What is the relationship between organizational slack and innovation? *.

Innovation is generally viewed as critical to the competitive health of organizations, industries, and nations. As such, how, why, when, and which organizations innovate has been a subject of interest in such diverse literatures as economics, business strategy, R&D management, organizational

psychology, and finance. One important aspect of innovation writings concerns the nature of organizational resources needed to support it. We can think of resources in terms of people, equipment, technology, information, special relationships, and even unique items such as patents or reputations. However, financial resources are probably the most necessary, if not sufficient, element in ensuring the translation of creative ideas into new processes, products or services. This is because innovation requires investments that go beyond those needed to address immediate operational needs, and because financial resources are a primary means for acquiring and supporting the people, equipment, technology and infrastructure involved in innovation. For example, Griliches (1990), in his survey, noted that the relationship between R&D spending and innovation has been repeatedly demonstrated.

Organizational slack is one possible source of funding for innovation. It consists of resources available to the firm above-and-beyond those necessary to meet immediate business requirements, fund ongoing programs, or meet explicit objectives (Cyert and March, 1963; March and Simon, 1958). Although, as will be discussed later, there are various definitions of slack (Bourgeois, 1981), all of them reflect the notion of excess resources that both cushion the organization from environmental changes and represent an opportunity for discretionary allocations, such as to innovation activities. According to Rosner (1968), slack allows firms to purchase innovation, absorb failure, bear the cost of developing and implementing innovations, and explore ideas in advance of" an actual need. As an illustration, Lee and Grewal (2004) showed that slack was related to retailers' adoption of the Internet as a communication channel.

However, this "cushion," which can enhance the competitive position of the firm and be a source of funding for innovation activities, may also be viewed as an impediment to organizational performance in general and to innovation activities in particular. At the organizational level, some have argued that slack reflects inefficiencies in organizations (e.g., Thompson, 1967; Yasai-Ardekani, 1986)--witness the emphasis on consolidation, streamlining, downsizing, and other efficiency-driven initiatives of the last 20 years--and that poor internal control systems in public corporations contribute to inefficiencies in deploying such resources (Jensen, 1993). At the innovation level, it has been proposed that slack gives rise to reduced discipline in the management of projects, thereby impairing innovation outcomes (Nohria and Gulati, 1996).

Thus, slack has been viewed as both creating funding opportunities for innovation, and as encouraging wasteful, undisciplined spending that hurts innovation outcomes. Trying to reconcile these positions, Bourgeois (1981) postulated a curvilinear relationship between slack and organizational "success" in general, while Nohria and Gulati (1996) and Geiger and Cashen (2002) extended the argument to the slack-innovation relationship in particular and offered empirical evidence to support such a relationship. These researchers suggest that too little slack may inhibit experimentation leading to innovation, while too much slack may lead to diminishing returns, or accelerating diminishing returns, as a consequence of undisciplined spending and improper oversight of innovation projects.

The purpose of this study is to further explore and extend previous research on the nature of the relationship between organizational slack and innovation in several important ways. First, this study addresses some of the methodological limitations of the oft-cited Nohria and Gulati (1996) study, thereby testing the robustness and generalizability of their findings. Second, given the prominent role played by patents as evidence of inventiveness in innovation research, this study investigates the nature of the relationship between organizational slack and patent-based outcomes--in particular, the importance or impact of firms' patents on subsequent work in the area. Third, to the degree that innovation-related processes and, in particular, patenting activity are at least partially a function of industry dynamics, this study investigates the moderating effects of industries' propensity to patent on the slack-inventions relationship.

Below we review the relevant research from which we generate our hypotheses concerning the role of organizational slack and patent-based innovation. We then proceed to describe our methodology, present our results, and discuss the research and practice implications of our findings.

THEORY AND HYPOTHESES

Different Types of Slack

Slack resources are usually assessed in terms of financial indicators. In the most common typology, researchers distinguish between available, recoverable, and potential slack resources (Bourgeois, 1981; Bourgeois and Singh, 1983; Sharfman et al., 1988; Singh, 1986). Available slack (also referred to as unabsorbed) represents resources available and not yet committed for particular allocations (e.g., resources available to fund innovation or an increase in dividend payments to owners). Recoverable slack (also referred to as absorbed) represents resources that have been absorbed by the organization (e.g., excess overhead), but which could be recovered through increased efficiencies. Potential slack represents future ability to generate resources (e.g., raising additional equity capital).

The research activity resulting in patents is heavily dependent on internal funding requiring internal resources to support it. Hao and Jaffe (1993) noted that the long and uncertain payback, the secrecy required, the inability to explain future concepts to investors, and the intangible nature of the assets produced make it difficult to finance R&D activities with external sources of funding, making firms more reliant on internal funding. Thus, the effects of the availability of such funding, above-and-beyond that required for ongoing operations should be an especially important determinant of patent-based innovation.

Of the three types of slack, potential slack, because it is not a current resource, is the least likely to play a major part in the internal funding thought to be important for funding R&D activities. Furthermore, debt financing has many other implications for a firm, including the expenses of servicing the debt, and its impact on the firm's credit worthiness and stock price, and thus is not going to be tapped lightly. While "absorbed" slack is, in theory, recoverable, the recovery may be difficult and is likely to evoke strong emotions and negative consequences. In other words, to the degree that slack has been absorbed into the organization, there are forces, such as power and politics (Pfeffer, 1992), that work against its recovery. It is usually during times of economic crisis that such recovery is attempted, and then any recovered resources are more likely to be used to cut costs and increase efficiency than to be redeployed to innovation activities.

Thus, available or unabsorbed slack is the most appropriate measure for investigating the slack-patent-based-innovation relationship, since it represents untapped, internal resources that could be used for innovation purposes. In studying the slack-innovation relationship, it seems far more likely that available or "unabsorbed" slack will make salient the allocation choices open to the firm, one such choice being innovation programs.

The Slack-organizational Performance Relationship

The relationship between slack and innovation can be viewed as a special case of the relationship between slack and organizational performance. Most discussions of slack and its impact on organizations go back to the work of Cyert, March, and Simon on the behavioral theory of the firm (Cyert and March, 1963; March and Simon, 1958). Taking a positive view, March and Simon noted that slack supports a differentiation of goals within the organization. Explicitly addressing the slack-innovation linkage, they note that "when an organization has slack money or manpower not committed to going programs, various specializations of function may arise with respect to commitment to new programs and program elaboration" (1958: 187). Thus, their view supports the notion that increased slack will foster innovation, though little is said about when such slack will, in fact, be devoted to innovation activities as opposed to other potential uses. Using similar arguments, others (e.g., Singh, 1986) have argued that slack resources facilitate innovation because innovation activities consume resources, and the risk associated with innovation can be better borne by those firms that enjoy greater resources.

Taking a more negative view of slack, some (e.g., Bourgeois, 1981; Yasai-Ardekani, 1986) have noted that while slack may act as a buffer from environmental pressures, or as a "shock absorber" (Bourgeois, 1981), it may also act to facilitate the creation of sub-optimal systems, processes, and structures that reduce a firm's aggressive explorations of new responses (e.g., Child, 1972). Similarly, Cheng and Kesner point out that for many, "the term 'slack' conjures up a host of negative perceptions" (1997: 1).

The notion of inefficiencies associated with expanding resource availability is also explored in the finance and economics literatures. For example, Jensen (1993) documents a general failure of internal control systems in dealing effectively with environmental changes in economic conditions, technology, productivity, and innovation. He notes that "substantial data support the proposition that the internal control systems of publicly-held corporations have generally failed to cause managers to maximize efficiency and value" (1993: 850). Applying his analysis to R&D productivity, Jensen estimates the varying degrees of efficiency attained by corporations' R&D expenditures, noting that they vary greatly, and that some firms' R&D expenditures actually caused "losses" when compared to alternative investments.

Thus, while it is intuitively obvious that increased resources will enable more initiatives to be pursued, will encourage experimentation, and will enable more projects to be funded to completion, there are also ample arguments that these additional resources may not be efficiently deployed, leading to diminishing, or possibly even negative, outcomes.

The Slack-innovation Relationship

The thinking regarding organizational slack and innovation parallels that regarding slack and organizational performance, in general. In examining whether slack will facilitate or inhibit innovation, Nohria and Gulati (1996) proposed that the seeming contradiction between those who say slack encourages exploration and those who say it breeds inefficiency could be reconciled if the relationship is viewed as non-linear. They note that their proposition rests on the following, interrelated observations. First, there is general agreement that slack promotes experimentation or spending on various projects. The only difference between the "pro-" and "anti-slack" camps is their opinion concerning whether or not the resources will be expended wisely. Second, the number of new initiatives logically increases with an increase in slack, though the outcomes from such initiatives may reflect diminishing returns. Third, there is evidence of diminishing discipline over the selection, support, and timely termination of projects with increases in slack. Although they phrase their reconciliation of these competing explanations in terms of slack having a "non-linear influence on innovation" (1996: 1250), suggesting that "there is an intermediate level of slack in any given organizational setting that is optimal for innovation" (1996: 1250), the hypothesis they test is one for a special case of curvilinearity, namely, that the relationship is inverse U-shaped. They find support for this hypothesis, making a case not only for diminishing marginal returns, but also diminishing total returns beyond a certain point.

Although the Nohria and Gulati (1996) study is important in that it is the first study to empirically demonstrate a curvilinear relationship between slack and innovation, there are several important questions that it could not answer. (1) First, the study was conducted at the departmental rather than organizational level of analysis, using departments from two firms. Accordingly, slack and innovation were operationalized at the department level. Since departments within the same organization are, to a great extent, operating under the general munificence or constraints of the larger organization's resources, and since much of the slack literature is focused on firm-level relationships, the generalizability of these findings to firm-level phenomena needs to be tested.

Second, the study used perceptual, self-report measures of both slack and innovation. This again calls for a test of the generalization of findings to more objective measures of both slack and innovation. Such self-report measures are subject to social desirability effects as well as to "same-source" and "same-method" biases.

Finally, the theoretical justification for an inverted U-shaped relationship also needs to be further examined and tested. There seems to be wide agreement that there are dynamics associated with increased slack that work to both promote and inhibit organizational outcomes such as innovation. It also seems reasonable, based on the studies cited above, that these opposing forces can be reconciled by the proposed curvilinear relationship. However, a diminishing marginal returns model would be a more parsimonious reconciliation of these arguments, and more consistent with such demonstrations in many areas of research. The proposition that innovation actually decreases beyond some point would have to be based on the notion that slack not only creates conditions which foster the undisciplined management of innovation projects (hence, diminishing returns), but that it actually leads to the funding of projects that should not be funded, the failure to fund projects that should be funded, and even the termination of projects that should be continued.

Although some justifications can be found for these types of outcomes (e.g., agency theory, escalating commitment phenomena, power and politics), such explanations are suggestive of a major breakdown of the processes associated with the management of innovation. While one can probably find illustrations of such breakdowns, little exists in the way of theory or empirical evidence to support this extreme case as the more general, universal model. Even in Jensen's (1993) work on agency theory, which is often cited for evidence of inefficiencies due to poor internal control mechanisms, firms varied widely between positive and negative returns on their R&D expenditures, evidencing varying degrees of internal control effectiveness.

Thus, there is a need to extend the Nohria and Gulati (1996) findings using independent and more objective assessments of both slack and innovation, using firm-level data, and more closely examining the nature of any found relationship between slack and innovation, and even factors that may moderate it.

Slack and Patent-based Innovation

Given the earlier-noted diversity of the innovation literature, it is not surprising that there is a corresponding diversity of innovation definitions. However, it can be argued that patent-related statistics may be among the better proxies for innovation (Griliches, 1990). Although the development of a patentable idea may be labeled an "invention," with the term "innovation" generally referring to the successful, market-based implementation of such an invention, patent-based inventions are the starting point for patent-based innovation and, as such, are an important input helping us understand this domain of innovation. Griliches (1990) notes that while we ask many questions about the relative inventiveness of firms and even countries, we have, in fact, very few measures of innovation. Those we do have, tend to be "distantly related, 'residual' measures and other proxies" (1990: 1661). On the other hand, patent statistics, he notes, are readily available and are by definition related to inventiveness. He concludes that "patent statistics are interesting in spite of all the difficulties that arise in their use and interpretation" (1990: 1662).

Some of these difficulties in using patent statistics arise from the fact that firms, and especially industries, vary widely in their propensity to seek patent protection, limiting the generalizability of findings to firms who do so. Even for these firms, simple patent counts may not be a good measure of innovation since some firms are likely to try and patent a myriad of insignificant developments, thus confounding the quantity and quality or impact of innovation represented by patent statistics. To address the quality vs. quantity of patents issue, some have argued against using simple patent counts. For example, Deng, Levand, and Narin (1999) note that patent applications may provide a richer source of data because applications cite other patents upon which the current application is based. Such citations provide evidence concerning the impact any given patent has had on future applications/innovations, thus serving as a proxy for the impact or importance of a given patent.

Important patents get cited more often because they show the direction for further research and, as such, are built upon by others. Trajtenberg notes that "these citations could be taken as first-hand evidence of the path-breaking nature of the original patent" (1990: 184). Similarly, Jaffe, Fogarty, and Banks (1998) concluded that patent citations measure technological impact and knowledge spillover, both of which indicate the quality of the innovation. Finally, Lanjouw and Schankerman (1997) found that litigated patents are cited far more heavily than non-litigated ones, suggesting that companies value their oft-cited patents more and thus defend them more vigorously. Thus, the assessment of patents' quality or impact, as evidenced by frequent citations, may be a potentially important outcome for investigating the slack-innovation relationships.

In the absence of any specific literature guiding predictions about the slack-patenting relationship, it seems reasonable that the general slack-organizational performance and slackinnovation literatures cited above are an appropriate starting point, especially literature addressing R&D activities. On the positive side, Hall (1992) found strong evidence of the effect of liquidity constraints on R&D spending in manufacturing firms. Singh (1986) showed that slack has a positive relationship with organizational performance, risk-taking, and decentralization of decision making, all conceivably being related to the level of R&D activity and its outcomes. The presence of slack should relax managerial controls and allow more discretion in the allocation of funds to new projects. During times of economic stress, slack buffers the organization from the uncertainties of new project initiations and reduces the personal risks or consequences associated with failure, thus fostering such activities (Bourgeois, 1981).

On the other hand, three arguments can be used to support the negative side of the slack-patenting debate. First, we have the above-noted arguments that slack leads to undisciplined project management (e.g., March and Simon, 1958; Nohria and Gulati, 1996). Second, the economic theories and demonstrations concerning the inefficient use of corporate resources (e.g., Jensen, 1993) also suggest that increased slack may not improve organizational performance. Finally, we have the limited evidence from previous research (e.g., Geiger and Cashen, 2002; Nohria and Gulati, 1996), though, again, none of these were specifically addressing patent-based outcomes as performance measures.

Thus, there is ample support for the notion of increasing and then decreasing returns from slack resources to guide our prediction that the same would be true for patent-based inventions. However, the exact nature of the diminishing returns argument is still somewhat problematic. At this point, it seems more prudent to expect increasing slack to produce diminishing marginal returns for patent-based innovations, but not necessarily negative or diminishing total returns.

Hypothesis 1: The relationship between organizational slack resources and the importance of patent-based inventions will be positive, but diminish in strength beyond some intermediate point.

Researchers have repeatedly demonstrated that the intensity of patenting activity varies widely across industries (e.g., Hicks et al., 2001; Mansfield, 1986). Some industries prefer using trade secrets and other means of protecting intellectual capital. Some industries have such short product life cycles that patents would be useless by the time they were granted. Other industries do not want to reveal the amount of information that a patent application requires, while still others do not consider such protection important. Similarly, industries differ in the degree to which innovation and R&D spending are central to their strategy, influencing their reliance on slack resources. O'Brien (2003) found that firms competing on the basis of innovation make financial slack a strategic priority.

Given our focus on patent-related outcomes, the centrality of patenting to the strategies of firms in a particular industry ought to moderate any slack-patenting relationship. In general, one might expect that since the main arguments for diminishing returns from slack revolve around inefficient and undisciplined management of innovation projects, firms in industries where patenting is critical to success and survival will have learned to minimize such tendencies as a part of their survival mechanisms. However, firms that are in industries that are less patent-intensive may not have focused organizational resources or learning to protect against such inefficiencies or lack of discipline in managing patent-focused projects.

Support for this proposition can be found in several streams of research. Although early population ecology theorists (e.g., Hannan and Freeman, 1984) suggested that firms' success or failure was a function of the match between their "demographics" and environmental conditions, later work (e.g., Singh, 1990) has included considerations for how the environment shapes change or adaptation in organizations, a more evolutionary view. If patentable inventions are important for survival in particular industries, the surviving firms will have developed capabilities appropriate for managing this process, learning to counter any tendencies toward inefficient or undisciplined innovation project management.

Another theoretical perspective on industry differences comes from resource-based views of the firm (e.g., Barney, 1991) that see resources as an important determinant of firms' performance. When such resources are "valuable, rare, difficult to imitate and non-substitutable," they "yield sustained competitive advantage" (Meyer, 1991: 823). This perspective would suggest that the management of innovation or R&D is one such core competence that firms possess as a requirement for keeping up or excelling in industries more heavily dependent on patenting activity. Thus, both ecological and resource-based views of firms' success would support an evolutionary explanation of why firms that survive in patent-intensive environments have developed better internal control mechanisms so as to better counter inefficiencies associated with poor internal control.

Finally, we can look to the project management literature for evidence that the management of innovation projects may be an important determinant of innovation outcomes. This literature provides evidence supporting the link between project management practices and R&D outcomes, further suggesting moderated, rather than universal, relationships between organizational slack and outputs such as patents.

Though based on mathematical models more so than field research studies, several recent studies have shown that complex projects, such as R&D projects, are subject to serious coordination problems, but that such problems can be mitigated by improved management of the process. For example, Mihm, Loch, and Huchzermeier (2003) modeled conditions under which complex design projects oscillate or diverge to low performance solutions, from which they derive managerial actions aimed at improving the performance dynamics of such projects. These actions are clearly related to the efficient and effective management of innovation projects (e.g., dealing with system size, communication of updates, broadcasting of preliminary and intermediate information, engendering cooperation). Similarly, Gutierrez and Kouvelis (1991) address managerial actions aimed at addressing pervasive evidence about the difficulties of meeting project completion deadlines.

Based on both the general observation that firms in some industries will have better mastered the efficient use of resources for patent-based innovation projects, thus mitigating the expectation of decreased efficiency and discipline attributable to excess slack, and supported by both firm-level and project-level research in related areas, we posit that the patenting intensity of industries will moderate the stack-innovation relationship. In other words, firms in industries that rely more strongly on patenting activity for competitive survival will have developed better mechanisms for initiating, managing, and even terminating projects, and thus are less subject to the diminishing re turns associated with increased slack.

Hypothesis 2: Industry patenting intensity will moderate the relationship between slack resources and the impact of firms' patents. Specifically, diminishing returns will be more evident in low-patent intensity industries.

METHODS

Variables

In this section we describe the operationalization of the dependent and independent variables used in the study, as well as the considerations that went into selecting appropriate time lags between the availability of resources and patent-based outcomes. We also discuss the selection of control variables to reflect organizational size and the propensity to invest in R&D activities.

Dependent Variable. The innovation measure used in this study assessed the impact of a firm's inventions, focusing on the "quality" of innovations rather than the "quantity." This measure is based on patent-related information used to measure patent impact from TECH-LINE, a technology indicator database marketed by CHI Research, Inc. TECH-LINE compiles data for the most active patentees in the U.S. patent system, including about 350 U.S. companies. Companies in the database must have approximately 50 patents during the previous five years, and the database accounts for mergers or acquisitions of the patenting company. In a recent study, Deng et al. (1999) used this database to investigate the relationship between science and technology activities and stock performance, whereas Hicks et al. (2001) used it to study the changing composition, by geography and industry, of innovative activity in the U.S.

In this database, the Citation Impact Index (CII) is a measure of the importance of a firm's innovation relative to other firms, as indicated by the citations of the firm's patents in others' subsequent patent applications. The index is a count of the number of citations in a given year referencing a firm's patents issued in the most recent five years, divided by the average number of citations of all patents in the database for the corresponding years. Thus, a value of "1" reflects average impact, whereas higher values indicate greater impact. Data for the CII were available for the period from 1990-1999, but large amounts of missing data for 1999 resulted in that year being excluded from the study.

Independent Variables. Slack was measured using data available from the Compustat database. Since we wanted the measure of slack to reflect unabsorbed or available resources, we used the "quick ratio" as a measure of liquidity (Geiger and Cashen, 2002; Hao and Jaffe, 1993). The quick ratio is defined as current assets, minus inventories, divided by current liabilities. Current assets are assets that can be converted to cash or used in the business within a relatively short period of time. They consist of cash, marketable securities, receivables, inventories, and other short-term assets. Current liabilities are financial obligations that have to be paid within a year. Thus, the quick ratio indicates the extent to which current assets, not counting inventories, cover the current liabilities. (An alternative measure of slack, the "current ratio," was also examined, and produced essentially the same results.)

Because the hypotheses predicting that slack resources impact innovation activity imply some sort of time-lag between the availability of the resources, their deployment, and the production of innovation, we needed to allow for that lag in our analyses. We assumed that it takes at least a year for slack resources to produce patentable outcomes. In addition, a large majority of patent applications take at least a year to be approved, and these patents would then have to be in the public domain for some period of time so that they can be cited. Thus, we chose a three-year time lag, averaging the slack variable for five-year periods (same as the CII variable) that reflected this three-year time lag. In other words, 1987-1991 quick ratios were averaged to predict the 1994 CII, while 1991-1995 quick ratios were averaged to predict the 1998 CII. (Two- and four-year time lags were also explored and produced essentially the same results.)

Industries' propensity to patent was assessed by dividing the total number of patents for a given industry in the database by the total number of patents in the database. This provided a measure of the relative intensity of patenting activity in that industry.

Control Variables. Since size of the organization is likely to influence the absolute level of resources devoted to R&D activities, we used total assets as a control variable. Because the proportion of total resources devoted to R&D activities may also influence patent-related activities, R&D intensity (consisting of R&D expenses divided by total sales) was also used as a control variable.

Analyses

To examine the stability of results across time periods and economic cycles, the nine-year period of available CII data was split into two time periods. To minimize overlap in time periods, since each year's CII reflects citations to patents issued in the most recent five years, the 1994 and 1998 CII data were used, resulting in only one year (1994) overlapping both time periods.

Cross-referencing the patent database with the financial database yielded 261 companies for which both innovation and financial data were available for the nine-year period. To test Hypothesis 1, that the slack-innovation relationship is curvilinear, stepwise multiple linear regression was used to test the impact of both the slack ratio and the square of the slack ratio on the patent measure, after controlling for total assets and R&D intensity. To test Hypothesis 2, that industry patenting intensity will interact with slack to influence innovation outcomes, we entered the interaction term for these two variables.

RESULTS

The results of our analyses are presented in three parts. First, we present summary statistics and zero-order correlations for all study variables for both time periods. Next, we present the tests of our hypotheses concerning the slack-innovation relationship and the moderating effect of industry patenting intensity. Finally, we will describe a supplemental analysis to further explore the shape of the curvilinear relationships found.

Table 1 reports the means, standard deviations, and intercorrelations for study variables for each of the time periods. From Table 1, we note that the Citation Impact Index is significantly negatively correlated with firm size for both time periods, significantly positively related to firms' R&D intensity for the 1994 data only, significantly positively related to slack for both time periods, and significantly positively related to the industry patenting intensity measure for 1998 data only. The within-firm variable correlations (size, R&D intensity and slack) are all moderate and significant, though assets are negatively correlated with both R&D intensity and slack for both time periods. The fact that both total assets and R&D intensity are used as control variables in our analyses makes any test for the effects of slack more conservative. Finally, it is interesting to note that the industry patenting intensity variable is unrelated to any of the firm-level measures.

Table 2 reports the results for the different models predicting the CII. After controlling for firm size and R&D intensity, the quick ratio is significant and accounts for significant increases in explained variance (Model 2). The subsequent inclusion of the squared term yields a significant negative coefficient for both time periods, with a 25% increase in [R.sup.2] for the 1994 data, and a 130% increase for the 1998 data (Model 3). This model supports Hypothesis 1 and confirms the negative curvilinear relationship found by Nohria and Gulati (1996). Adding the industry patenting intensity variable (Model 4) results in a significant coefficient for the 1998 data, along with a major increase in [R.sup.2], but not for the 1994 data. Finally, the interaction term is significant for the 1998 data, and marginally so for the 1994 data (Model 5).

Graphing the interaction for each time period using points +/- one standard deviation from the mean (Figure I), we note that for both time periods, industries that are more patent intensive display a much stronger positive relationship between slack and the impact of their patents than do firms that are lower in patenting intensity. This finding provides support for Hypothesis 2 and suggests that the nature of curvilinear effects when investigating slack-innovation relationships may mask industry differences.

[FIGURE I OMITTED]

Next, we investigated whether the negative squared slack term should be taken to represent an inverted U-shaped relationship, as others have suggested. While a quadratic term indicates a parabolic curve, and the negative sign indicates direction, these factors do not necessarily demonstrate an inverted U-shape. To do so would require the demonstration of an inflection point beyond which the curve becomes downward sloping, as opposed to just asymptotic, and a demonstration that this point is not just a statistical abstraction, but that it is within the range of acceptable or realistic values of the independent variable. Plotting the values of the quick ratio beyond which our dependent variable would actually decrease in value (as opposed to increasing more slowly), we found that this value was 4.57 for the 1994 data and 4.14 for the 1998 data. Only two firms out of 212 and four firms out of 242 in the 1994 and 1998 data, respectively, had quick ratios greater than these values. Thus, the relationship between slack and innovation, at least for the sample and outcome assessed here, may be better thought of as one of diminishing marginal returns, but not necessarily diminishing total returns, except, possibly, for extreme values of slack.

DISCUSSION

This study's findings provide support for the general proposition that there is a curvilinear relationship between organizations' unabsorbed slack resources and at least one aspect of innovation--the importance of their inventions as measured by the impact they have on subsequent research in the field. As such, it represents another demonstration that the seemingly contradictory attributions made to organizational slack can be reconciled when explaining the impact of slack on innovations. In essence, too little slack may inhibit exploration programs that lead to innovation, while too much slack may, in fact, result in reduced benefits beyond a certain point.

This finding is an important extension of the Nohria and Gulati (1996) findings in that it was based on a cross-sectional sample, using different and objective measures of both slack and innovation, and replicated for two different time periods. Furthermore, the nature of this relationship remains robust for different financial measures of slack, as well as for different time lags for the slack-invention relationship. This increases our confidence that the relationships found are not a function of a particular time's economic conditions, political or regulatory environment, or other transient or cyclical forces, though it should be noted that the strengths of relationships varied between the time periods. Such differences may be due to changes in business conditions (the 1994 data reflect a period of slower economic growth, including a recession), or a general change in the rate and pattern of patenting activities, as suggested by Hicks et al. (2001), who noted explosive growth in patenting activity in the U.S. between 1980 and 1999. An examination of our total patent data for the two time periods did reveal a 27% increase in patenting activity across all firms between the 1994 and 1998 data.

More importantly, however, this study found support for more complex, contingency notions related to the slack-innovation relationship. Specifically, for industries that rely more on patent protection, the relationship between slack and patenting outcomes was considerably more positive than for those who do not (though this finding was only marginally significant for the 1994 data). This suggests that the primary explanations for curvilinearity--that inefficient and undisciplined management of the additional resources accounts for diminishing returns--may be less plausible for firms in industries that need to be efficient in these activities in order to remain competitive.

At a general level, this finding alerts us to an important contingency factor that may need to be worked into our theories of innovation or, at minimum, be understood as a limit to generalizations from studies that do not specify or try to account for such industry differences. Other industry factors also need to be investigated.

For example, Martinez and Artz (2006) found that industries' regulatory environments affected the relationship between a firm's slack resources and managers' propensity to invest in higher-risk activities, such as those associated with innovation. In addition to environmental, industry, or firm-level variables, other contingencies worth investigating relate to the nature of the innovations themselves and their dependency on internal funding from slack resources. Incremental or sustaining innovations may require fewer resources than disruptive ones (Christensen and Raynor, 2003), and thus may be differentially related to slack resources. At a more specific level, these findings suggest that theories of innovation may need to specify industry effects, and that certain propositions about innovation may be industry-dependent. Without such contingency or moderator hypotheses, findings may mask or average differential relationships for different sub-groups. Furthermore, we need to do a better job of incorporating existing theories and models of organizations, such as ecological/evolutionary views, resource-based views, and organizational learning views, in order to develop testable propositions about why certain types of relationships between innovation inputs and outputs may occur.

The question of whether increased slack simply decreases the efficiency of innovation outputs or actually results in negative outputs beyond a certain point, as some have suggested, was addressed by our results with the finding that for very extreme values of slack, "negative" innovation outcomes may, in fact, be found. However, these extreme cases clearly constitute "outliers," and one should be careful about specifying a more general, inverted U-shaped curve for the slack-innovation relationship. Future research will need to investigate the underlying causes for this effect, when it does occur. The common explanations given for diminishing returns (i.e., lack of discipline, decreased vigilance, etc.) are not quite adequate for explaining why innovation would actually decrease beyond some point unless we hypothesize and test propositions specifically designed to assess program or resource management deficiencies that may occur. Given the very high liquidity ratio beyond which this effect occurs in our samples, one would need to better understand the dynamics of firms that maintain such high ratios. Perhaps there are correlates or causes of such high ratios that might explain the negative returns on innovation programs. Thus, our results caution against the "negative innovation" returns interpretation of the earlier slack-innovation studies.

Our findings also point to the importance of developing multiple measures of innovation, and the need to better map the conceptual distinctions between these different measures and delineate the conditions under which each conceptualization is most appropriate. For example, when company data about innovation activities are unavailable (which they usually are, for trade secrecy reasons), the ability to specify and use a patent-based, publicly-available measure represents a major benefit. Given the public nature of patent data, the explosive growth in patenting activity (Hicks et al., 2001), and the support for the use of patent data as proxies for innovation (Deng et al., 1999; Griliches, 1990; Trajtenberg, 1990), our results demonstrate the salience of such patenting data for testing hypotheses concerning various aspects of organizational resources and innovation outcomes. However, even then, we may specify different patent-based measures. Our measure of "impact" may be said to have provided a demonstration of the impact of slack resources on the quality or importance of a firm's innovations, but not on their quantity or their commercial success.

In conclusion, this study has shown that organizational resources constitute an important determinant of innovation outcomes, and that the relationship between resources and outcomes may not be linear, but may be best represented as curvilinear. However, the strength of the relationship varies widely for different industries. These findings have theoretical significance for research on innovation and our results hopefully shed light on some issues of conceptualization and operationalization of both the innovation and slack resources variables.

Our findings also have practical significance in that they point to the need to balance competing pressures to be efficient and yet make investments in future performance through innovation. While "lean-and-mean" may be a popular mantra, being too lean may mortgage one's future. Some (e.g., Hamel and Prahalad, 1994) have voiced these concerns, but little theory or research exists for guiding managers in dealing with this dilemma. Our findings suggest that restricting the availability of resources that may be used for innovation projects has the effect of diminishing the importance or impact of the innovation thus produced. However, when resources are plentiful, the significance of the innovations produced depends on industry factors (among others) that may be associated with the experience and associated project-related discipline that is embedded in industries heavily dependent on patenting activity.

If that is the case, those firms not possessing such experience or discipline may, in fact, be squandering resources as they accumulate and allocate organizational slack in the hope of increasing innovativeness. On the other hand, the experienced and disciplined firms may yield significant rewards from the accumulation of slack and its allocation to innovation activities. In fact, Ping-Hung et al. (2003) found that for the pharmaceutical and chemical industries, investments in R&D earned an operating margin return much higher than the industry cost of capital and the return on investments in fixed assets.

Hopefully, this research can point the way toward some approaches for managing this trade-off. Keeping resources tight seems to hurt innovation. Amabile, writing about human creativity, notes that the:

   main resources that affect creativity are
   time and money.... Interestingly, adding
   more resources above a "threshold of sufficiency"
   does not boost creativity. Unfortunately,
   many managers don't realize this
   and therefore often make another mistake.
   They keep resources tight, which pushes
   people to channel their creativity into finding
   additional resources, not in actually developing
   new products or services (1998:
   82).

This conclusion concerning individual creativity seems to parallel our conclusions concerning organizational innovation.

If the diminishing returns associated with increased levels of slack are due to lack of programmatic discipline, firms may want to tighten their administration of R&D, the milestone review process, and the choice of criteria for deciding upon continuing funding for research projects. On the other hand, if it is found that increased resources above a certain point do not sufficiently boost innovation, regardless of how well these resources are managed, then resource allocation decisions should take this into account when deciding about channeling funds into innovation-related activities or other corporate needs.

Strengths and Limitations

As with all research, our results should be evaluated in terms of several limitations that will need to be addressed by future research. Most of these limitations pertain to the patent-based data set. Although the publicly-available data used as our dependent measures offer a good alternative for the assessment of the usually more secret innovation activities of a firm, the use of such data represents several limitations. First, such data are skewed toward firms and industries that are likely to rely on patenting as a means of protecting their innovations. To the degree that firms do not rely heavily on patenting, use copyrights, trademarks or other means of protection, or even believe that trade secrets will be exposed through the patent application process, they will be underrepresented in our data. The use of patent-based data is thus most appropriate for research about industries that make regular use of patent law provisions.

Second, the TECH-LINE database covers only companies that exhibit a certain level of patenting activities, thus skewing the data toward larger, better-capitalized firms. To the degree that smaller, entrepreneurial firms account for significant patenting activity, they are not reflected in such databases. Furthermore, to the degree that these larger, better-capitalized firms may be operating in more mature industries and focus more on sustaining than disruptive innovations (Christensen and Raynor, 2003), our results may generalize better to larger firms and sustaining innovations. Finally, although we examined two different time periods, we do not know how these results may hold up as business conditions change. Thus, future research needs to address the trade-offs between more objective, publicly-available data, and the limits to generalizations from such data. Finally, the database represents only companies filing for patent protection in the U.S.

In spite of these limitations, our research benefited from several strengths. First, we used a relatively large, cross-sectional sample of firms. Second, we used objective and independent measures for both the slack and the innovation variables, eliminating major sources of methodological threats to validity. Third, we examined the stability of our results across two different time periods, thus minimizing the likelihood that results are due to the effects of unique or temporary environmental events, such as economic cycles, interest rates, etc. Finally, we time-lagged our dependent variable to more closely reflect the temporal relationship between investments in innovation and innovation outcomes (though it should be noted that there are various ways of operationalizing such time lags; see Ping-Hung, Mishra, and Gobeli (2003) for an alternative method). Hopefully, these study attributes allow the results to make a contribution to furthering our understanding of how corporate resources impact innovation outcomes.

References

Amabile, T. M. 1998. "How to Kill Creativity." Harvard Business Review (September-October): 77-87.

Barney, J. B. 1991. "Firm Resources and Sustained Competitive Advantage." Journal of Management 17: 99-120.

Bourgeois, L.J. 1981. "On the Measurement of Organizational Slack." Academy of Management Review 6: 29-39.

--and J.V. Singh. 1983. "Organizational Slack and Political Behavior within Top Management Teams." Academy of Management Proceedings (August): 43-47.

Cheng, J. L. C. and I. F. Kesner. 1997. "Organizational Slack and Response to Environmental Shifts: The Impact of Resource Allocation Patterns." Journal of Management 23: 1-18.

Child, J. 1972. "Organizational Structure, Environment and Performance: The Role of Strategic Choice." Sociology 6: 1-22.

Christensen, C. M. and M. E. Raynor. 2003. The Innovator's Solution. Boston, MA: Harvard Business School Publishing Corporation.

Cyert, R. M. and J. G. March. 1963. A Behavioral Theory of the Firm. Cliffs, NJ: Prentice-Hall, Inc.

Deng, Z., B. Levand and F. Narin. 1999. "Science and Technology as Predictors of Stock Performance." Financial Analysts Journal 55: 20-32.

Geiger, S. W. and L. H. Cashen. 2002. "A Multidimensional Examination of Slack and its Impact on Innovation." Journal of Managerial Issues 14 (1): 68-84.

Griliches, Z. 1990. "Patent Statistics as Economic Indicators: A Survey." Journal of Economic Literature 28: 1661-1707.

Gutierrez, G.J. and P. Kouvelis, 1991. "Parkinson's Law and its Implications for Project Management." Management Science 37: 990-1001.

Hall, B. 1992. "Investment in Research and Development at the Firm Level: Does the Source of Financing Matter?" National Bureau of Economic Research, Working Paper No. 4096.

Hamel, G. and C. K. Prahalad. 1994. Competing for the Future. Boston, MA: Harvard Business School Press.

Hannah, M. T. and J. Freeman. 1984. "Structural Inertia and Organizational Change." American Sociological Review 49: 149-164.

Hao, K. Y. and A. B. Jaffe. 1993. "Effect of Liquidity on Firms' R&D Spending." Economics of Innovation and New Technology 2: 275-282.

Hicks, D., T. Breitzman, D. Olivastro and K. Hamilton. 2001. "The Changing Composition of Innovative Activity in the U.S.: A Portrait Based on Patent Analysis." Research Policy 30: 681-703.

Jaffe, A. B., M. S. Fogarty and B. A. Banks. 1998. "Evidence from Patents and Patent Citations on the Impact of NASA and Other Federal Labs on Commercial Innovation." The Journal of Industrial Economics XLVI: 183-205.

Jensen, M. C. 1993. "The Modern Industrial Revolution, Exit, and the Failure of Internal Control Systems." Journal of Finance 48: 831-880.

Lanjouw, J. and M. Schankerman. 1997. "Stylized Facts on Patent Litigation: Value, Scope and Ownership." National Bureau of Economic Research, Working Paper No. 6297.

Lee, R. P. and R. Grewal. 2004. "Strategic Responses to New Technologies and Their Impact on Firm Performance." Journal of Marketing 68 (4): 157-172.

Mansfield, E. 1986. "Patents and Innovation: An Empirical Study." Management Science 32 (2): 173-181.

March, J. A. and H. A. Simon. 1958. Organizations. New York, NY: Wiley.

Martinez, R.J. and K. Artz, 2006. "An Examination of Firm Slack and Risk-taking in Regulated and De-regulated Airlines." Journal of Managerial Issues 18 (1): 11-31.

Meyer, A. D. 1991. "What is Strategy's Distinctive Competence?" Journal of Management 17: 821-833.

Mihm, J., C. Loch and A. Huchzermeier. 2003. "Problem-solving Oscillations in Complex Engineering Projects." Management Science 49: 733-750.

Nohria, N. and R. Gulati. 1996. "Is Slack Good or Bad for Innovation?" Academy of Management Journal 39 (5): 1245-1264.

O'Brien, J. P. 2003. "The Capital Structure Implications of Pursuing a Strategy of Innovation." Strategic Management Journal 24 (5): 415-432.

Pfeffer, J. 1992. Managing with Power. Boston, MA: Harvard Business School Press.

Ping-Hung, H., C. S. Mishra and D. H. Gobeli. 2003. "The Return on R&D versus Capital Expenditures in Pharmaceutical and Chemical Industries." IEEE Transactions on Engineering Management 50: 141-150.

Rosner, M. M. 1968. "Economic Determinants of Organizational Innovation." Administrative Science Quarterly 12: 614-625.

Sharfman, M. P., G. Wolf, R. B. Chase and D. A. Tansik. 1988. "Antecedents of Organizational Slack." Academy of Management Review 13: 601-614.

Singh, J. V. 1990. Organizational Evolution: New Directions. Newberry Park, CA: Sage.

--. 1986. "Performance, Slack, and Risk Taking in Organizational Decision Making." Academy of Management Journal 29: 562-585.

Thompson, J. D. 1967. Organizations in Action. New York, NY: McGraw Hill.

Trajtenberg, M. 1990. "A Penny for Your Quotes: Patent Citations and Value of Innovations." RAND Journal of Economics 21:172-187.

Yasai-Ardekani, M. 1986. "Structural Adaptations to Environments." Academy of Management Review 11 : 9-21.

David M. Herold

Professor of Organizational Behavior

Georgia Institute of Technology

Narayanan Jayaraman

Professor of Finance

Georgia Institute of Technology

C. R. Narayanaswamy

Associate Professor of Finance

Clayton State University

* The authors would like to thank the Editor, an anonymous reviewer, Steve Caldwell, Don Fedor, Jinsoo Lee, Charles Parsons and Frank Rothaermel for their suggestions on an earlier version of this article.

(1) A study by Geiger and Cashen (2002) demonstrated a curvilinear relationship between slack and R&D expenditures. Since R&D expenditures are inputs into the innovation process, and thus not subject to the arguments concerning efficient or inefficient use of the resources in innovation activities, this study is not reviewed here.

Table 1
Intercorrelations among Study Variables

1994 Data (N = 212)

Variable             Mean   S.D.

CII                  1.31    .78
Total Assets         7.21   1.65   -.27 **
R&D Intensity         .10    .38   .19 **    -.29 **
Quick Ratio          1.21    .87   .33 **    -.47 **   .48 **
Industry Patenting
  Intensity           .07    .04       .08       .05      -12    -.01
                                                       ([dagger])

** p < .01, ([dagger]) p <.10

1998 Data (N = 242)

Variable             Mean   S.D.

CII                  1.42   1.19
Total Assets         7.34   1.57   -.23 **
R&D Intensity         .48   2.99       .06   -.27 **
Quick Ratio          1.18    .86   .33 **    -.50 **   .24 **
Industry Patenting
  Intensity           .07    .05   .39 **       -.02      -.10    .06

** p <.01

Table 2
Regression of CII on Financial and Patenting Data

                                  1994 CII
Independent
Variable             Model 1       Model 2      Model 3

Intercept            2.09 **        1.52 **      .86 *

Total Assets         -.11 **        -.07        -.03

R&D Intensity        -.24           -.05         .18
                    ([dagger])

Quick Ratio                          .23 **      .64 **

Quick Ratio
Squared                                         -.07 **

Industry
  Patenting
  Intensity

Industry Patent
  Intensity X
Quick Ratio

N                      212           212          212

Adjusted
  [R.sup.2]           .08 **         .12 **      .15 **

[DELTA]
  [R.sup.2]                          .04         .03

                              1994 CII
Independent
Variable             Model 4       Model 5

Intercept              .81 *        1.02 **

Total Assets          -.03          -.02

R&D Intensity          .01           .26

Quick Ratio            .62 **        .38
                                  ([dagger])
Quick Ratio
Squared               -.07 **       -.06 *

Industry              1.54         -2.17
  Patenting
  Intensity

Industry Patent
  Intensity X
Quick Ratio                         3.20

N                      212          212

Adjusted
  [R.sup.2]            .15 **        .16 **

[DELTA]
  [R.sup.2]            .00           .01

                                   1998 CII
Independent
Variable             Model 1       Model 2      Model 3

Intercept            2.73 **         1.54 **     -.47

Total Assets         -.18 **         -.08         .04

R&D Intensity         .00            -.02         .00

                                      .39 **     1.74**

Quick Ratio                                      -.21**

Quick Ratio
Squared

Industry
  Patenting
  Intensity

Industry Patent
  Intensity X
Quick Ratio
                      242             242         242
N
                      .05 **          .10 **      .23 **
Adjusted
  [R.sup.2]

[DELTA]                               .05         .13
  [R.sup.2]

                              1998 CII
Independent
Variable             Model 4       Model 5

Intercept            -.53           .31

Total Assets          .02           .00

R&D Intensity         .01           .02

                     1.40 **        .62
                                  ([dagger])

Quick Ratio          -.16 **       -.09 *

Quick Ratio
Squared              7.26 **       -.77

Industry
  Patenting
  Intensity

Industry Patent
  Intensity X                       6.79 **
Quick Ratio
                      242           242
N
                      .31 **        .33 **
Adjusted
  [R.sup.2]

[DELTA]               .08           .02
  [R.sup.2]

** p < .01, * p < .05, * p < .10