The relationship between firm investments and technological innovation and political action.

I. Introduction

Economic theory recognizes the importance of investments in both technological innovation and political activity in firm behavior, and extensive research exists which examines independently these expenditures. However, surprisingly little attention is devoted to analyzing technological innovation and political action expenditures jointly. Technological innovation and political action are two of the options firms face regarding how to achieve lower costs or market power; the characteristics of these choices not only affect the degree to which each is used, but also have widely divergent normative and policy implications for society. Furthermore, many forms of technological innovation require regulatory approval, so that firms may use the two forms of investment in conjunction with each other in addition to treating them as separate investment alternatives.

This paper investigates the relationships between firm investments in technological innovation and political activity. Both forms of investment seek to lower costs and/or create market advantage, but economists have long recognized the social importance of technological innovation to economic growth, while treating most firm political activity as inefficient. Growing social concerns surrounding declining innovation, productivity, and growth and increasing firm political activity suggest that these are timely dimensions to explore jointly.

In the next section I discuss the relationship between firm investments in technological innovation and political action. This discussion motivates the subsequent empirical modelling presented in section III. Because of the possibility of simultaneity between firm spending in research and development (R&D) and Political Action Committee (PAC) spending, and heavy censoring of PAC data, estimation is in the context of a simultaneous censored ("Tobit") model suggested by Smith and Blundell [29]. Data for the study, and expected relationships between R&D, PAC, and control variables, are discussed in section IV. Estimation results are presented in section V. The endogeneity of firm R&D expenditures as a determinant of firm PAC expenditures is not rejected, and there is significant evidence that the two forms of investment are complements; these results hold true to various degrees over industries and industry groups. Section VI offers conclusions and suggestions for further research.

II Background

Consider the firm's choices concerning discretionary investments in technical innovation and political activity. The two forms of investment may be substitutes, complements, or have no significant relationship. In this context, investments are defined as substitutes if a marginal change in the use of one corresponds with a change in the opposite direction of the other, and are defined as complements if a marginal change in the use of one corresponds with a change in the same direction in the other.

The relationship between the two forms of investment generally depends on the relationship between marginal returns to each form of investment, in the same vein as the relationship between labor and capital in a standard profit maximization situation. If there is an increase in the relative return to technological innovation, intuition suggests that firms are likely to invest more in R&D. What is less immediately apparent is that such an exogenous change in the technological environment may alter the relative return, and hence optimal investment levels, in the political environment. A firm may increase its investment in political action in response to increased profitability in that environment and concurrently lower its investment in technological innovation because the expected change in the political environment makes investing in the technological environment less attractive or less necessary. Likewise, increased or decreased profitability in technological innovation could lead to the attractiveness of investment in political action to move in the opposite direction.

Numerous economic theories suggest that returns to technological innovation are not static over time, so that alternative investments such as political activity may be attractive substitutes [3; 6; 10; 11; 15; 17; 22; 27; 28]. For example, in the context of Schumpetarian "waves" of technological innovation, if a firm finds itself moving into an innovation "trough," marginal returns to political action may become greater, so that firm managers turn to political action as a means of achieving advantage.

Alternatively, technological innovation and political action could be complements, at least for some firms or industries. In particular, the political/regulatory environment may lag behind technical innovation, so that firms or industries that invest heavily in technological innovation may find it necessary or beneficial to make political investments to permit the introduction of new technology. In such a situation, firms' political contributions may simply "buy access" to politicians, and/or they may contain information themselves.(1) Once new technology is introduced, political action may be necessary for the firm to prevent other firms from appropriating part or all of the advantage the technological innovation creates, thus raising the marginal return to political action. A third explanation for complementarities is that a firm may seek to increase its sales (via political activity) to further exploit a technological gain that, say, increases per unit profits or depends on economies of scale.

Recognizing that changes in technological possibilities (and hence profitability) can alter political activity is crucial. Alterations in regulation may be a necessary condition for a firm to realize profits from technological innovation. Any event that impacts profitability in one environment can also impact the other; complete analysis of policy that affects one environment includes analysis in terms of the policy's effects on the other environment. In particular, if firm political activity is a necessary component of technological innovation, then policy designed to increase R&D investment may also have the effect on encouraging increases in PAC expenditures, an activity many citizens think should be limited. Conversely, limits on firm political investment may have a chilling effect on firm R&D expenditures.

Numerous researchers in several disciplines have examined firm PAC expenditures.(2) Zardkoohi, Munger, and Grier, Munger and Roberts (hereafter referred to as GMR) offer recent economics-oriented studies that investigate the determinants of firm PAC contributions [31; 23; 13; 14]. Regardless of disciplinary origin, most studies focus on or pay considerable attention to a firm's and/or industry's ability to overcome collective action problems, if even to refute the importance of such problems.(3) None of these studies have considered the influence of finn technological innovation on finn political innovation or included R&D expenditures as an explanatory variable in their empirical analysis.

The few studies that consider technological innovation and political action jointly are in the context of environmental regulation. Research by Downing and by Abbott and Brady offer models in this context, and suggest the possibility of complementarity between firm technological innovation and political action [8; 1; 2]. In most contexts, the firm has a natural incentive to adopt new technology. However, firms subject to environmental regulation may be restricted by the relevant regulatory agency as to the technology it can use to meet pollution emission goals. If cost-reducing technology is available, the firm must lobby the regulatory agency to permit use of the new technology. Presumably there will be economic actors with interests in maintaining use of the current technology, so the firm's lobbying is met with resistance by those actors.

III. Empirical Modelling

Taylor [30] provides a firm-level theoretical model that suggests that firms determine optimal technological innovation and political action investment levels jointly, even when one form of investment does not affect the probability of outcomes in the other environment, and establishes conditions under which the investments will be substitutes or complements. The theoretical demonstration of the joint determination of technological innovation and political action investment levels is empirically testable. It is also possible to investigate the nature of the relationship between the two forms of investment to see whether they behave predominantly as complements, substitutes, or exhibit no significant relationship across firms.

Empirical modelling is complicated by significant censoring of the PAC contributions variable. Ignoring this feature of the data produces inconsistent parameter estimates, including those parameters used in testing for endogeneity. Because of this, the primary empirical model used below is of the form suggested by Smith and Blundell [29].(4) The authors develop both a test for endogeneity and an efficient method of estimation in the context of simultaneity and censoring. I also employ a secondary empirical model, of the form suggested by Nelson and Olson, that involves less efficient estimation than that used by Smith and Blundell, but readily permits the use of dummy variables to identify industry and group specific relationships between endogenous variables [24; 29].

Smith and Blundell formulate a two-equation simultaneous system in which the dependent variable in the first equation is censored [29]. The first equation is structural, while the second is in reduced form. The simultaneous system for firm i(i = 1, ... H) is given as

[Mathematical Expression Omitted] (1)

[y.sub.2i] = [x[prime].sub.i][[Pi].sub.2] + [v.sub.2i]. (2)

The endogenous variable [y.sub.2i] is continuous, whereas [y.sub.1i] is limited and takes on the value

[Mathematical Expression Omitted], (3)

0 otherwise. In the case at hand, [y.sub.1] is firm PAC contributions, which is highly censored, and [y.sub.2] is firm R&D expenditure. The error terms [[Mu].sub.1i] and [v.sub.2i] are jointly normally distributed,

[Mathematical Expression Omitted] (4)

and

[x[prime].sub.i] = ([x[prime].sub.1i],[x[prime].sub.2i]) (5)

is a 1 x K([equivalent to] [K.sub.1] + [K.sub.2]) vector of exogenous variables that exactly identify both equations.

If firms do determine their technological innovation and political action investments jointly, then R&D spending should be an endogenous determinant of PAC expenditures. The test for endogeneity is developed by writing [[Mu].sub.1i] conditional on [v.sub.2i] as

[[Mu].sub.1i] = [v.sub.2i][Alpha] + [[Epsilon].sub.1i]. (6)

Substituting [[Mu].sub.1i] into (1) produces the conditional model

[Mathematical Expression Omitted]. (7)

A test of [H.sub.0] : [Alpha] = 0 constitutes a test for exogeneity of [y.sub.2i]; specifically, if the hypothesis that [Alpha] is equal to zero is not rejected, then the hypothesis that R&D expenditures are an exogenous determinant of PAC contributions is not rejected.(5) The test is implemented by estimating (2) via OLS and then using the residuals from that estimation as estimates of [v.sub.2i]. Equation (7) is then estimated via a standard censored regression. If the coefficient on the residual term is significant, then exogeneity is rejected. Should this occur, equations (1) and (2) are estimated jointly. The log likelihood for this specification is

ln L = [summation over i] {[[Xi].sub.i]ln([Phi]([[Epsilon].sub.1i],[[Sigma].sub.11.2])) + (1 - [[Xi].sub.i]) ln(1 - [Phi]([[Epsilon].sub.1i],[[Sigma].sub.11.2])) - .5ln([summation over j] [v.sub.2j][v[prime].sub.2j]/H)} (8)

where [[Sigma].sub.11.2] is the conditional variance of [[Epsilon].sub.1i], given by

[Mathematical Expression Omitted] (9)

and [[Xi].sub.i] is a binary variable corresponding with (3). The coefficient on R&D expenditures in the equation for firm PAC spending in (1) is of primary interest in determining the relationship the two forms of investment.(6)

An interesting question is whether or not different industries or groups of industries (e.g., those subject to considerable environmental regulation) exhibit different relationships in their technological innovation and political action investments. Using dummy variables is a standard method of exploring this kind of question, but the Smith and Blundell model does not readily lend itself to the use of dummy variables [29].(7) Instead, I use a two-stage least squares (2SLS) model suggested by Nelson and Olson [24]. This model is also described by (1) through (5), but instead of using the conditional model given by (7), estimation is by a 2SLS method. Fitted values for [y.sub.2i] are obtained using (2) and then interacted with industry-specific dummy variables. Equation (1) is then estimated, using a standard censored regression likelihood function and including the interactive terms, so that the use of (8) is avoided. In the interest of conserving degrees of freedom, the industry/industry group dummies are not used interactively with other variables.

IV. Data

Previous studies of both R&D and PAC determinants use various measures of firm size, firm financial position, and industry concentration; many also incorporate unique variables such as degree of government regulation, percentage of industry output that is not final output, and industry standard deviations of certain variables. The data used to estimate the empirical models in this study consist of information on 2383 firms.(8) Most of the variables are generally measures of firm size and/or "health" (e.g., net profit margin, debt). Data are from Compustat, the Federal Election Commission (FEC), and the Census of Manufacturers; most of the firms are manufacturers. Eighty nine percent of the firms in the data set do not have PACs, so their PAC spending is treated as equal to zero.(9)

Many of the variables have ambiguous expected relationships with PAC contributions and R&D expenditures. It is not clear that if a variable's value for a particular firm is, say, low relative to other firms (e.g., low return on assets), this prevents the firm from making R&D and/or PAC expenditures, or encourages the firm to make these investments in hopes of improving its position. Beyond intuition, theory, and past research, the primary criteria for specification is significance, as measured by the t-statistic. Each variable in the data set is discussed below.

Sales is primarily a control variable for the size of the firm, although it also serves to measure the general health of the firm. Firms with large sales relative to otherwise similar firms should be in a better position to invest in technological innovation and political activity, so that the expected relationship is positive for both PAC contributions and R&D expenditures.

Firms with higher debt may have less discretionary investment. High debt, however, may also compel firms to seek political rents and/or innovation in order to improve their debt situation. Furthermore, while firms with heavy debt loads may seem to be in a weaker position relative to other firms, debt may also be a measure of confidence in the firm; the same argument can be made for changes in debt from year to year. Thus, the expected relationship between debt, and differences in debt, and R&D and PAC expenditures is ambiguous.

Concentration is the focus of most of the research regarding PAC contributions and much of the research regarding R&D expenditures. More concentrated industries should be better at solving free-rider problems that may otherwise limit political cooperation, and thus in this regard should be more politically active. Zardkoohi points out that, on the other hand, more concentrated industries may rent seek to a lesser extent because they already have a degree of market power [31, 808]. The key finding of GMR [13] results from the significance of a quadratic concentration term, indicating that the effect of industry concentration on firm and industry PAC contributions varies over concentration. Researchers have found a variety of relationships between concentration and R&D expenditures.

The number of employees a firm uses is primarily a proxy for firm size. Employees are an important source of PAC contributions, but the variable includes all employees, while firm PAC contributions are more likely to come from white-collar workers. Most unions have PACs, so blue-collar workers represented by a union may be more likely to contribute to their union PAC. Furthermore, a firm with many employees represents more votes for the politician, so that, to the extent that a firm can sway the votes of its employees, firms with many employees may find PAC contributions unnecessary because it has political influence from the votes it represents. Thus, the relationship between employees and PAC expenditures is ambiguous. On the other hand, the variable may be a measure of emphasis toward or away from labor (as an input) for a firm relative to similarly-sized firms. Highly innovative firms are likely to be labor-intensive relative to their use of capital. The employees variable is therefore used as a restriction to identify (1).

The number of business segments in which a firm is involved is included as a measure of firm diversification, and was used by GMR [13]. More diverse firms would be expected to have more (in quantity) political interests, but may be too diverse to focus on (or care about) a particular issue. The information problems of more diverse firms cited in the technology literature may also apply to investments in political activity. McEachern and Romeo study the relationship between innovative activity and product diversification [20]. They find an important distinction between in-product line diversification and out-of-product line diversification, finding a significant positive relationship between in-product line diversification and R&D intensity, while the relationship between out-of-product line diversification and R&D intensity is dependent on the technological opportunity in the new product lines.(10) Given that at this point I cannot distinguish between in-product line and out-of-product line diversification, I do not expect product diversity to be a significant predictor of R&D expenditures, so that product diversity is used to identify the structural equation for R&D expenditures.

Stockholder equity is a measure of firm size, the degree to which the firm is controlled by stockholders, and the degree to which it is equity financed. Stockholders also represent a source of both PAC funds and political influence, which may be a substitute for PAC funds. These possibilities make the expected relationship between stockholder equity and firm PAC expenditures to be ambiguous. Based on the literature regarding management structure and firm PAC expenditures, I would expect a positive relationship between stockholder equity and firm R&D expenditures, as increased stockholder equity may force management to seek higher returns than they might otherwise.(11)

Retained earnings are a measure of a firm's general health and its ability to finance both PAC contributions and R&D expenditures. While both forms of investment can be considered as business expenses and need not be financed out of retained earnings, retained earnings should be a measure of the firm's ability to engage in investment. Firms with greater retained earnings would be expected to invest more on both technological innovation and political action. As with other variables, however, it may be that firms with low retained earnings may be more motivated to seek rents via technological innovation and/or political action.

Net profit margin is another firm characteristic that has an ambiguous expected relationship with firm investments in both technological innovation and political action. The Schumpeterian perspective suggests that more profitable firms should be able to invest more, in either R&D or PAC expenditures. Again, however, firms with low profits may have greater motivation to invest, in order to improve their position.

V. Estimation Results

Estimations of (7) using two different specifications are provided in Table I, with standard errors in parentheses. Specification A excludes the sales squared, stockholder equity, retained earnings, and net profit margin terms. Specification B includes these terms.(12) In both specifications, the estimate of interest is the coefficient on [v.sub.2i]. A significant estimate leads to the rejection of the hypothesis of exogeneity of R&D as a determinant of PAC.(13) The hypothesis is rejected for both specifications. This finding supports the hypothesis that firms jointly determine their technological innovation and political action investments.

Tables II and III show estimates for three different specifications of (1) and (2), the equations that form the simultaneous model.(14) The reported specifications are based on determining the influences of squared terms of sales and concentration, since, as noted above, previous studies found that they can affect both sign and significance of their non-squared counterparts. Additionally, most previous studies on the determinants of firm PAC spending do not include stockholder [TABULAR DATA FOR TABLE I OMITTED] equity, retained earnings, and net profit margin as regressors, so their effects are also studied in the reported specifications.

The coefficients of primary interest are those on R&D. For all specifications, the coefficient on R&D is positive and significant, suggesting that many firms treat investments in technological innovation and political action as complements.(15) This finding does not explain why the two forms of investment behave as complements; for example, it is possible that highly technologically innovative firms also happen to rent seek. Nevertheless, the joint determination and the positive relationship between R&D expenditures and PAC contributions do not contradict the hypothesis [TABULAR DATA FOR TABLE II OMITTED] that firm political activity is a necessary component of technological innovation. From a certain perspective, the finding is not surprising. Firms are likely to have information that is valuable, in the sense that the information can make regulation more efficient, and most technological innovation requires some form of regulatory approval. Consider an economy in which firms were prohibited from making political contributions, yet still had a myriad of concerns to communicate to the politician/regulator. All firms would "look" the same, in that they all would present apparently important interests for the attention of the regulator, whose time and other resources are limited. Regardless of the interpretation, the finding does suggest that policy to limit PAC spending should take into account potential effects on other firm activities, including investment in technological innovation.

The estimate of the coefficients on sales in Specification A is negative but insignificant, while the coefficients are significant and positive for Specifications B and C, which both have squared sales as variables. This suggests a quadratic relationship between sales and PAC contributions. The relationship changes from positive to negative around sales of $37 million, well below the mean of sales, which is $1.369 billion. However, of the 1889 firms in the sample, 765 firms have [TABULAR DATA FOR TABLE III OMITTED] sales less than $39 million, while 1124 firms have sales greater than $39 million, so the relationship is positive for a significant range of firms. This finding suggests that smaller firms find political action necessary, while larger firms find it less necessary. A possible explanation for this finding is that larger firms have considerable political power independent of their PAC contributions, perhaps because they can marshall the votes of more employees and stockholders, and thus need less PAC-derived political influence. Another possibility is that smaller firms are better able to solve internal free-rider problems, and thus are more capable of raising PAC funds.

Estimates of the coefficient on debt suggest that firms with higher debt relative to similar firms are likely to invest heavier in political action. These findings support those of GMR and would seem to indicate that the need to improve firm position (via political action) outweighs the difficulties that firms in poor position might have in raising funds for political activities [13]. The inclusion of the squared sales variable improves the explanatory power of debt. The coefficient on difdebt, however, implies that firms with a worsening debt position (indicated by a positive value of debtdif) relative to other firms are likely to invest less heavily in political activity. The explanatory power of this variable is also improved by the inclusion of the squared sales term.

The positive and significant coefficients on concentration in Specifications A and B suggests that firms in less concentrated industries are less likely to invest relatively heavily in political action. A quadratic concentration term does not have significant explanatory power, and lowers the significance of the linear concentration term. This finding supports those of Pittman and GMR and is contrary to the results in Zardkoohi and Munger [25; 13; 31; 23]. Zardkoohi ignored firms without PACs, and GMR aggregated to the industry level, methodologies which may have affected their results [31; 13]. When one considers the nature of firm political contributions, it is easy to imagine why the free-rider problem is difficult to overcome, so that firms in less concentrated industries are reluctant to become politically involved. Executives at General Motors know that a great portion of their political efforts will be accrued by General Motors, and likely have little trouble enlisting the aid of their counterparts at Ford and Chrysler. Firms in less concentrated industries may feel that they have little chance of making a political difference, and may face difficulty in enlisting the aid of other firms in the industry.(16)

The positive and significant estimates of the coefficient on business segments in all three specifications implies that firms with a more diverse product line are likely to invest heavier in political action relative to firms with a less diverse product line. This finding is consistent with that of GMR [13] and counter to that of GMR [14]. It suggests that the broader interests of more diverse firms translates into greater political activity, and that the firms are able to overcome the informational problems that previous researchers have suggested act as a hindrance on investing in technological innovation.

The coefficient estimate for stockholder equity is negative and significant in the specifications in which is included. This suggests that firms with greater stockholder control and greater equity financing are less politically active. To the extent that political action is a less risky investment and enters more positively into management utility than does investment in technological innovation, this finding supports those of McEachern and Romeo and Baysinger and Hoskisson, both of which associated R&D expenditures with more conservative and direct management [20; 4]. The coefficient sign is also consistent with the positive relationship between debt and PAC expenditures. Firms with greater debt financing are more politically active, while firms with relatively more equity financing are less politically active.

The coefficient estimates on both retained earnings and net profit margin are positive and significant, suggesting that "healthier" firms are more likely to be politically active. More politically active firms are those that have the resources for the requisite political action, not those firms with low retained earnings and net profit margins that may be more desperate for political favors.

Estimates of (1) using the 2SLS procedure suggested by Nelson and Olson are provided in Table IV [24].(17) The estimation produces results similar to those of the Smith and Blundell MLE model, with minor losses in efficiency [29].

Table V provides estimates from equations of the form of (1) using Specification B, but with the inclusion of industry group dummy variables and an interactive variable derived by multiplying the dummy variable by the R&D variable. The groups are based on those industries that may be highly innovative and/or subject to considerable regulation. Firms not in any of these groupings comprise the group of firms whose R&D observation is not interacted with a dummy [TABULAR DATA FOR TABLE IV OMITTED] variable. Coefficients on interactive terms measure the degree to which the represented group differs from the data set as a whole.(18) The findings indicate that certain groups of industries exhibit an even more positive relationship between their investments in technological innovation and political activity. No group subject to rapid technological innovation and/or regulation exhibits a significantly negative relationship between the two forms of investment.

The different findings for the Food and Drug Administration Regulation and the Drugs, Medical Supplies, and Medical Services is an interesting and potentially insightful result. The two groups have some firms in common, but the former is based on regulation, while the latter is based on technological innovation and regulatory approval of that innovation.(19) The dissimilarity of results between the two groups seems to lend further support to the hypothesis that firm political activity can be a necessary part of technological innovation. The group that relies on technological innovation perhaps more so than any other industry or industry group has a very powerful relationship between its investments in technological innovation and political activity. On the other hand, a group that has some firms in common, yet is more oriented toward regulation alone, shows no significant incremental relationship.(20) The other group based more on technological innovation, Telecommunications and Computers, exhibits a relationship similar to that of the Drugs, Medical Supplies, and Medical Services group.

Table V. Industry Group Analysis Estimates

                                                Coefficient Estimate
Variable                                          (Standard Error)

R&D (fitted value)                                    .3307
                                                     (.1137)(**)

R&D * Environmental Regulation dummy (mining,         .2290
chemicals, petroleum and coal, plastics,             (.1113)(**)
paper, lumber and wood, stone/clay/glass,
primary metals)

R&D * Telecommunications and Computers dummy          .9500
(communications equipment, communications,           (.2310)(**)
computer and data processing services)

R&D * Food and Drug Administration Regulation        -.1438
dummy (food, tobacco, drugs)                         (.1537)

R&D * Transportation dummy (transportation            .3678
equipment, trucking and warehousing,                 (.1336)(**)
wholesale motor vehicles, parts and supplies,
automotive dealers and service stations)

R&D * Transportation and Petroleum dummy (the        -.05462
above, plus oil and gas extraction, petroleum        (.1119)
and coal)

R&D * Drugs, Medical Supplies, and Health             .7801
Services dummy (drugs, medical instruments           (.2603)(**)
and supplies, opthalmic supplies, health
services)

R&D * Wood Products dummy (general building           .5497
contractors, special trade contractors,              (.2597)(**)
lumber and wood products, paper and allied
products, printing and publishing, lumber and
construction materials, wholesale paper and
paper products, retail lumber)

Observations                                           1887

Log Likelihood                                        -1592.9

* p [less than or equal to].05 ** p [less than or equal to] .01

Table VI shows the estimates from a similar version of (1); in this case, the industries listed are specific 2 digit Standard Industrial Classification (SIC) groups, with the exceptions that the petroleum and coal industries and the rail and trucking industries are combined. These estimates are from a single equation, with multiple interactive (and intercept) terms. Like the more tailored groupings above, they are singled out because they may be more technologically innovative or subject to considerable regulation. Firms not in any of these industries comprise the group of firms whose R&D observation is not interacted with a dummy variable.

Fabricated Metals shows a positive and significant incremental relationship between investments in technological innovation and political activity, while Machinery and Equipment, Electrical Equipment, and Petroleum and Coal exhibit relatively negative and significant incremental relationships. However, in these latter groups, the magnitude of the (negative) coefficients on the interactive terms are less than the magnitude of the (positive) coefficient on R&D, so that the overall relationship is still positive. Because the classifications are less specific than those provided in Table V, the results are more difficult to interpret. Why firms in Fabricated Metals have an incrementally more positive relationship, while those in Machinery and Equipment have a less positive relationship, is not apparent. The findings do indicate that the relationship between firm investments in technological innovation and political action differ considerably across industries.

Table VI. SIC Industry Analysis Estimates

                                            Coefficient Estimate
Variable                                      (Standard Error)

R&D (fitted value)                                .3423
                                                 (.1162)(**)

Railroads and Trucking                           3.288
                                                (2.289)

Fabricated Metals                                2.108
                                                (1.025)(**)

Machinery and Equipment                          -.2613
                                                 (.0998)(**)

Electrical Equipment                             -.2880
                                                 (.0977)(**)

Chemicals                                        -.1854
                                                 (.1243)

Instruments and Related Products                 -.1806
                                                 (.1671)

Petroleum and Coal                               -.2619
                                                 (.1182)(**)

Observations                                       1889

Log Likelihood                                    -1614.7

* p [less than or equal to] .05 ** p [less than or equal to] .01

VI. Conclusion and Extensions

This paper explores the relationship between firm investments in technological innovation and political action. After discussing this relationship, previous research, and available data, I use several estimations to explore empirically the possible relationships. My findings indicate that firms do consider the investments jointly, indicated by the endogeneity of R&D as a determinant of PAC expenditures. Furthermore, I provide evidence that the two are complements, raising the possibility that political action is a necessary component of technological innovation. This possibility is further supported by my analysis of the relationship over several industries and industry groups.

There are numerous ways in which this line of research could potentially be extended; two areas stand out in particular. First, cooperation between firms in both the technological and political environments is likely. Despite the long-standing interest within the discipline of Industrial Organization in the effects of cooperation and the free-riding problem, relatively little attention has been given to cooperative efforts in the realm of firm political activity. Cooperation in the form of production and pricing is illegal and hence difficult to observe, but political cooperation is not. Incorporating cooperation, particularly the role of trade associations, into the analytical framework provided here would be a very useful extension. Secondly, the question remains as to why there appears to be a complementary relationship between firm R&D and PAC expenditures. A direct approach, although one with inherent limitations, would be to ask those in industry about the nature of the relationship. Anecdotally, firm leaders object to limitations on firm political activity, but they rarely detail the negative effects they claim such limitations would have; if asked, they may be willing to identify explicitly, or dispel the notion of, a chilling effect that PAC contribution limitations might have on firm R&D investment. Another approach would be a specific industry or case study. Given the social importance of technological innovation and the suspicion with which many hold firm political activity, this is an important issue to resolve.

1. Sabato suggests and discusses the former [26, 92]. Kolpin and Taylor explore the role of political contributions as signals [16].

2. Grier, Munger, and Roberts (hereafter referred to as GMR) offers a thorough literature review of material published in economics, political science, and sociology sources [14]. I am indebted to an anonymous referee for bringing this and several additional sources [5; 19; 21] outside of the economics literature to my attention.

3. Boies, for example, rejects the importance of the free-rider problem in favor of individual firm relationships as the dominant force behind firm political action [5]. Research by Mizruchi is a notable exception to the focus on collective action [21]. That work examines the influence of interlocking boards of directors on corporate political activity.

4. GMR point out that the censored "Tobit" specification has the implicit restriction of forcing the qualitative (whether or not a firm has a PAC) and quantitative (how much the PAC spends) decisions by the firm to be explained by the same determinants [14]. A more general specification is developed by Cragg, the test for which is described in Greene [7; 12,700-701]. However, the effects of simultaneity on the specification and the test for it have not been studied, and are therefore not used here. GMR find that a more general specification does not materially affect the quality of their particular results, and there does not appear to be any theoretical or intuitive reason to suggest that the qualitative and quantitative relationships would be significantly different in the study at hand [14].

5. See Smith and Blundell for full details [29].

6. Because (2) is a reduced form equation, its coefficients are not immediately interpretable in terms of establishing the relationship between firm investment in technological innovation and political activity. While one can solve for the structural coefficients, their use is hindered by the difficulty of calculating associated standard errors for the purposes of inference and hypothesis testing.

7. Difficulty arises because the estimates of [v.sub.2i] (from (2) and used in the likelihood function given by (8)) are affected by all observations, even if one were to use dummy variables in (2); industry specific R&D effects from (2) cannot be incorporated into (1) when (8) is used for the estimation. As a result, the MLE model cannot be used to estimate inter-industry differences.

8. Many observations have missing data, so that the number of observations used in estimations range from 1354 to 1914.

9. The 1987-1988 election cycle was the most recent for which data were available when this study was initiated; PAC expenditure figures for each finn are for this entire cycle. Other data are annual values from 1988; 1987 data are used for differenced variables.

10. Many studies create measures of "technological opportunity," such as industry average R&D intensity.

11. See for example McEachern and Romeo and Baysinger and Hoskisson [20; 4].

12. These specifications of (7) correspond with specifications of (1) and (2) presented and explained below, The significance of the coefficient estimate on [v.sub.2i] is generally invariant to minor changes in specification.

13. Equation (7) does not estimate the other parameters jointly with those from (2), so these parameter estimates do not have as much interpretive value as subsequent estimations.

14. PAC expenditure values were scaled by 1000 to aid in convergence.

15. In a censored regression, the magnitude of the coefficient estimates is not immediately interpretable as the marginal effect on the observed dependent variable of changes in the regressors variables. The coefficients represent the marginal effect of the latent, censored dependent variable. Maddala offers a full explication of the various versions of marginal effects that may be calculated [18, 159-60]. Of importance in their use here, all versions of marginal effects have the same sign as the estimated coefficient.

16. This discussion begs the question of the role of trade associations in industry political activity. This topic is discussed in section VI.

17. Since Specification C does not provide results significantly different from those of Specification B, the estimates from the former are not presented. Because estimates of (2) are not estimated jointly, they are not available.

18. For purposes of exposition, only the coefficients on the interactive term are presented.

19. Excluding medical services from the Drugs, Medical Supplies, and Medical Services group does not have a significant impact on the group's estimates.

20. I have no measure of technological opportunity, but I would think that food and tobacco are not extremely technologically innovative industries.

References

1. Abbott, Alden F. and Gorden L. Brady, "Tollison Costs and Technological Innovation: The Case of Environmental Regulation." Public Choice, May 1990, 157-65.

2. -----, "Welfare Gains from Innovation-induced Rent-seeking." Cato Journal, Spring/Summer 1991, 89-97.

3. Ayres, Robert, "Barriers and Breakthroughs: An 'Expanding Frontiers' Model of the Technology-Industry Life Cycle." Technovation, May 1988, 87-115.

4. Baysinger, Borly and Robert E. Hoskisson, "Diversification Strategy and R&D Intensity in Multiproduct Firms." Academy of Management Journal. June 1989, 310-32.

5. Boies, John L., "Money, Business, and the State: Material Interests of Fortune 500 Corporations, and the Size of Political Action Committees." American Sociological Review, October 1989, 821-33.

6. Cohen, Wesley M. and Steven Clepper, "The Anatomy of Industry R&D Intensity Distributions." American Economic Review, September 1992, 773-79.

7. Cragg, John G., "Some Statistical Methods for Limited Dependent Variables with Application to the Demand for Durable Goods." Econometrica, September 1971, 829-44.

8. Downing, Paul B., "A Political Economy Model of Implementing Pollution Laws." Journal of Environmental Economics and Management, September 1981, 255-71.

9. Federal Election Commission. Campaign Expenditures in the United States, 1987-1988: Freedom of Information Act (FOIA) Data [computer file]. 2nd release, 1990. Washington, D.C.: Federal Election Commission (producer). Ann Arbor, Michigan: InterUniversity Consortium for Political and Social Research (distributor).

10. Gort, Michael and Richard A. Wall, "The Evolution of Technologies and Investment in Innovation." Economic Journal, September 1986, 741-57.

11. Grabowski, Henry G. and Dennis C. Mueller. "Life-Cycle Effects on Corporate Returns on Retentions." Review of Economics and Statistics, November 1975, 400-16.

12. Greene, William H. Econometric Analysis. 2nd Ed. Englewood Cliffs, N.J.: Prentice-Hall Inc., 1976, pp. 700-701.

13. Grief, Kevin B., Michael C. Munger, and Brian E. Roberts, "The Industrial Organization of Corporate Political Participation." Southern Economic Journal, January 1991, 727-38.

14. -----, "The Determinants of Industry Political Activity, 1978-1986." American Political Science Review, December 1994, 911-26.

15. Kleinknecht, Alfred, "Are There Schumpetarian Waves of Innovation?" Cambridge Journal of Economics, March 1990, 81-92.

16. Kolpin, Van and Davis F. Taylor. "The Information Content of Political Contributions: Effects on Social Welfare." Working paper, University of Oregon, 1995.

17. Kuznets, Simon S. Secular Movements in Production and Prices. Boston: Houghton Mifflin, 1930.

18. Maddala, G. S. Limited-Dependent and Qualitative Variables in Econometrics. Cambridge: Cambridge University Press, 1983, pp. 159-60.

19. Master, Marick F. and Gerald D. Keim, "Determinants of PAC Participation among Large Corporations." Journal of Politics, November 1985, 1158-73.

20. McEachern, William H. and Anthony A. Romeo, "Stockholder Control, Uncertainty, and the Allocation of Resources to Research and Development." Journal of Industrial Economics, June 1978, 349-61.

21. Mizruchi, Mark S. The Structure of Corporate Political Action: Interfirm Relations and Their Consequences. Cambridge, Mass.: Harvard University Press, 1992.

22. Mueller, Dennis C., "A Life Cycle Theory of the Firm." Journal of Industrial Economics, July 1972, 199-219.

23. Munger, Michael C., "On the Political Participation of the Firm in the Electoral Process: An Update." Public Choice, March 1988, 295-98.

24. Nelson, Forrest and Lawrence Olson, "Specification and Estimation of a Simultaneous Equation Model with Limited Dependent Variables." International Economic Review, October 1978, 695-709.

25. Pittman, Russel, "Market Structure and Campaign Contributions." Public Choice, Fall 1977, 37-52.

26. Sabato, Larry. PAC Power: Inside the World of Political Action Committees. London: Norton, 1984, p. 94.

27. Schmookler, Jacob. Invention and Economic Growth. Cambridge: Harvard University Press, 1966.

28. Schumpeter, Joseph A. Capitalism, Socialism, and Democracy. New York: Harper, 1942.

29. Smith, Richard J. and Richard W. Blundell, "An Exogeneity Test for Simultaneous Equation Tobit Model with an Application to Labor Supply." Econometrica, May 1986, 679-85.

30. Taylor, Davis F. "Technological Innovation, Political Action, and Strategic Behavior of the Firm." Ph.D. dissertation, University of Oregon, 1995.

31. Zardkoohi, Asghar. "On the Political Participation of the Firm in the Electoral Process." Southern Economic Journal, January 1985, 804-17.