Accounting for Performance in Contracting for Services: Are Successful Contractual Relationships Controlled or Managed? Sergio Fernandez School of Public and Environmental Affairs Indiana University 1315 E. Tenth Street Bloomington, Indiana 47405-1701 812-856-4873 Tel. 812-855-7802 Fax sefernan@indiana.edu Paper presented at the 8th Public Management Research Conference, Los Angeles, California, and hosted by the School of Policy, Planning, and Development at the University of Southern California, September 29 to October 1, 2005. The term privatization refers not to a single idea or innovation, but rather to a broad group of alternative service delivery arrangements (Savas, 1987; 2000; ICMA, 1989; Seidenstat, 1999; Lavery, 1999; Sclar, 2000; Hodge, 2000). In the United States, the most pervasive form of privatization is contracting for services (Miranda and Andersen, 1994; Martin, 1999). Experts have observed a huge upsurge in contracting for services at all levels of government during the last three decades (Savas, 1987, 2000; Osborne and Gaebler, 1993; Hanrahan, 1983; Kettl, 1993; Kelman, 2002). Every year, hundreds of billions of dollars in services are contracted out to forprofit and non-profit providers. Moreover, the contracting trend is not a fad that is likely to dissipate in the foreseeable future. Indeed, its use continues to spread in the United States and has gained a significant degree of legitimacy in the American system of governance (Kettl, 2000; Donahue, 2002; Kelman, 2002; Martin, 1999; Brudney, et al., 2005; Miranda and Andersen, 1994). These developments have created a significant demand for research on the topic of contract management and contracting performance. What factors account for success in contracting for services? Although researchers have produced a large body of work on contracting out, the answer to this question remains elusive for several reasons. First, considerable confusion has been caused by competing theoretical perspectives on contracting, each with its own set of propositions about the determinants of contracting performance. In addition, most of the empirical research on the outcomes of contracting for services has focused narrowly on efficiency or quality of service, while neglecting other important outcomes such as responsiveness to the government’s requirements, legal compliance, and customer satisfaction (see Hodge, 2000). Researchers also have exhibited a strong tendency to approach the issue of contracting performance from a single theoretical perspective (e.g., principal-agent theory) and to focus on a small set of explanatory variables 2 (e.g., competition, contract monitoring capacity). With the exception of a handful of studies (e.g., Domberger and Hensher, 1993; Romzek and Johnston, 2005; 2002), there have been no systematic efforts to develop a comprehensive explanatory model of contracting performance and test it empirically. Recent large n studies by Brown and Potoski (2002, 2003), Hefetz and Warner (2004), and Brudney, et al. (2005) represent important contributions to the literature, but none of these studies tried to explain contracting performance. This paper seeks to meet the challenge of accounting for success in contracting for services by developing a comprehensive model of performance in this area and testing the model using a large n dataset of contractual relationships between local governments and private service providers in the United States. The discussion begins with an overview of the literature on privatization and contracting out. Then, a model of contracting performance is developed inductively from different streams of research. The methodology and data are then described, followed by the analysis and its results. The paper concludes with a discussion of the limitations of the study and suggestions for future research on the topic of contracting performance. Overview of the Privatization Literature Interest in privatization has spawned a huge literature, with different streams of research focusing on a variety of issues relating to privatization. A sizable number of studies have focused on the conditions under which to adopt privatization (Moe, 1987; Cohen, 2001; Donahue, 1989; Brown and Potoski, 2003). This stream of research points to a variety of factors elected officials and public managers should consider in deciding whether or not to contract out a service or function, such as whether or not the service is inherently governmental in nature, the potential for moral hazard and adverse selection problems, and the government’s contract monitoring capacity. A number of other studies have explored the determinants of privatization 3 (see Boyne, 1998; Warner and Hebdon, 2001; Brudney, et al., 2005). Research findings indicate that a variety of factors influence the adoption of contracting out, including fiscal stress and fiscal capacity of governments; the strength of public unions and size of the public work force; contract management capacity; the density of markets for private service providers; political factors, such as the ideology and political affiliation of citizens and elected officials; and characteristics of the services that are contracted out. The largest stream of research on privatization consists of studies comparing public with private service delivery in terms of efficiency and quality of service. While some experts have concluded that the evidence points to contracting out as more efficient than in house service delivery (Savas, 1987, 2000; Siegel, 1999; Greene, 2002; Dilger, Moffett, and Struyk, 1997), others maintain that the evidence in more mixed and inconclusive (Hodge, 2000; Sclar, 2000; Brudney, et al., 2005). Regarding quality of service, the evidence is limited and also mixed, with some studies finding improvements in quality of service through contracting out (e.g., Dilger, Moffett, and Struyk, 1997), no improvement (Hodge, 2000), and a decline in the quality of service (Kamerman and Kahn, 1989). In short, improvements in efficiency and quality appear to be contingent on a variety of factors, in particular, how well public managers manage the contracting process. Other privatization experts have focused on human resources management issues, including public employee and union opposition to privatization and the effects of contracting out on public sector employment, wages, and benefits. This literature indicates that public employees and their unions are ardent opponents of privatization (American Federation of State, County, and Municipal Employees, 2004; Savas, 2000; Rehfuss, 1989; Chandler and Feuille, 1991; Fernandez and Smith, 2005), even though there is scant evidence that large numbers of 4 public employees lose their jobs as a result of privatization in the United States (National Commission for Employment Policy, 1989; Donahue, 2002; Light, 1999; Greene, 2002). Developing a Model of Contracting Performance In the privatization literature, one also finds may empirical studies and other publications that speak to the issue of performance in contractual relationships. Contributions to this stream of research come from a variety of fields or areas of study, including public administration, public policy, business, generic management, economics, and legal studies. Some of this work is published in academic journals, monographs, and edited volumes, but a lot of it also can be found in government reports and publications intended for an audience of practitioners and policymakers. Trying to bring some order to this seemingly amorphous collection of work is no small undertaking. However, a closer reading of the literature reveals two distinct theoretical perspectives or approaches to contracting: the conventional or classical approach to contracting and the more recent relational or collaborative perspective. In addition, policy implementation research offers its own set of insights regarding successful contracting for services. Let us review these three bodies of work in turn to build a comprehensive model of contracting performance. The conventional or classical approach to contracting reflects the influence of principalagent theory, economic theory of competitive markets, and standard procurement practices (DeHoog, 1990; Sclar, 2000; Macneil, 1974, 1978). This approach generally treats contracts as discrete arms-length transactions between adversaries with divergent interests. The agent, or contractor, is assumed to conceal information about his capabilities and to engage in opportunistic behavior. Success in contracting, therefore, depends on a set of factors that help the government, or principal, program the contractual relationship and limit the contractor’s 5 discretion. Nearly all of the experts writing from this perspective point to a common set of variables as determinants of contracting performance (Savas, 1987, 2000; Rehfuss, 1989; Kettl, 1993; ICMA, 1989, 1992; Lavery, 1999; Seidenstat, 1999; Marlin, 1984; Wesemann, 1981). These explanatory variables, all of which are expected to be positively correlated with contracting performance, are competition (X1 and X2); ex ante evaluation of the contractor (X3); contract specificity (X4); contract monitoring scope (X5); contract monitoring intensity (X6); reliance on legal means for resolving disputes (X7); expertise in contract administration (X8); and technical knowledge of the service being contracted out (X9) (see Table 1). Before discussing these nine explanatory variables more closely, we must define our dependent variable, contracting performance (Y). Contracting performance is a multidimensional concept that captures various aspects of a provider’s performance on a contract. In this model, the dependent variable is measured using a factor score created from four-point ordinal indicators of the following eight dimensions of contractor performance: actual cost in comparison to projected cost; actual cost in comparison to in house service delivery; quality of work; responsiveness to the government’s requirements; timeliness; service continuity; compliance with the law; and customer satisfaction. Principal components factor analysis is a statistical technique that allows us to analyze and summarize the inter-relationships among a set of variables in order to reduce these variables down to a smaller number of underlying factors. Observations (or cases) are assigned a score on each factor that is derived from the analysis. This factor score represents the weighted combination of scores of different variables on that observation; the weights assigned to each variable depend on the factor loading of that variable (i.e., on the variable’s contribution to the meaning of the factor, or the variable’s correlation with the factor) (Kachigan, 1991). When 6 using factor analysis to create a new measure of a concept from multiple variables, therefore, the new measure (i.e., the factor) takes into account all of the variables that were included in the factor analysis, but it weights them based on their correlation with, or contribution to, the factor. The factor analysis of these eight indicators of contracting performance yielded a onefactor solution, with the component having an eigenvalue of 4.81 that explains 60.1% of the variance; all other components had eigenvalues of less than 1.001. Six of the eight indicators had factor loadings of .77 or higher on this component; the first two indicators measuring cost had loadings of .60 and .60, respectively. The factor scores for this component were saved as a new variable, which serves as the measure of contracting performance2. Let us now return to the first nine explanatory variables in the model. Competition is one of the most frequently cited factors relating to successful contracting (Savas, 1987, 2000; ICMA, 1992; Kettl, 1993; Greene, 2002; Hodge, 2000). Competitive bidding between providers should enhance efficiency by keeping bidders honest and compelling them to minimize their price; it should also encourage providers to deliver the best quality of service possible. Two measures of competition are included in the model: the number of bidders on the contract (X1) and whether or not public employees were allowed to bid on the contract (X2). The operational definitions for these and other explanatory variables are presented in Table 1. The conventional approach to contracting underscores the need for the agency to perform a comprehensive evaluation of the bidders’ performance capacity prior to awarding the contract, to help ensure that the most competent provider is chosen (Rehfuss, 1989; ICMA, 1989, 1992; Wasemann, 1981; Marlin, 1984; Romzek and Johnston, 2002). This evaluation should be based on a variety of criteria that capture a provider’s potential for meeting the requirements of the contract. 7 A count variable of seven evaluation criteria serves as the measure of ex ante contractor evaluation (X3). Up until recently, most of the literature on contracting for services placed a strong emphasis on contract specificity (X4). Experts suggest that contract specifications should be very precise, comprehensive, and written in language that is clearly understandable to the parties (ICMA, 1992; Savas, 2000; Rehfuss, 1989, Wasemann, 1981; Marlin, 1984). A high level of specificity should serve to make the government’s expectations clear to the contractor and helps public officials hold the contractor accountable for its performance. Responses to a ten-point ordinal indicator serve as the measure of contract specificity. The conventional approach to contracting places a premium on rigorous contract monitoring (Rehfuss, 1989; Savas, 2000; ICMA, 1989, 1992; Marlin, 1984; Wesemann, 1981; Prager, 1994; Lavery, 1999; Seidenstat, 1999; Brown and Potoski, 2003; Hefetz and Warner, 2004). Rigorous monitoring of the contractor’s behavior is supposed to improve success in contracting by indicating to public managers whether the contractor is performing according to the agreement and not in an inappropriate or opportunistic manner; rigorous monitoring also allows public managers to detect problems or fluctuations in performance that might be indicative of major problems to come. The literature suggests that effective contract monitoring is broad in scope and high in frequency or intensity. That is, public managers should monitor different aspects of performance simultaneously (e.g., inputs, processes, outputs, cost, timeliness, legal compliance), using a variety of monitoring tools and procedures at frequent intervals. Multiple ordinal indicators were factor analyzed to develop measures of contract monitoring scope (X5) and contract monitoring intensity (X6). Another explanatory variable cited in this literature is reliance on legal means for resolving disputes, particularly sanctions to rein in an unresponsive contractor found to be in violation of the agreement (ICMA, 1992; DeHoog, 1990; Macneil, 1974, 1978). Legal sanctions can include 8 financial penalties, reduced payments, litigation, and the threat of or actual termination of the contract. The conventional approach to contracting largely fails to provide internal flexibility or the possibility of planning for changes once the transaction commences; needed flexibility must come from the internal planning of each party prior to entering the transaction. Three four-point ordinal indicators were factor analyzed to develop a measure of reliance on legal means for resolving disputes (X7). Finally, having agency personnel with expertise in contract administration (X8) and with indepth knowledge of the workings of the service (X9) are frequently cited as keys to successful contracting. The public agency should have a staff trained in contract administration that is capable of preparing bid documents, evaluating bids, handling questions posed during a pre-bid conference, monitoring performance, and dealing with complaints and performance problems (ICMA, 1992; DeHoog, 1990; Rehfuss, 1989; Romzek and Johnston, 2002). In-house technical knowledge allows the agency to set its own goals (i.e., knowing what to buy), thereby preventing the erosion of accountability and the emergence of conflicts of interest that can result when private contractors are permitted to set goals (Kettl, 1993); it also makes the job of contract monitoring (i.e., knowing what has been bought) easier and more effective. A newer approach to contracting, often referred to as relational contracting, has surfaced in the literature to rival the conventional or classical approach (DeHoog, 1990; Lawther, 2003; Williamson, 1985; 1996; Sclar; 2000; Smith, 1996; Kettner and Martin, 1987; Bennett and Ferlie, 1996)3. Relational contracting represents a more flexible and cooperative approach to managing contractual relationships. In relational contracting, obligations and sanctions are diffuse, nonspecific, and non-measurable (Macneil, 1978, 1974). Consequently, the parties come to rely on extensive mutual planning, bargaining, and collaboration throughout the term of the contract in order to “fill in” the gaps in the original agreement. Mutual trust, shared norms and values, and 9 standards of behavior that evolve as the relationship unfolds serve to facilitate mutual adjustment between the parties and make the relationship more adaptable to emerging contingencies. The literature on relational contracting offers its own set of explanatory variables, all of which are hypothesized to have a positive effect on contracting performance: reliance on alternative means for resolving disputes (X10); frequency of communication between the parties (X12); joint problem solving between the parties (X13); contract duration (X15); and trust (X16) (see Table 1). While the early literature on contracting stressed strict enforcement of the original agreement and use of sanctions and threats to rein in a contractor, experts on relational contracting have begun to call for an approach to contract management that eschews confrontation and emphasizes flexibility and the use of alternative means for resolving disputes (X10), such as negotiation, arbitration, and mediation (Williamson, 1985, 1996; Sclar, 2000; DeHoog, 1990; Smith, 1996; Lawther, 2003; Kettner and Martin, 1987). Relying on legal sanctions to resolve disputes can be very expensive, as litigation imposes substantial costs in terms of money and time on both parties; a negotiated settlement often presents an efficient alternative to confrontation. Moreover, the threat or use of legal sanctions to enforce a breach of contract can heighten conflict and cause serious damage to an effective working relationship between the government and a contractor4. Factor analysis was used to develop a measure of reliance on alternative means for resolving disputes. The model also includes a variable for the inclusion of financial incentives in a contract (X11), which can encourage good performance by contractors and prevent conflicts from arising (Cooper, 2003; Behn and Kant, 1999). Frequent communication (X12) and joint problem solving between the parties (X13), throughout the contractual relationship are cited in the relational contracting literature as determinants of contracting performance (DeHoog, 1990; Sclar, 2000; Lawther, 2003: Macneil, 1978, 1974). The level of task complexity and uncertainty involved in many contracts for public 10 services preclude public managers from writing contractual requirements that are comprehensive and highly detailed. Moreover, even moderate levels of complexity and uncertainty increase the likelihood that at least some of the contract requirements will be erroneous. Successful contracting, therefore, may often require that the parties communicate and interact frequently to make necessary adjustments in work processes, performance standards, quantities and prices, as well as to fill in the gaps in the original contract document (DeHoog, 1990, Sclar, 2000; Lawther, 2003). Multiple ordinal indicators were factor analyzed to develop measures of the extent of communication (X12) and of joint problem solving efforts (X13). The model includes variables for the degree of task uncertainty that characterizes the contractor’s work (X14) and for the duration of the contract (X15). Previous research indicates a higher incidence of performance problems when contracting for “soft” services, such as public safety and human services, which typically involve more complex processes and technologies and which can be more difficult to specify and measure (Bendick, 1984; Hodge, 2000). Task uncertainty should be negatively correlated with contracting performance, therefore. Contracts of longer duration may facilitate learning and allow the parties more time to iron out the kinks in service delivery. Thus, contract duration should be positively correlated with the dependent variable. A number of experts have commented on the benefits of trust for performance in contractual relationships and partnerships (Arrow, 1974; Macneil, 1980; Williamson, 1985; Bennett and Ferlie, 1996; Sclar, 2000; Lawther, 2003; Zaheer and Venkatraman, 1995). Interorganizational trust is a construct involving the three components of dependability, predictability and faith in a relationship between two or more parties (see Young-Ybarra and Wiersema, 1999; and Zaheer, McEvily, and Perrone, 1998). How does trust affect contracting performance? The literature suggests three different causal paths. First, trust serves as a deterrent against opportunistic behavior. Trust 11 between partnering organizations seems to reduce each partners’ motivation to behave opportunistically and increases the likelihood that they will forgo short-term advantages in favor or mutual long-term gains (Aulakh, 1996; Bradach and Eccles, 1989; Jeffries and Reed, 2000; Luhmann, 1979; Macaulay, 1963). Second, trust can be a cost-effective substitute for authority and control mechanisms (Mayer, et al., 1995; Aulakh, et al., 1996; Bradach and Eccles, 1989; Hill, 1990; Zaheer, McEvily, and Perrone, 1998). If parties that trust each other are discouraged from behaving opportunistically and tend to behave in ways that conform to mutual expectations, their behavior becomes much more predictable. Predictable patterns of behavior in turn reduce transaction costs in contractual relationships by diminishing the need for highly detailed contract requirements and rigorous contract monitoring. Finally, as Young-Ybarra and Wiersema (1999) explain, “in the literature on interorganizational relationships, there has been a somewhat consistent argument that the existence of relationships based on trust between partners has a positive impact on the ability of the partners to adjust to changing environmental demands or unintended problems that may arise” (p. 443). In their study, the authors found that trust was positively and significantly correlated with the parties’ ability to modify an existing relationship (see also Doz, 1996; Williamson, 1985). Four Likert-type ordinal indicators were factor analyzed to develop a measure of trust. The third and final body of work that will be brought to bear on the question of contracting performance is the policy implementation literature, including research on the hollow state. With increasing frequency, governments in the United States are implementing policies through contractual arrangements with one or more private organizations that entail a sharing of implementation action across organizational boundaries and sectors (Romzek and Johnstone, 2002; Brown, O’Toole, and Brudney, 1998; Hall and O’Toole, 2000). Several of the variables that have been discussed so far also have received attention in the policy implementation literature. Three 12 additional policy implementation variables that appear to account for contracting performance are political support for contracting out (X17), financial resources committed to the contracting initiative (X18), and the complexity of implementation structures (i.e., the number of subcontractors involved in service delivery) (X19). The first two of these variables should be positively correlated with contracting performance, while the third variable is expected to be negatively correlated with the dependent variable. Some privatization experts have observed that the level of political support for contracting (X17) is crucial for the adoption and implementation of privatization initiatives (e.g., Sclar, 2000; Wallin, 1997). One barrier to privatization that is cited often is political opposition from public employees and their unions (Chandler and Feuille, 1991; Fernandez and Smith, 2005 forthcoming; Fernandez, Lowman, and Rainey, 2005). As Mazmanian and Sabatier (1989) and others have asserted, the likelihood of success in policy implementation is improved if officials in the implementing agency are committed to the objectives of the policy in question and are persistent when obstacles crop up or when other actors, including members of the target group, resist change. Implementation scholars have underscored the importance of providing adequate financial resources to public organizations involved in policy implementation (Mazmanian and Sabatier, 1989; Provan and Milward, 1995; Montjoy and O’Toole, 1979). Ample funding is necessary to staff implementing agencies and provide them with the administrative and technical capacity to carry out their mission. In regards to contracting, adequate financial resources (X18) can help to enhance performance by enabling the agency to effectively administer the solicitation process, monitor performance, and manage the ongoing relationship; adequate funding of the contracting initiative also is needed to protect the provider from financial stress (Romzek and Johnston, 2002). Multiple ordinal indicators were factor analyzed to develop measures of political support for contracting out and financial resources committed to the contractual relationship. 13 Lastly, the policy implementation literature has demonstrated that the structure and complexity of interactions among actors can have a significant effect on the outcomes of implementation (Pressman and Wildavsky, 1984; Van Meter and Van Horn, 1975; Agranoff and McGuire, 1998). In regards to contracting performance, the lesson appears to be that complex service delivery arrangements involving multiple subcontractors imposes additional burdens on the prime contractor, including higher coordination costs, the likelihood of delays, and sometimes even conflict over the choice of goals and means, all of which ultimately weaken performance. An ordinal indicator is used to measure the number of subcontractors involved in the contractual relationship. Methodology and Data The model presented above will be tested using ordinal least squares (OLS) regression and a large dataset of over 450 local government contracts from across the U.S. As the previous review of the literature indicated, very few studies of contracting performance have been explanatory in nature or have employed a large number of observations to test theoretical propositions. Even the three previous attempts to test comprehensive explanatory models of contracting effectiveness have been hampered by small sample sizes (Romzek and Johnston, 2002; 2005; Domberger and Hensher, 1993). A multivariate analysis like this one using such a large dataset offers several advantages over previous studies, including statistical control, more precise estimates of causal effect, and greater external validity. The choice of level of analysis is crucial for explaining contracting performance. Most of the factors that are hypothesized to influence contracting performance are at the contract level of analysis. Thus, the unit of analysis chosen for this study is the individual contractual relationship between a local government and a private contractor. Adopting a higher level of analysis (e.g., the 14 local government level) when trying to explain why some contractual relationships fail while others succeed would be akin to committing the ecological fallacy. The data for this study were gathered through a mail survey of local governments throughout the country. Obtaining a good sampling frame is a serious obstacle to doing a survey of local governments when the contractual relationship is the unit of analysis. The sampling frame used for this survey was the set of responses to the 2002-2003 International City/County Management Association’s alternative service delivery survey, which identifies contracts in sixtyseven different service areas5. A sample of 982 local government contracts (with for-profit and non-profit providers) was drawn randomly from the 2002-2003 ICMA data. The sample was stratified by the seven broad categories of services used by ICMA to identify these services and functions. The survey was addressed to the chief administrative office of the local government, asking him or her to answer a series of questions about a specific contract with a private for-profit or not-for-profit provider. The response rate for the survey was forty eight percent, which compares favorably to other recent large-n privatization studies (Brown and Potoski, 2003; Brudney, et al., 2005). Two approaches were taken to gauge the generalizability of the survey data. First, respondents and non-respondents were compared. The two groups are remarkably similar in terms of type of local government, type of jurisdiction, population size, geographic region, metropolitan status, and form of government; thus offering no indication of non-response error (Groves, 1989; Keeter, et al., 2000)6. Second, a comparison was made between the contracts reported in the survey and all of the contracts in the ICMA sampling frame by the seven broad service categories that were used to stratify the survey sample. The results show that while contracts for public works/transportation are somewhat overrepresented in the survey sample and contracts for cultural and arts programs are underrepresented in the survey sample, overall, the contracts in the sample of 15 survey respondents are quite comparable to the contracts in the sampling frame, which is the best estimate of the population of contracts at the local level in the United States. This fact bodes well for the external validity of the study. Since the data used in this study is in part attitudinal, the model includes control variables for four respondent characteristics (years in government, political ideology, level of education, and overall opinion about privatization), which may create spurious relationships and bias the findings. Finally, it should be noted that some of the literature on contracting suggests interactive effects involving the variables task complexity and competition (see DeHoog, 1990 and Williamson, 1985). Several specifications that included interactive terms were tested using moderated multiple regression, but the results failed to produce a better fit. OLS regression is used, therefore, to test an additive version of the model. Analysis and Results The results of the OLS regression are shown in Table 2. The R-square is 0.503, indicating that model accounts for half of the variance in contracting performance (F test value of 18.218, p < 0.001). Two diagnostic tests revealed no multicollinearity in the model7. To examine the regression errors, the model’s standardized residuals were plotted against the predicted values. The plot reveals no extreme outliers; all of the standardized residuals are within three standard deviations of zero. The model also seems to be robust and not sensitive to leverage points8 None of the four control variables for respondent characteristics achieved statistical significance in the model. Since the data is in part attitudinal, this finding is encouraging and should help to allay fears about biased responses and spurious relationships influencing the regression results. Let us begin by examining the results for the first nine independent variables in the model that were derived from the classical or conventional approach to contracting. All nine of these 16 variables were hypothesized to have a positive effect on contracting performance. As noted in the literature review, competition is perhaps the most frequently cited determinant of success in contracting for services. Thus, we expect the two measures of competition to be positively correlated with the dependent variable. As we see, however, neither the number of bidders on a contract (X1) or allowing public employees to bid on the contract (X2) is statistically significant at the p < 0.05 level. Given the somewhat surprising nature of this finding, an additional measure of competition—on-going competition among multiple contractors—was included in another specification of the model, but this variable also failed to achieve statistical significance. In short, the failure of three measures of competition to achieve statistical significance indicates that competition has little if any effect on contracting performance as measured in this analysis. One plausible explanation for the absence of a competition effect in the model is that fostering competition can be costly for government, as Donahue (1989) and Sclar (2002) observed. A second possibility is that stability in contractual relationships contributes much more to success than competition, as some students of contracting with the non-profit sector have suggested (Smith and Lipsky, 1993; Smith, 1996; Smith and Smyth, 1996; Provan and Milward, 1995)9. It was hypothesized that the variables ex ante evaluation (X3) and contract specificity (X4) would be positively correlated with the dependent variable. Neither coefficient achieves statistical significance at the p < 0.05 level. It appears, therefore, that undertaking a comprehensive evaluation of the bidders’ performance capacity prior to awarding the contract and developing highly detailed and specific contractual requirements have little if any effect on contracting performance. The prevailing view in much of the literature on contracting is that rigorous monitoring is a key determinant of success. The model includes two measures of contract monitoring, contract monitoring scope (X5) and contract monitoring intensity (X6). We find that the coefficient for 17 contracting monitoring scope (X5) fails to achieve statistical significance at the p < 0.05 level. Collecting a wide range of performance data has no effect on contracting performance. The coefficient for contracting monitoring intensity (X6) also fails to achieve statistical significance at the p < 0.05 level. Monitoring the contract using a variety of monitoring tools and procedures at frequent intervals seemingly has no influence on contracting performance. To see if any of the monitoring tools and approaches is correlated individually with the dependent variable, OLS regression was run again with the monitoring intensity variable (X6) disaggregated into the six original indicators. None of the six coefficients were statistically significant even at the p < 0.10 level10. In short, these findings suggest that the discovery of performance problems through monitoring by itself does little to improve contracting performance. It is possible that contracting performance is improved when the parties act on such information, by working together to find a solution to a performance problem and to implement the solution. As discussed below, other findings lend support to this proposition. The conventional approach to contracting stresses the need for principals to use legal sanctions in order to discourage poor performance and rein in a contractor who has breached the agreement. The results fail to support this proposition. The coefficient for reliance on legal means for resolving disputes (X7) fails to achieve statistical significance at the p < 0.05 level. Two variables were included in the model to test Kettl’s (1993) “smart buyer” hypothesis. The results provide partial support for this hypothesis. While the coefficient for expertise in contract administration (X8) fails to achieve statistical significance at the p < 0.05 level, the coefficient for technical knowledge of the service (X9) is statistically significant and in the anticipated direction. Having public managers with knowledge of the technical complexities of service delivery results in higher levels of contracting performance. In-depth knowledge of the workings of a service helps public managers to set sensible goals for the contract, to independently 18 judge the contractor’s performance without relying on the expertise of consultants or other contractors, and to propose the correct modifications to a contract as predicated by changes in technology and in the needs of service recipients. To summarize, nine explanatory variables were derived from the classical or conventional approach to contracting. Only the variable technical knowledge of the service (X9) is statistically significant and has the expected positive effect on contracting performance. The other eight variables fail to achieve statistical significance and appear to have no impact on the dependent variable. Let us now turn to the explanatory variables derived from the literature on relational contracting. The variables X10 through X16 are expected to have a positive effect on contracting performance, except for X14 (task uncertainty), which is hypothesized to have a negative impact on performance. The coefficient for reliance on alternative means for resolving disputes (X10) is positive and statistically significant at the p < 0.05 level, as the literature on relational contracting suggests. Negotiation, arbitration, and mediation appear to be the more effective means for resolving disputes, perhaps because of their lower cost compared to litigation, and also because they minimize conflict between the parties and preserve an effective working relationship (see Macaulay, 1963; Williamson, 1985). Conversely, including financial incentives (X11) in the contractual agreement appears to have no effect on contracting performance, as this variable’s coefficient fails to achieve statistical significance at the p < 0.05 level. The frequency of communication between the parties (X12) does not have the anticipated positive effect on contracting performance. Although the coefficient is in the expected direction, it fails to achieve statistical significance at the p < 0.05 level. The model includes a variable for efforts by the parties to work together to identify and solve problems during the contractual relationship. The coefficient for joint problem solving after contract award (X13) is positive and 19 statistically significant at the p < 0.01 level in the model. As public managers work more closely with the contractor’s staff to solve performance issues, the level of contracting performance tends to increase, as experts on relational contracting have stipulated. This finding supports the notion expressed above that contracting performance is improved not by the mere monitoring of performance but by deliberate action to understand and solve performance problems. The coefficient for task uncertainty (X14) is negative (-0.098) and statistically significant at the p < 0.01 level. This is a sensible finding, insofar as we would expect uncertainty about the best means for accomplishing the work, difficulty measuring outcomes, and shifting technical requirements to complicate the work of the contractor and diminish the chances of the government getting the outcomes it desires (see Bendick, 1984; Hodge, 2000). Contracts of longer duration should allow the parties more time for learning and for resolving problems that often arise as the contractor takes over responsibility for service delivery. The results do not support this proposition, however. The coefficient for contract duration (X15) is in the expected direction but it fails to achieve statistical significance at the p < 0.05 level. The most powerful predictor of contracting performance in the model is the level of trust between the parties (X16) (beta of 0.452), a factor that was hardly mentioned in the early literature on contracting but one that is cited with much frequency in research on relational contracting and interorganizational collaboration. The coefficient for trust is positive and statistically significant at the p < 0.001 level11. As previous research on trust indicated, the presence of trust in a contractual relationship seems to reduce the incidence of shirking and other forms of opportunistic behavior that weaken performance. Trust also may be contributing to contracting performance by facilitating the parties’ efforts to respond in a flexible manner to unforeseen contingencies and other disturbances in the relationship12. 20 Some of the literature on trust suggests that trust and performance have a reciprocal causal relationship: a higher level of trust between the parties results in better performance, and better performance in turn raises the level of trust. A reciprocal causal relationship in an OLS model would pose an endogeneity problem, which causes OLS regression coefficients to become inconsistent, making it more difficult to estimate causal impact. In this analysis, we are concerned that endogeneity is biasing upwards the coefficient for trust, making its effect seem larger than it is and biasing the other coefficients in the model. The Durbin-Wu-Hausman test (Davidson and MacKinnon, 1993; Wooldridge, 2002) was performed to detect endogeneity and assess the need for developing an instrumental variable for trust (X16). The results failed to detect an endogenous relationship between trust and the dependent variable, thus allowing us to infer with greater certainty the causal impact of trust and of the other independent and control variables in the model13. Theoretically, however, a pattern of reciprocal causation could be at play in some form. To summarize, four of the seven explanatory variables derived from the relational contracting literature have the expected effects on contracting performance. Reliance on alternative means for resolving disputes (X10), joint problem solving efforts after contract award (X13), and trust between the parties (X16) are statistically significant and have positive effects on contracting performance, while task uncertainty (X14) has a negative effect on contracting performance. Conversely, financial incentives (X11), the frequency of communication (X12), and contract duration (X15) fail to achieve statistical significance and appear have no effect on contracting performance. Finally, as suggested by policy implementation researchers, we expected the variables political support for contracting out among public employees (X17) and financial resources dedicated to the contracting initiative (X18) to be positively correlated with contracting performance and for the number of subcontractors involved in the contractual relationship (X19) to be negatively correlated with contracting performance. The results support all three propositions. The 21 coefficients for political support (X17) and financial resources (X18) are positive and statistically significant at the p < 0.05 level and they have the second and third highest standardized coefficients in the model (0.242 and 0.139, respectively). Mid- and low-level managers and street–level employees who are supportive of the contractual relationship appear less likely to fear privatization and more inclined to work in partnership with the contractor’s staff to manage problems that arise during the contractual relationship. Adequate funding to support various managerial activities and functions—including initial planning efforts, the subsequent solicitation process, and the various ongoing activities involved in managing the contractual relationship—also contributes to successful contracting. A significant commitment of financial resources is also likely to send a signal to the contractor that the government is committed to the success of the contractual relationship. The coefficient for the number of subcontractors involved in the contractual relationship (X19) is negative and statistically significant at the p < 0.05 level. The need to coordinate the actions of multiple service providers may increase the cost of the contract and create delays. Moreover, the presence of multiple clearance points (i.e., subcontractors) could be weakening the government’s control over the service delivery process and undercutting its ability to influence the outcomes of the contractual relationship. Discussion and Conclusion This study set out to identify the determinants of contracting performance. Two general conclusions emerge from the analysis. First, the conventional wisdom about how to effectively manage the contracting process seems to miss the mark, as most of the propositions derived from this approach are rejected by the findings. Robust competition, tight contract specifications, rigorous contract monitoring, and the use of legal means to enforce the contract are factors that are supposed to work together to program the contractual relationship and to exert control over the contractor. The findings suggest, however, that successful contractual relationships are not discrete, 22 arms-length transactions that are programmed and controlled but rather dynamic relationships that are managed in a flexible and cooperative manner, with the parties acting more like equals than principal and agent. Contractual relationships often are messy affairs susceptible to a variety of setbacks and disturbances that can frustrate efforts to reach mutually agreed-upon objectives. According to Landau and Stout (1979), “solutions to problems cannot be commanded. They must be discovered, found on the basis of imagination, analysis, experiment, and criticism” (p. 152). As they note, efforts to control a situation that calls for flexibility and experimentation are at best ineffective and at worst counter productive. This may explain in part why key elements of the conventional approach to contracting appear to have no impact on contracting performance. Instead, the findings show that some of the strongest determinants of contracting performance are factors that facilitate adaptive decision making, problem solving and learning, among them trust, a willingness to work together to identify and solve problems, and reliance on negotiations and other alternative means for resolving disputes14. A second and related conclusion that emerges from the findings is that ex post factors that come into play after the contract is awarded have as much of a bearing on the outcomes of a contractual relationship as ex ante factors, which set the groundwork for the relationship. Public managers can take various ex ante steps to enhance their chances for success before a contract is awarded, including securing the support of public employees, adequately funding the contracting initiative, ensuring that the agency has in-house knowledge of the workings of service delivery, and limiting the number of subcontractors involved in the contractual relationship. These ex ante factors alone do not determine the outcomes of the relationship, however. A number of other issues that arise during the life of the contract also must be managed effectively to ensure a high level of contracting performance. Parties that work together to identify and solve problems in service delivery, that settle disputes in a constructive and amiable manner so that conflict does not escalate, 23 and that take steps to build mutual trust achieve higher levels of contracting performance; maintaining adequate levels of funding and sustaining political support from employees throughout the life of a contract also contribute to performance. Several potential methodological problems were addressed above and do not appear to be influencing the results, including endogeneity involving the variable trust and non-response bias. Also, although the data used for the analysis is attitudinal, the findings are strengthened by the fact that the four control variables for respondent characteristics failed to achieve statistical significance. The potential effects of social desirability should be raised here15. One consequence of social desirability is to raise the bar on tests of statistical significance by reducing variability in the values of the independent variables in the model. Two of the variables that fail to achieve statistical significance but that are somewhat close to doing so—contract specificity and expertise in contract administration—have very skewed frequencies that seem to indicate some social desirability effect. Since the coefficients for these two variables were in the expected direction, the influence of social desirability may explain in part why these variables are not statistically significant. On the whole, however, the data for the independent and control variables in the model exhibit sufficient variability to allay fears of social desirability bias affecting the overall results of this study16. The results of this study should be fairly generalizable to most instances of contracting out by local governments across the United States, especially among medium and large local governments. The response rate for the survey was nearly fifty percent, respondents and nonrespondents were highly comparable on a number of theoretically important variables, and the data capture contracts for over sixty different services. Whether or not these findings are generalizeable beyond local government contracting to contracting out by state and federal agencies is a question that is much more difficult to answer. The findings may be generalizeable to state agency contracting for similar types of services on a similar scale (e.g., public works, transportation, health 24 and human services, support services). However, at the federal level, the procurement process is more elaborate and formalized, with many policies and regulations governing how public managers undertake the make-or-buy decision, sourcing, and contract management. Public managers operating in such a constraining environment may find it more difficult to bargain, to make quick and efficient adjustments to the contractual relationship, and to take steps to gain the trust of the other party. For these reasons, a more relational or collaborative approach to managing contractual relationships could be less effective at the federal level, and more research on contracting performance at the federal level is clearly needed. The significance of trust and cooperation as determinants of contracting performance raises some concerns about loss of accountability in contracting for services. Various experts suggest that external mechanisms of accountability, such as oversight and regulation, are needed to hold private contractors accountable for their behavior (Kamarck, 2002; Bardach and Lesser, 1996; Smith and Smyth, 1996)17. On one level, we might be concerned about whether too much trust weakens the ability of public managers to hold contractors accountable for their behavior and performance. Little if any empirical evidence from this analysis or from the relational contracting literature suggests that this is the case. The data indicates that public managers do not necessarily engage in less contract monitoring when they have high levels of trust in the contractor18. In addition, trust is positively correlated with all of the eight indicators used to measure contacting performance. Thus, trust not only contributes to greater efficiency, it also has a positive effect on respondents’ ratings of their contractor’s responsiveness to the government’s requirements, on compliance with the law, and on customer satisfaction, three key forms of contractor accountability (Cooper, 2003). Finally, previous work on the role of trust in economic exchanges indicates that like authority, trust acts as a mechanism for holding a party accountable for its behavior (Bradach and Eccles, 1989; see also Macaulay, 1963; Aulakh, Kotabe, and Sahay, 1996; Luhmann, 1979). 25 On another level, we might be concerned with whether public managers who trust their contractors and develop close personal relationships with them become unresponsive to political institutions, the public, and the law. One could reasonably argue that public managers who repeatedly interact with a contractor’s staff might develop a close bond with the staff, and as a result, would be inclined to give them preferential treatment. It is also possible that public managers involved in relational contracting could come to view the continuation of the contractual relationship as more important than the legal requirement to rebid the contract or to terminate it for a breach of agreement. These are empirical questions that have not been explored in a systematic fashion by researchers but are certainty worthy of further investigation. The survey data used for this study does not lend itself to answering these kinds of questions about a public contract manager’s accountability to political institutions, the public, and the law, since all of the data are survey responses from public managers. Answering such questions would require asking other actors in the political system, such has elected officials or service recipients, to rate the level of responsiveness of public managers involved in relational contracting and to provide their own evaluation of the contractor’s performance, so that their evaluation can then be compared to the public managers’ judgment of performance. Future research is needed to fill this gap in our understanding of accountability in the area of contracting for services. 26 References Agranoff, R. and McGuire, M. 1998. “Multinetwork Management: Collaboration and the Hollow State in Local Economic Policy.” Journal of Public Administration Research and Theory, 8: 67-91. American Federation of State, County, and Municipal Employees. 2004. Privatization—The Public Pays 2004. Washington, D. C. Arrow, K. 1974. The Limits of Organization. New York: Norton. Aulakh, P S., Kotabe, M. and Sahay, A. 1996. “Trust and Performance in Cross-Border Marketing Partnerships: A Behavioral Approach.” Journal of International Business Studies, special issue: 1005-1032. Bardach, E. and Lesser, C. 1996. “Accountability in Human Service Collaboratives—For What? And to Whom?” Journal of Public Administration Research and Theory, 6: 197-212. Behn, R. D. and Kant, P. A. 1999. “Strategies for Avoiding the Pitfalls of Performance Contracting.” Public Productivity and Management Review, 22: 470-89. Bendick, M., Jr. 1984. “Privatization of Public Services: Recent Experience.” In Brooks, H. and Lieman, L., and Schelling, C. S. (Eds.) Public-Private Partnership: New Opportunities for Meeting Social Needs. Cambridge, MA: Ballinger. Bennett, C. and Ferlie, E. 1996. “Contracting in Theory and in Practice: Some Evidence from the NHS.” Public Administration, 74: 49-66. Boyne, G. A. 1998. “The Determinants of Variations in Local Service Contracting: Garbage In, Garbage Out?” Urban Affairs Review 34, 149-162. Brown, M. M., O’Toole, L. J., Jr., and J. L. Brudney. 1998. “Implementing Information Technology in Government: An Empirical Assessment of the Role of Local Partnerships.” Journal of Public Administration Research and Theory, 4: 499-525. Brown, T. L. and Potoski, M. 2002. “Managing Contract Performance: A Transaction Costs Approach.” Journal of Policy Analysis and Management, 22: 275-297. Brown, T. L. and Potoski, M. 2003. "Contract–Management Capacity in Municipal and County Governments." Public Administration Review, 63: 153-164. Brudney, J. L., Fernandez, S., Ryu, J. E., and Wright, D. S. 2005. “Exploring and Explaining Contracting Out: Patterns Among the American States.” Journal of Public Administration Research and Theory, 15: 393-419. Chandler, T. and Feuille, P. 1991. “Municipal Unions and Privatization.” Public Administration Review, 51: 15-22. Cohen, S. 2001. “A Strategic Framework for Devolving Responsibility and Functions From Government to the Private Sector.” Public Administration Review, 61: 432-440. Cooper, P. J. 2003. Governing by Contract. Washington, DC: CQ Press. Davidson, R. and MacKinnon, J. G. 1993. Estimation and Inference in Econometrics. New York: Oxford University Press. DeHoog, R. H. 1990. “Competition, Negotiation, or Cooperation: Three Models for Service Contracting.” Administration and Society, 22: 317-40. Dilger, R. J., Moffett, R. R., and Struyk, L. 1997. “Privatization of Municipal Services in America’s Largest Cities.” Public Administration Review, 57: 21-26. Domberger, S. and Hensher, D. 1993. “On the Performance of Competitively Tendered, Public Sector Cleaning Contracts.” Public Administration, 71: 441-454. 27 Donahue, J. D. 2002. “The Problem of Public Jobs.” In J. Nye and Donahue, D. (Eds.). Marketbased Governance: Supply Side, Demand Side, Upside, and Downside. Washington, DC: Brookings Institution Press. Donahue, J. D. 1989. The Privatization Decision: Public Ends, Private Means. BasicBooks. Doz, Y. L. 1996. “The Evolution of Cooperation is Strategic Alliances: Initial Conditions or Learning Processes?” Strategic Management Journal, 17: 55-83. Fernandez, S. Lowman, C. E., and Rainey, H. G. 2005. “Privatization and Human Resources Management.” In N. N. Riccucci (Ed.). Public Personnel Management: Current Concerns, Future Challenges. Fourth Edition. New York: Longman. Fernandez, S. and Smith, C. 2005, forthcoming. “Looking for Evidence of Public Employee Opposition to Privatization: An Empirical Study with Implications for Practice.” Review of Public Personnel Administration. Greene, J. D. 2002. Cities and Privatization: Prospects for the New Century. Upper Saddle River, New Jersey: Prentice Hall. Groves, R., M. 1989. Survey Error and Survey Cost. Wiley, John & Sons. Hall, T. E. and L. J. O’Toole, Jr. 2000. “Structures for Policy Implementation: An Analysis of National Legislation, 1965-1966 and 1993-1994.” Administration and Society, 31: 67-86. Hefetz, A. and Warner, M. 2004. "Privatization and its Reverse: Explaining the Dynamics of the Government Contracting Process." Journal of Public Administration Research and Theory. 14: 171-190. Hill, C. W. 1990. “Cooperation, Opportunism, and the Invisible Hand: Implications for Transaction Cost Theory.” Academy of Management Review, 15: 500-513. Hodge, G. A. 2000. Privatization: An International Review of Performance. Boulder, CO: Westview Press. International City/County Management Association (ICMA). 1989. Service Delivery in the 90s: Alternative Approaches for Local Governments. Washington, DC: ICMA. International City/County Management Association (ICMA). 1992. Service Contracting: A Local Government Guide. Washington, DC: ICMA. Jeffries, F. L. and Reed, R. 2000. “Trust and Adaptation in Relational Contracting.” Academy of Management Review, 25: 873-882. Kachigan, S. K. 1991. Multivariate Statistical Analysis: A Conceptual Introduction. Second Edition. New York: Radius Press. Kamarck, E. C. 2002. “The End of Government as We Know It.” In Donahue, J. D. and Nye, J. S., Jr. Market-Based Governance: Supply Side, Demand Side, Upside, and Downside. Cambridge, MA: Visions of Governance in the 21st Century. Kamerman, S. B. and Kahn, A. J. (Eds.) 1989. Privatization and the Welfare State. Princeton: Princeton University Press. Kelman, S. J. 2002. “Contracting Out.” In L. M. Salamon. (Ed.). The Tools of Government: A Guide to the New Governance. New York: Oxford University Press. Keeter, S., Miller, C., Kohut, A., Groves, R. M., and Presser, S. 2000. “Consequences of Reducing Nonresponse in a National Survey.” Public Opinion Quarterly, 64: 125-148. Kettl, D. F. 2000. The Global Public Management Revolution: A Report on the Transformation of Governance. Washington, DC: Brookings Institution. Kettl, D. F. 1993. Sharing Power: Public Governance and Private Markets. Washington, D. C.: The Brookings Institution. Kettner, P. M. and Martin, L. L. 1990. “Purchase of Service Contracting: Two Models.” Administration in Social Work, 14: 15-31. 28 Landau, M. and Stout, R., Jr. 1979. “To Manage is Not to Control: Or the Folly of Type II Error.” Public Administration Review, 39: 148-56. Lavery, K. 1999. Smart Contracting for Local Government Services: Processes and Experience. Westport: Praeger. Lawther, W. C. 2003. "Privatizing Personnel: Outsourcing Public Sector Functions." In Public Personnel Administration: Problems and Prospects. Kearney, H. (Ed.) Upper Saddle River, NJ: Prentice Hall. Light, P. C. 1999. The True Size of Government. Washington, DC: Brookings Institution Press. Luhmann, N. 1979. Trust and Power. Chichester, UK: Wiley. Macneil, I. R. 1974. “The Many Futures of Contracts.” Southern California Law Review, 47: 691816. Macneil, I. R. 1978. “Contracts: Adjustments of Long-term Economic Relations under Classical, Neoclassical and Relational Contract Law.” Northwestern University Law Review, 72: 855905. Macneil, I. R. 1980. The New Social Contract. New Haven, CT: Yale University Press. Macaulay, S. 1963. “Non-contractual Relations in Business.” American Sociological Review, 28: 55-70. Marlin, J. T. (Ed.). 1984. Contracting Municipal Services: A Guide for Purchase from the Private Sector. New York: Wiley. Martin, L. L. 1999. Contracting for Service Delivery: Local Government Choices. Washington, DC: International City/County Management Association. Mayer, R. C., Davis, J. H., and Schoorman, F. D. 1995. “An Integrative Model of Organizational Trust.” Academy of Management Review, 20: 709-734. Mazmanian, D. A. and Sabatier, P. A. 1989. Implementation and Public Policy: With a New Postcript. Latham, MD: University Press of America. Miranda, R. and Andersen, K. 1994. “Alternative Service Delivery in Local Government, 19821992.” In International City/County Management Association (ICMA). The Municipal Year Book, 1994. Washington DC: ICMA. Montjoy, R. S. and O’Toole, L. J., Jr. 1979. “Towards a Theory of Policy Implementation: An Organizational Perspective.” Public Administration Review, 39: 456-75. National Commission of Employment Policy. 1988. Privatization and Public Employees: The Impact of City and County Contracting Out on Government Workers. Dedek and Company for National Commission of Employment Policy. Osborne, D. and Gaebler, T. 1992. Reinventing Government: How the Entrepreneurial Spirit is Transforming the Public Sector. Reading, Mass: Addison-Wesley. Prager, J. 1994. “Contracting Out Government Services: Lessons from the Private Sector.” Public Administration Review, 54: 176-184. Pressman, J. and Wildavsky, A. 1984. Implementation. Third Edition. Berkeley: University of California Press. Provan, K. G. and Milward, H. B. 1995. “A Preliminary Theory of Interorganizational Network Effectiveness: a Comparative Study of Four Community Mental Health Systems.” Administrative Science Quarterly, 40: 1-33. Rehfuss, J. A. 1989. Contracting Out in Government. San Francisco: Jossey-Bass. Romzek, B. S. and Johnston, J. M. 2005. “State Social Services Contracting: Exploring the Determinants of Effective Contract Accountability.” Public Administration Review, 65: 436-449. Romzek, B. S. and Johnston, J. M. 2002. “Effective Contract Implementation and Management: A Preliminary Model.” Journal of Public Administration Research and Theory, 12: 423-453. 29 Savas, E. S. 1987. Privatization: The Key to Better Government. Chatham, N.J.: Chatham House. Savas, E. S. 2000. Privatization and Public-Private Partnerships. New York: Chatham House. Sclar, E. D. 2000. You Don't Always Get What You Pay For: The Economics of Privatization. Ithaca: Cornell University Press. Seidenstat, P. (Ed.). 1999. Contracting Out Government Services. Westport: Praeger. Siegel, G. B. 1999. “Where Are We on Local Government Service Contracting?” Public Productivity and Management Review, 22: 365-388. Smith, S. R. 1996. “Transforming Public Services: Contracting for Social and Health Services in the US.” Public Administration, 74: 113-27. Smith, S. R. and Lipsky, M. 1993. Nonprofits for Hire: The Welfare State in the Age of Contracting. Cambridge, MA: Harvard University Press. Smith, S. R. and Smyth, J. 1996. “Contracting for Services in a Decentralized System.” Journal of Public Administration Research and Theory, 6: 277-296. Van Meter, D. S. and Van Horn, C. E. 1975. “The Policy Implementation Process: A Conceptual Framework.” Administration and Society, 6: 445-488. Wallin, B. A. 1997. “The Need for a Privatization Process: Lessons From Development and Implementation.” Public Administration Review, 57: 11-20. Warner, M. and Hebdon, R. 2001. “Local Government Restructuring: Privatization and its Alternatives.” Journal of Policy Analysis and Management 20, 315-336. Wesemann, H. E. 1981. Contracting for City Services. Pittsburgh: Innovations Press. Williamson, O. E. 1985. The Economic Institutions of Capitalism. New York: Free Press. Williamson, O. E. 1996. The Mechanisms of Governance. New York: Oxford University Press. Wooldridge, J. M. 2002. Introductory Econometrics: A Modern Approach. Second Edition. Mason, OH: Thompson South-Western. Young-Yabarra, C. and Wiersema, M. 1999. “Strategic Flexibility in Information Technology Alliances: The Influence of Transaction Cost Economics and Social Exchange Theory.” Organization Science, 10: 439-459. Zaheer, A. and Venkatraman, N. 1995. “Relational Governance as an Interorganizational Strategy: An Empirical Test of the Role of Trust in Economic Exchange.” Strategic Management Journal, 16: 373-392. Zaheer, A., McEvily, B., and Perrone, V. 1998. “Does Trust Matter? Exploring Effects of Interorganizational and Interpersonal Trust on Performance.” Organization Science, 9: 141159. 30 Table 1: Explanatory Variables and Measures Variable Number of bidders (X1) Public-private competition (X2) Ex ante evaluation (X3) Contract specificity (X4) Contract monitoring scope (X5) Contract monitoring intensity (X6) Reliance on legal means for resolving disputes (X7) Expertise in contract administration (X8) Technical knowledge of the service (X9) Reliance on alternative means for resolving disputes (X10) Financial incentives (X11) Frequency of Measure(s) A 5-point ordinal indicator for the approximate number of providers who submitted bids or proposals for the contract. Response to the question “Where public employees allowed to bid on the contract?” The total number of the following seven factors that were taken into consideration when evaluating the contractor’s capacity to perform the work prior to awarding the contract: the provider’s financial capacity or financial health; technical capacity; staffing capacity; cost for service delivery; reputation; total previous experience performing the work; and previous performance on other contracts with city/county. Response to the question “On a scale from 1 to 10, with 1 being vague and 10 being very specific, how specific was the language used to write the contract’s scope of work (or technical specifications)?” The total number of the following eight types of contractor performance data collected by the local government: work inputs; work processes; work outputs; timeliness; cost; accuracy of invoicing; legal compliance; and complaints. Factor score created from responses to how frequently the local government used each of the following six monitoring tools or procedures to assess the contractor’s performance: inspections of work in progress; inspections of work completed; complaints monitoring; examining contractor reports; performance measurement systems; and citizen surveys (6-point ordinal scale). One factor solution, eigenvalue of 2.73, 45.5% of the variance explained. Factor score created from responses to how frequently the local government relied on following three means for resolving contract disputes: financial penalties; the threat of contract termination; and litigation to resolve disputes (4-point ordinal scale). One factor solution, eigenvalue of 1.72, 56.7% of the variance explained. Response to the question “How would you rate the level of expertise in contract administration among local government employees managing the contract?” (5-point ordinal scale). Response to the question “How much technical knowledge about the service do local government employees have?” (4-point ordinal scale). Factor score created from responses to how frequently the local government relied on following three means for resolving contract disputes: negotiations; arbitration; mediation (4-point ordinal scale). One factor solution, eigenvalue of 1.80, 60.07% of the variance explained. The total number of the following three types of financial incentives that were included as part of the contractual agreement: gain sharing, contract renewal based on good performance; and bonus for reaching certain goals. Factor score created from responses to how frequently the local government 31 communication (X12) Joint problem solving (X13) Task uncertainty (X14) Contract duration (X15) Trust (X16) used each of the following four channels of communication to share information with the contractor’s staff: scheduled face-to-face meetings; informal face-to-face conversations; telephone conversations; and written communications (6-point ordinal scale). One factor solution, eigenvalue of 2.48, 61.8% of the variance explained. Factor score created from responses to the following two Likert type questions: “We always work together with the contractor to identify problems” and “We always work together with the contractor to solve problems” (5-point ordinal scale). One factor solution, eigenvalue of 1.81, 90.3% of the variance explained. Factor score created from the responses to the following five Likert type questions: “The contract has numerous desired outcomes”; “When the contractor achieves one desired outcome, it conflicts with other desired outcomes”; “There is more than one method for achieving each desired outcome”; “It is difficult to measure or evaluate the quality of the work performed by the contractor”: and “The work performed by the contractor is always affected by unforeseen technical changes or developments” (f-point ordinal scale). Two factor solution, with the first factor having an eigenvalue of 1.68 (33.5% of the variance explained), and the second factor having an eigenvalue of 1.11 (22.4% of the variance explained). The first factor is used as the measure of task uncertainty (X14). The number of months that the contract has been in operation. Factor score created from responses to the following four Likert type questions: “When we encounter difficult and new circumstances, we do not feel worried or threatened by letting the contractor do what it wants”; “We are familiar with the patterns of behavior the contractor has established, and we can rely on the contractor to behave in certain ways”; “We have found that the contractor is always dependable”; and “The contractor can never be trusted to act properly” (5-point ordinal scale). One factor solution, eigenvalue of 1.95, 48.8% of the variance explained. Political support Factor score created from responses to how supportive top management, for contracting out lower-level management, and frontline (street-level workers) are of the (X17) contracting initiative (4-point ordinal scale). One factor solution, eigenvalue of 2.08, 69.5% of the variance explained. Financial resources Factor score created from responses to how adequate is the amount of (X18) funding your local government has allocated for administering the contracting process, monitoring the contract, and paying or reimbursing the contractor (4-point ordinal scale). One factor solution, eigenvalue of 2.10, 70.0% of the variance explained. Subcontractors Response to the question ‘Approximately how many subcontractors does (X19) the contractor use to deliver the service?” (5-point ordinal scale). 32 Table 2. Regression Results Number of bidders (X1) Public-private competition (X2) Ex ante evaluation (X3) Contract specificity (X4) Contract monitoring scope (X5) Contract monitoring intensity (X6) Reliance on legal means for resolving disputes (X7) Expertise in contract administration (X8) Technical knowledge of service (X9) Reliance on alternative means for resolving disputes (X10) Financial incentives (X11) Frequency of communication (X12) Joint problem solving after contract award (X13) Task uncertainty (X14) Contract duration (X15) Trust between the parties (X16) Political support for contracting out (X17) Financial resources (X18) Subcontractors (X19) R-square = 0.503 * p < 0.05 Adj. R-square = 0.475 ** p < 0.01 Coefficient -0.035 -0.086 0.033 0.024 -0.006 -0.001 0.005 0.060 0.111* Std. Error 0.039 0.095 0.023 0.018 0.021 0.047 0.040 0.045 0.060 Beta -0.034 -0.032 0.058 0.054 -0.012 -0.001 0.005 0.053 0.073 0.076* -0.072 0.047 0.114** -0.108** 0.001 0.452*** 0.242*** 0.139*** -0.092* 0.039 0.064 0.044 0.039 -0.037 0.001 0.041 0.040 0.037 0.044 0.076 -0.041 0.047 0.114 -0.108 0.058 0.452 0.242 0.139 -0.076 F = 18.218 N = 439 *** p < 0.001 Note: Significance values are for a one-tail t-test. A reliability test of these eight indicators yielded a Cronbach’s coefficient of 0.90. It is important to note that there appear to be no trade-offs between the eight dimensions of performance used to create the dependent variable, as all eight indicators are positively correlated with each other and have positive loadings on the factor that was derived through principal components factor analysis. 3 Relational contracting has origins in work on interorganizational collaboration and partnership, non-profit management, social work, and legal studies. 4 The need to preserve an effective working relationship is particularly pressing when the government has few or no alternative providers to turn to (i.e., situations of contracting in thin markets), or when significant asset specific investments have been made by the parties, which creates a “lock in” effect that creates incentives for the parties to negotiate a settlement to a dispute (Williamson, 1985, 1996). 5 Every five years, ICMA conducts an alternative service delivery survey of all larger local governments (cities with populations of 10,000 and over and counties with populations of 25,000 and over) and of a sample of smaller local governments. The survey asks local governments to identify the services that are contracted out and the type of external provider that is used (e.g., a governmental agency, a for-profit provider, or a non-profit provider). The response rate for these surveys has ranged from approximately forty percent to about twenty-five percent for the most recent survey from 2002-2003; the total number of observations has ranged from about 1,200 to 1,500 local governments. 6 Interestingly, recent studies have found that population estimates are affected only slightly by very large changes in the response rate (e.g., Keeter, et al., 2000; Curtin, et al., 2000). 1 2 33 7 None of the independent variables has a tolerance statistic of less than 0.526 and none has a VIF of greater than 1.900. There are eight observations with a leverage value greater than the 2p/N threshold (.114 for the model), where p is the number of independent/control variables and N is the number of observations (Norusis, 1993). A leverage value that is greater than 2p/N is indicative of an observation that is exerting a great deal of influence on the path of the fitted equation. To test the sensitivity of the model to these eight leverage points, I ran OLS regression eight additional times, each time dropping one of the leverage points from the analysis. Dropping each of the leverage points fails to cause any meaningful change in the regression coefficients, t-scores, the R2 or the F value. The eight leverage points are included in the analysis, therefore. 9 A third possible explanation is the somewhat low level of variability in the responses to the questions about whether public employees were allowed to bid on the contract and about the number of contracts that were awarded. About 80% of respondents indicated that public employees were not allowed to bid on the contract, and over 80% of respondents indicated that the contract was awarded to just one provider, as opposed to multiple providers. Lack of significant variability in the responses may help to explain why neither one of these variables is statistically significant in the OLS regression. 10 The coefficients for inspections of work in progress, complaints monitoring, and citizens surveys are much closer to achieving statistical significance than the other three tools and approaches, and the coefficients for complaints monitoring and citizen surveys are in the anticipated positive direction. 11 Jeffries and Reed (2000), who have acknowledged that trust often has a positive effect on the outcomes of a contractual relationship or partnership, have argued that there is a downside associated with too much trust. According to them, very high levels of interpersonal trust can have a negative impact on the outcomes of a contractual relationship, thus suggesting the possibility of a curvilinear relationship between trust and performance. As they explain, parties that trust each other too much tend to economize on time (and overlook sound alternatives) in order to reach the first mutually acceptable solution, even though this solution is unlikely to be the optimal one. The survey data offered no indication of a nonlinear relationship between trust and contracting effectiveness. The bivariate plot between trust and the dependent variable exhibits a clearly linear relationship. Moreover, various nonlinear transformations of trust (including inverse, quadratic, and cubic) were tried, but none of them yielded a better fit for the bivariate equation. Even so, future research should explore the possibility of a curvilinear relationship between trust and performance. 12 The positive and statistically significant correlation between trust (X16) and joint problem solving (X13) (r = 0.29) adds support to this line of reasoning. 8 13 The Durbin-Wu-Hausman test requires developing a new equation to predict the independent variable perceived to be endogenous (e.g., trust). This was done by regressing trust on all of the exogenous variables in the original structural equation predicting contracting performance, as well as the following variables: a dummy variable for whether or not the contractor had provided the service in the past; a dummy variable for for-profit/not-for-profit contractor; a dummy variable for whether or not the contract could be renewed indefinitely; the number of years the contracted was awarded; the yearly dollar amount paid to the contractor; and the number of contracts awarded for the service (R square = 0.361). The residual from this equation was then saved and inserted into the original structural equation that predicts contracting performance. The coefficient for the residual is not statistically different from zero (p value = 0.621), indicating no statistical endogeneity involving the variables trust and contracting performance. 14 Over the last several decades, the field of organization theory has witnessed the emergence of several prominent rational adaptive theories of organizational change and behavior, including Lawrence and Lorsch’s contingency theory (1967), the structural contingency framework developed by James Thompson (1967), resource dependence theory (Pfeffer and Salancik, 1978), and transaction cost economics (Williamson, 1985). Rational adaptive theories describe how managerial leaders cope with environmental uncertainty and turbulence by modifying the organization’s structures and processes to “fit” the particular environment so that the organization can thrive in it. Taken as a whole, the findings from this study seem to be in harmony with the rational adaptive perspective in organization theory, which, like relational contracting, underscores the importance of flexibility and adaptability in management. 15 One concern in survey research is that responses are biased to reflect mainstream norms and expectations in the field. For instance, in the area of contracting for services, public managers frequently hear or read about the need to monitor contracts rigorously, and as a result, we would expect their answers to reflect this expectation. 16 There is additional evidence to suggest that social desirability bias is not widespread in this study. Public managers who respond that they behave in ways that conform to the norms and expectations of the field (e.g., fostering competition, monitoring contracts rigorously, strictly enforcing contracts) probably also have a tendency to indicate high levels of contracting performance. Because the conventional wisdom on contracting reiterates these norms and expectations again and again, it seems likely that we would have found results supporting this dominant perspective (e.g., a positive correlation between competition and contracting performance, a positive correlation between contract 34 monitoring and contracting performance, etc.). Instead, the results tend to support a relational or collaborative view of contracting that is a more recent perspective and one that does not carry as much currency in the field as the dominant perspective influenced by principal-agent theory, microeconomics, and well-established procurement practices. 17 Conversely, others such as Kelman (1990), Cooper (2003), and Dicke (2002) have echoed the words of Friedrich (1940) when they warn of the shortcomings inherent in relying on external mechanisms of accountability. From the perspective of this later set of experts, the answer to the problem of accountability lies in striking a more even balance between external mechanisms of accountability and internal ones, such as shared values and trust. 18 Bivariate correlations between trust and the two measures of contract monitoring used in this study show that the extent of contract monitoring is not significantly related to the level of trust (Pearson correlations of 0.01 and -0.10). 35