
In a second study, using data on 28 components from two divisions of one auto manufacturer, Monteverde and Teece (1982b) find support for a hypothesized positive relationship between appropriable quasi rents and quasi-vertical integration. This hypothesis is derived from work by Williamson (1975) and Klein, Crawford, and Alchian (1978).The authors construct a proxy for appropriable quasi rents from measures of tooling cost and the degree of specialization of tooling (the percentage of tooling cost needed to convert tooling to next best alternative use). Again, the results support the notion that specialized assets create hazards which may be mitigated, in part, through increase in vertical integration.
A study by Walker and Weber (1984) uses a structural equation model to estimate the relationship between stages of the decision-making process as well as measures of transaction characteristics. Data on the decision to make-or-buy 60 components within one firm were gathered from the discussions of a committee comprised of engineering and purchasing managers entrusted with the As one pathway in their model, the authors found support for a positive relationship between volume uncertainty and a firm’s decision to make rather than buy a component. Of particular interest to this research, they also use a measure of supplier competition which is based in theory, on the number of available suppliers for a given component. The number of suppliers is considered a proxy for asset specificity. Unfortunately, rather than a true measure of supplier density, the authors rely on answers to a survey question which asks managers to estimate the extent to which there is supplier competition.
In sum, the role of asset specificity is of primary importance in determining choice of organizational form. According to Williamson (e.g. 1985, 1991), of the three characteristics which drive integration decisions- asset specificity, frequency and uncertainty- asset specificity often has the greatest impact on the magnitude of contracting hazards and thus should be of paramount importance in any analysis. In brief, as asset specificity increases, appropriability and hold-up hazards intensify, encouraging firms to bring in-house highly asset specific transactions. Thus, based on the theory and prior empirical evidence, we predict a positive relationship between asset specificity and vertical integration. It is a simple exercise to construct the first hypothesis:
H1: As the asset specificity of a primary component increases, the likelihood of the integration of that component increases.