Editorial Manager(tm) for American Journal of Political Science Manuscript Draft

Transcription

1 Editorial Manager(tm) for American Journal of Political Science Manuscript Draft Manuscript Number: AJPS Title: #33423(Resubmission) WHO LOBBIES WHOM? Electoral Systems and Organized Interests' Choice of Bureaucrats vs. Politicians in Japan Article Type: Article Keywords: lobbying; interest groups; electoral systems; legislators; bureaucrats; Japan Corresponding Author: Megumi Naoi, Ph.D. Corresponding Author's Institution: University of California, San Diego First Author: Megumi Naoi, Ph.D. Order of Authors: Megumi Naoi, Ph.D.; Ellis Krauss, Ph.D. Abstract: How do interest groups choose across different venues of lobbying to influence policy? Why do some interest groups lobby politicians and others lobby bureaucrats? We theorize lobbying as an organized interests' effort to form and enforce a contract with policy-makers, politicians or bureaucrats. We argue that organizational structures of interest groups--whether they are centralized or decentralized--substantially affect their lobbying strategies because they are associated with different ability to monitor and enforce contracts with policy-makers and punish them when they fail. Different types of electoral systems, we further argue, affect the effectiveness of various instruments of monitoring and punishment (votes, political funds, and candidate endorsement) and hence interest groups' lobbying choices. We test these arguments with the case of a major democracy which recently went through an electoral reform, Japan. We use a unique longitudinal survey data on lobbying which spans two decades, covers around 250 organized interest groups in various sectors and issue areas. The results lend strong support to our organizational structure argument and the effect of electoral reform on interest groups'lobbying tactics.

2 Suggested Reviewers: Opposed Reviewers:

3 Manuscript ( Not to include ANY author-identifying information) Click here to download Manuscript ( Not to include ANY author-identifying information): AJPS_Anonymous_Sep2_2008.pdf Revision Memo Dear Reviewers and the Editor: We truly appreciate excellent comments and suggestions we have received on our manuscript, Who Lobbies Whom? Electoral Systems and Organized Interests Choice of Bureaucrats vs. Politicians in Japan. The reviews are the best kind we have had: intelligent, thoughtful, and constructive. In the revised version, we believe we have addressed all the points raised by reviewers and the editor. The editor asked us to respond to four areas, but we have incorporated these into three (for e.g., by including areas 3 and 4 of her revision instructions into the Empirics section). Below, we organize our responses to (1) Theory, (2) Empirics, (3) Methodology and (4) Others (e.g., styles and presentations) and mark each response with [R1][R2][R3][R4][R5] on the left corresponding with a point raised by each of the five reviewers. Before proceeding further, we would like to raise one key point. There are two opposing suggestions by [R1] and [R4] about whether we should refocus our paper on how interest groups adopt to political institutional change and jettison the venue choice question (R1) or focus on the venue choice question and underplay the electoral reform effects (R4). Both are excellent suggestions. Three other reviewers seem to find both venue choice question and the effect of electoral reform on lobbying strategies to be important contribution of our paper (R2 and R3 find the electoral system angle to be interesting, R5 states explicitly that both a venue choice and electoral system angles are important contribution). We have asked the editor for guidance as to which we should weigh more, as it substantially changes the direction of our revision. However, the editor seems to wish to leave this choice to us as we did not receive guidance on this question. After careful reading, we believe that R1 s comment is due to the fact that we did not make it clear that the two questions (the venue choice and the effect of electoral reform) are systematically related: the electoral systems affect the contract enforceability with politicians and the contract enforceability accounts for the venue choices. Thus, in the revised manuscript, we made it clear that the two questions are systematically related (explained in detail later). We then address R1 s comment to focus on the effect of electoral reform on lobbying strategies by (i) detailing how the policy-making environments under MMD and SMD/PR systems shaped lobbying strategies, and (ii) 1

4 developing two-by-two framework (Table 1) for understanding how legislators electoral incentives shape interest groups lobbying strategies. We also address the R1 s point that we lack a theory to account for why interest groups lobby bureaucrats. In particular, R1 is concerned that the venue choice question seems trivial in light of a preference-centered approach because in a strong party, parliamentary regime like Japan s (as De Figueiredo argues (2002 APSR)), preference of politicians and bureaucrats should be similar. We agree that if we take a preferencebased approach, the venue choice question is trivial. The question is not trivial, however, if we consider the venue choice as a function of contract enforceability the most serious neglect in the preference-centered approach according to Hall and Deardorff (APSR, 2006). Our contractual approach addresses this neglect by modeling repeated personal contact between interest groups and legislators/bureaucrats who are already sympathetic to their preferred policies. Indeed, all anecdotes discussed in our paper are about special interest politicians who should have aligned interests with organized interests and corresponding ministries reneging on the contract or shirking the efforts. Even when politicians are sympathetic, contracts can break down. So the question about policy preference is irrelevant here. In the revised manuscript, we clarify our point that the venue choice question is critical to test our contractual approach to lobbying and also address R1 s concerns in three ways. First, in the literature review, we differentiated our approach more explicitly from the preference and resource-centered approach to lobbying and venue choices. We argue in page 4-9 that the venue choice question allows us to test core assumptions developed by the preference and resource-centered approach about why interest groups lobby. Simply studying interest groups intensity of lobbying in one venue i.e., legislative arena would not allow us to test these core assumptions (see de Figueiredo and de Figueiredo 2002 on how we should not study lobbying one venue in isolation from another when there exists interdependencies in group s choice of lobbying intensity across venues). Second, we also highlight several major puzzles about lobbying in general and in Japan in particular in page 4-9 and argue that the preferencebased approach cannot account for them. Finally, in the empirical section, we demonstrate the relative validity of our contractual approach on the venue choice by controlling for a number of preference and resource-based factors that existing studies found to be relevant for the venue choice. 2

5 We address R4 s concern that by simply analyzing three waves of surveys before and after the electoral reform, we won t be able to isolate the effect of electoral reform from other co-founding factors. We address this concern in two ways. First, we demonstrate that the data on venue choice allows us to isolate the effect of electoral reform from other co-founding determinants of lobbying strategies, such as recession and globalization, because the electoral reform should change lobbying patterns with politicians, while the reform alone should not change their lobbying patterns with bureaucrats. We also include the level of bureaucratic contacts as a control variable in one of our models for estimating the frequency of contact with politicians, as we could reasonably assume that the level of bureaucratic contacts might reflect the influence of unmodeled co-founding factors especially the changes in economic environment before and after the electoral reform. Even after controlling for the level of bureaucratic contact, we find that there is substantial change before and after the electoral reform in the lobbying pattern with politicians which gives a strong support to the effect of electoral reform. Second, we also show in Graph 1 in the appendix (discussed on page 26) that among 144 groups that were constantly surveyed between second and third surveys (total of 288 obs), how their frequency of contacts changed between the two surveys. This allows us to control for group-specific attributes that stayed constant or did not change much between 1994 and 2003, such as sector, the size of membership, and the nature of contacts and demonstrate the changes over time. By doing so, we also address R5 s comment about how our empirical analysis accurately tests our hypotheses (see II-(6)). In sum, while we focus our theory section to explaining the link among electoral systems, organizational structures of interest groups, and their lobbying strategies in detail in the revised version, we do need to keep the venue choice question for aforementioned reasons. I. THEORY (1) Policy-Making Environment under SNTV vs. Mixed Electoral Systems [R1, R3] both usefully suggested that we explain the policy-making environment under SNTV and mixed electoral systems and how this environment shaped interest groups lobbying strategies. As mentioned above, we followed this suggestion by explaining this and develop a 2x2 framework (see Table 1: centralized vs. decentralized, and 3

6 narrow vs. broad)(pp.9-11 and pp.15-16). And we proceed to discuss how policymaking environment shapes lobbying strategies under SNTV and mixed electoral systems and discuss their implications for lobbying behaviors (pp and pp.16-17). We also discuss party discipline under SNTV and the mixed electoral system and how the strength of party discipline affects groups incentives to lobby individual legislators using McGillivray s work (p.16 and footnote 14). [R4] also suggests that we parse out whether interest groups target backbenchers or party leaders under SNTV as it provides a finer test of our argument a great suggestion. Unfortunately, the surveys do not ask this question specifically. Even if we can design questionnaires, moreover, the dichotomy of backbenchers vs. party leaders gets little tricky as the PARC directors are likely to be occupied by junior to mid- career legislators whose mean numbers of terms served are three. These directors tend to serve as Parliamentary Minister (seimu-daijin or seimu-jikan) in corresponding ministries (Rosenbluth and McCubbins 1995). Thus, by including Parliamentary Minister as one of the five political positions we examine, our dependent variable captures interest groups contact with these junior to mid-tier politicians as well. (2) The Logic of Decentralizing Monitoring under SMD vs. SNTV [R3] We agree with R3 s suggestion about the logic of decentralized and centralized monitoring under the SMD/PR system and addressed these points in the revised version of Hypothesis 2 (p.17) and Table 1 in appendix. Indeed, Graph 1 presented in the appendix clearly supports R3 s point that decentralized groups with both headquarters in Tokyo and local organizations did moderately increase their contact with politicians as they continue to lobby through local organizations in SMD while increase lobbying through the headquarters, although this increase is much more moderate (18%) compared to centralized groups without local organizations (63%). (3) The Logic of Centralizing Monitoring under SNTV [R5] raised a sharp question about why centralized groups have difficulty monitoring politicians given their proximity to the Diet. This allowed us to rethink the issue of transparency in policy-making environment under SNTV. We address this question by discussing how, under one-party dominant system before 1994, the within-ldp policymaking organ (i.e., PARC) did not allow interest groups to effectively monitor politicians' positions and relative contribution to realizing policies (pp ). Also, as we explain in coding section (p.20), 94% of groups that have local organizations have 4

7 headquarters in Tokyo. Thus, decentralized groups are in general better positions to monitor politicians activities both at the center and in localities. Accordingly, [R4] also suggests that we differentiate interest group s structures into three categories: local only, local and central, and central only. Since more than 94% of groups with local organizations have headquarters in Tokyo, we address this question by including a new dummy variable, no_tokyo_hq which takes a value of one if a group has a headquarter outside of Tokyo (i.e., if no_tokyo_hq=1 & local=0, this is a localonly group. We have also interacted these two variables, but the results do not change). The table below shows the distribution of cases. This also allows us to differentiate Walker (1991) s access-based argument from our contractual argument thank you for suggesting this!. tab no_tokyo_hq localorg no_tokyo_hq localorg 0 1 Total Total The only problem of including this variable, though, is that because not many groups have headquarters outside of Tokyo vicinity (twelve groups are in this category for the third survey, only two groups for the second, and only one for the first), outliers weigh heavily in the final logit and regression results. For instance, no_tokyo_hq has strong, positive effects on the frequency and percentage of contacts with politicians as shown in Table 4 and 7 (note that results do not change if we drop no_tokyo_hq from the models). This is due to two outliers with very high contact score with politicians. We cannot reveal specific types of these groups in our paper as they should be kept anonymous, but our personal communication with the original survey team has suggested that these were large business associations organized by a broader regional block (not SMD-level, localized group) which makes perfect sense because under regionally-organized PR blocks, these associations can lobby through a regional PR block. 5

8 Cross tabulation below shows mean contact scores for 2x2 categories. Without these two outliers, the mean scores are consistent with our hypotheses & R4 s expectation as shown in table below. table no_tokyo_hq localorg, c(mean polcontact_four_s3) no_tokyo_ localorg hq (73) (140) (3) (7) Note: Frequency of observations in parenthesis. With the two outliers, the results look like below.. table no_tokyo_hq localorg, c(mean polcontact_four_s3) no_tokyo_ localorg hq (73) (140) (4) (8) Note: Frequency of observations in parenthesis. (4) Swing vs. Core Voters: McGillivray (1997; 2004) [R1] suggests that we discuss McGillivray s work on core vs. swing districts. We did so in page 16 and footnote 14. (5) Framing of Contribution and the Literature Review [R2, R5] We have located our contribution to the up-to-date literature on lobbying in the U.S. (see Theories of Lobbying Venue Choices pp.4-8). This includes an extensive discussion of Walker s work and how our approach is both similar and different from his. We also cite and/or discuss works by Wright, Hall and Deardorff, Ainsworth, 6

9 McCarty and Rothenberg, Snyder, Stratman, Furlong and Golden, Yackee and Yackee and Figueiredo and Figueiredo. II. EMPIRICS (1) Why Japan? Stress experimental design in introduction [R3, R4] Following R3 s and R4 s suggestion, on page 3 of Introduction, we underscore the quasi-experimental advantages of the Japanese case. (2) Missing membership data [R4, R2] We addressed the missing membership data by using survey questionnaires on the past membership in ten-year interval following R4 s suggestion. (see footnote 21 on page 21). As a result, the level of statistical significance for local organization variable for the first survey has improved from ten to five percent we hope this demonstrates that the results are actually substantial and will address [R2] s concern that the results are only modest especially for the first survey. (3) Industrial Structures and Organizational Structures [R3] raised the omitted variable bias issue that organizational structures of interest groups might be shaped by industry characteristics. In particular, R3 is concerned that industries dominated by small-medium size firms may exhibit particular type of organizational structure. We checked for the pairwise correlations among a group having local organization and organizations dominated by small-medium size firms. As the pairwise correlation table below suggests, they are only correlated at Thus, we do not consider this omitted variable bias to be a serious concern. We also address R3 s point more systematically by including interest group leaders response to a questionnaire on whether an organization represents large or small-medium size firms. The variable sm_industry takes a value of one if they are dominated by small-medium size firms and zero otherwise. Unfortunately, this question was asked only for the first survey. The pairwise correlation table below suggests that sm_industry is highly correlated with industry sector dummy at.67 and policy tribe variable at.66. Policy tribe (named zoku here for the space reason) and industry dummy also highly correlates at.83 for the first survey thus, we include sm_industry as a control and drop industry dummy and policy tribe variables instead. For a robustness check, we also ran the same model with the 7

10 policy tribe variable and without industry sector dummy the major results do not change as shown in [Robustness Check] in the appendix of this revision memo.. pwcorr localorg sm_industry agriculture labor economy zoku likeldp localorg sm_industry agriculture labor economy zoku likeldp localorg sm_industry agriculture labor economy zoku likeldp (4) Consult Other Data [R1] Consult other data who gives how much to whom. We were not fully clear whether this suggestion means we should go out to get campaign contribution data, or, it simply means that we should talk about the intensity of lobbying (which we do in both original & revised manuscript we use two dependent variables, one on intensity and another on the relative allocation of lobbying resources to the two venues as percentage variable see pp.19-20) as opposed to simply discussing a venue choice in a dichotomous way. We would have loved to include the campaign contribution data, however, it is extremely hard to do so as the data became only publicly available in a consistent manner since Even the available data is incomplete as it only reflects what parties and legislators get at the central office, not at local branches (see footnote 16 on page 18). Contribution to these local party branches is only available on sites for 300 electoral districts (see Sasaki et al. 1999). Moreover, the three waves of surveys used in this paper consistently show that interest group leaders consider that campaign contribution to be the last reason why legislators help them: only around 2% of respondents agree with the statement (see footnote 1 on page 5). [R4] Consult collateral data source One potential critique to using the survey data as opposed to behavioral data is that what interest group leaders report might not match their behaviors due to the reporting bias (i.e., certain interest groups might underreport their lobbying activities) and anchoring issues (e.g., their perception of frequent contacts differ among IGs). 8

11 In order to address this issue, we consulted interest group lobbying data on trade issues before the Diet committees starting in 1955 collected by Megumi Naoi and found that under the SNTV system, interest groups that have both local branches and headquarters in Tokyo are much more likely to show up to testify as locally-organized groups than sectorally-organized groups or headquarters confirming our point. See simulated effects on the LDP seat shares in prefectures on the probability of groups showing up to testify as geographically-organized groups than sectorally-organized groups under SNTV and SMD/PR presented a graph below. Simulated Effects of Electoral Competition and the Probability of Regional Lobbying: SNTV vs. SMD/PR Systems 0.2 Probability of Regional Lobbying SNTV SMD/PR The LDP Vote Share (Prefecture) Source: Naoi s trade lobbying data from 1955 to Simulation done using CLARIFY software available at The above graph suggests that under SNTV, the probability of interest groups showing up to testify as local branches, as opposed to the nationally-organized sectoral groups or headquarters, was three to four-fold higher than under SMD/PR system. This confirms our point about the importance of decentralized monitoring and enforcement by local groups under SNTV. Despite the fact that their central headquarters in Tokyo are geographically much closer to the Diet, local branches travel all the way to Tokyo to testify before the Diet committees. The effect of electoral competition on the probability of local groups showing up to testify becomes much weaker under SMD/PR. Also, the higher the LDP seat shares in prefectures, more likely that locally-organized groups show up to testify before the Diet committees over trade issues. This also makes sense in light our contractual approach because it is easier to enforce a contract with legislators with safe seats and thus a longer time-horizon (i.e., higher vote share). 9

12 We also contacted other scholars who have been collecting interest group survey data, such as Professor Yutaka Tsujinaka, but they are not ready to share the data with us at this point. (5) Generalizability beyond Japan How about the U.S.? [R2] raises a question about generalizability beyond the Japanese case and suggests that we test our theory using Walker s (1991) data available at ICPSR. Although this is an excellent suggestion, we believe that presenting the results for the U.S. case would be beyond the scope of our paper. Our goal is to demonstrate that electoral institutions and interest group organizations interact to shape lobbying patterns, and the U.S. case does not allow us to test this argument, either. We will definitely pursue this suggestion for the next paper, though, and truly appreciate the suggestion. III. METHODOLOGY (1) Testing our argument more accurately. [R5] raises a very good point that our tests were about cross-sectional differences between centralized vs. decentralized groups, although our argument and hypotheses are about both cross-sectional difference (H1, H2, and H3) and changes after the reform (H2 and H3). We address this point by focusing on 144 groups that were surveyed consistently for the 1994 and 2003 surveys and showing changes in their contact scores in Graph 1. This allows us to hold sectoral, issue areas, and organizational resources relatively constant, and show changes induced by the electoral reform. The changes are consistent with Hypothesis 2 and 3. (2) Omitted variable bias test As [R5] points out, since all the potential determinants of a group having local organizations are already included as control variables in the model estimating the frequency of contacts with politicians, we took out our discussion on the omitted variable bias test from the revised version. Following R3 s suggestion, moreover, we included a dummy for organizations dominated by small-medium size firms (see earlier discussion) to address one more potential determinant of a group having a local organization. (3) Should contact scores be aggregated? 10

13 [R5] We understood this question as whether we could aggregate an ordinal variable. We made sure that no contact is assigned zero (instead of one ) so that even when we aggregate the contact scores across different positions, aggregation of zero stays zero. Furthermore, we recode these aggregated scores into four-scale ordinal dependent variable in the revised version to address this point. IV. Minor Points Clarification, Presentation, and Styles [R2] Q1: Punishing or reward? An anecdote on page 12. Response: We clarified the point of the anecdote by adding more background to the quote and text on page 12. Q2: Why is the N in the three graphs 22? How were these particular interest groups selected from the total N of the surveys, which each included more than 200 interest groups? I did not find these graphs helpful and would recommend deleting them. In general, too many tables. Response: We deleted useless graphs and present only a few that deliver our results most clearly. [R4] Q1: More examples on monitoring tactics used by specific groups. We added a new example of construction industry keeping transcript of politicians contribution to realizing policies on page 12. Also see our discussion on the difficulty of monitoring at the center and how interest groups tried to deal with this difficulty under SNTV (p. 10 and fn11). Q2: Is Nokyo centralized or decentralized? A: See our definition of decentralized groups on page 8 and more specific discussion on Nokyo on page 13, fn.10. Q3: Agency slack and group s decisions to lobby bureaucrats A: This is a great point so we discuss this and the relevant literature (McCubbins and Schwartz 1984) when we discuss sectoral dummies in control section (page 22). 11

14 Q4: Label the surveys by years. A: We followed this suggestion. [R5] Q1: Ln_Y issue. A: We recalculated ln_y variable and turns out that we miswrote the equation in text but we did calculate ln_y in a correct manner. We corrected the equation to ln_y=ln(y/(1-y)) in text following R5 s suggestion. For groups that lobbied in only one venue, we used the following formula at Langche Zeng s suggestion: ln_y=ln(y/(1-y))=ln_y=ln( /( )) if Y=0 ; ln_y=ln(y/(1-y))=ln_y=ln( /( )) if Y=1. Q2: Proposition-Hypotheses Structure A: Changed to simple hypotheses. Q3: Can central headquarters discipline local organizations? A: This is an excellent question. Those who study organizational structures of industry and trade associations in Japan found that local organizations tend to have autonomy from the headquarters for two reasons. First, many industry associations are fiscally decentralized: local branches often raise much of their own revenues (e.g., Agricultural Cooperatives see fn.12 on page 14). Furthermore, government subsidies to these local organizations are often channeled through prefectural governments. Second, anecdotes abound that local business organizations deviated from the central order in implementing industrial policies suggested by the Ministry of International Trade and Industry because they felt the need to tailor the national policy to meet their local conditions (documented in detail in David Friedman, Misunderstood Japanese Miracle. Cornell UP). 12

15 Appendix for the Revision Memo Robustness Check: The Frequency of Contacts with Legislators (Not Controlling for Frequency of Contacts with Bureaucrats and Including vs. Excluding Policy Tribes for Survey 1) Organizational Structures Local Org (2.28)** (2.39)** (3.24)*** (0.69) No_tokyo_hq (0.00) (0.15) Organizational Resources ln_membership (0.87) (1.00) (2.36)** (0.36) Current MPs (1.13) (1.10) (0.79) (0.68) Policy Tribes (0.64) (2.38)** (0.68) AdvisoryCouncil (2.75)*** (2.72)*** (3.38)*** (2.01)** Retirement (0.29) (0.31) (0.67) (0.06) Subsidy (0.18) (0.10) (0.28) (0.93) Group Age (2.84)*** (2.82)*** (4.28)*** (1.84)* Issues and Sectors Regulation (0.45) (0.32) (1.00) (1.57) Agriculture (0.44) (0.34) (2.12)** (0.40) Labor (0.64) (0.45) (1.48) (0.27) sm_industry (0.12) (0.73) International (0.58) (0.93) Observations Absolute value of z statistics in parentheses * significant at 10%;** significant at 5%; *** significant at 1% Note: Cut points not shown. No_tokyo_hq is dropped from first two models due to collinearity. 13

16 WHO LOBBIES WHOM? Electoral Systems and Organized Interests Choice of Bureaucrats vs. Politicians in Japan Abstract How do interest groups choose across different venues of lobbying to influence policy? Why do some interest groups lobby politicians and others lobby bureaucrats? We theorize lobbying as an organized interests effort to form and enforce a contract with policy-makers, politicians or bureaucrats. We argue that organizational structures of interest groups substantially affect their choice of lobbying strategies because they are associated with different ability to monitor and enforce contracts with policy-makers and punish them when they fail. We demonstrate that whether structures are centralized or decentralized accounts for a good part of the variations in their venue choices. Different types of electoral systems, we further argue, affect the effectiveness of various instruments of monitoring and punishment (votes, political funds, and candidate endorsement) and hence interest groups choice to lobby politicians or bureaucrats. We test these arguments with the case of a major democracy which recently went through an electoral reform without major partisan change, Japan. We use a unique longitudinal survey data on lobbying which spans two decades (1980, 1994, 2003), covers around 250 organized interest groups in various sectors and issue areas. The results lend strong support to our organizational structure argument and the effect of electoral reform on interest groups choice of lobbying tactics Words (Excluding Appendix)

17 Who Lobbies Whom? P. 2 INTRODUCTION How do interest groups choose across different venues of lobbying to influence policy? Why do some interest groups lobby politicians and others lobby bureaucrats? These questions lie at the heart of much of the classic literature in political economy, ranging from Madison and the Federalist Papers and Dahl s seminal study of pluralism in New Haven (Dahl 1963) through studies of corporatism (e.g., Schmitter and Lembruch 1979) to the collective choice literature (e.g. Olson 1965). Yet the existing literature on lobbying tends to ask to what extent interest groups influence policy rather than a question of how they attempt to influence it. In particular, the question of venue selection, i.e., how interest groups choose across multiple venues of lobbying, has not been explored extensively to date. In contrast to the existing approaches that theorize interest groups goals of lobbying as changing legislators preferences or their policy-making resources (McCarty and Rothenberg 1996, Walker 1991; Hall and Deardorff 2006), we theorize lobbying as an organized interests effort to monitor and enforce a contract with policy-makers, politicians or bureaucrats. We argue that organizational structures of interest groups, in particular, whether organizational structures are centralized or decentralized, substantially affect their choice of lobbying strategies because they are associated with different ability to monitor and enforce contracts with policy-makers and punish them when they fail. We further demonstrate that the effect of centralized vs. decentralized structure on venue choices is conditional on the types of electoral system: majoritiarian, semi-proportional (single-non-transferrable-vote: SNTV), or proportional representation systems. Under a highly personalistic electoral system like SNTV, decentralized groups are more likely to go to a political route as they are better able to monitor, enforce, and punish politicians than centralized groups. Under a party-centered electoral systems, such as a

18 Who Lobbies Whom? P. 3 closed-list proportional representation system, centralized groups are more likely to go to a political route as they are better able to enforce and punish politicians via political parties. We test this argument with a major democracy and economy that underwent an electoral reform, Japan. Japan provides a quasi-experimental opportunity to test how interest group organizations and electoral institutions interact to shape lobbying strategies for two reasons. First, it is quasi-experimental because the electoral reform occurred without a major partisan change (Horiuchi and Saito 2003). Second, post-war Japan provides a laboratory of electoral systems because it has employed SNTV (pre-1994) and majoritarian combined with proportional representation systems (post-1994). We use a unique longitudinal elite survey data set of a large sample of interest group leaders at three data points spanning a quarter of century. We demonstrate that interest groups adapt their lobbying strategies and organizations to the change in electoral institutions in which they are embedded. Our argument proceeds in two steps. First, we theorize interest groups lobbying venue choices and intensity as an effort to enforce and monitor a contract with politicians and hypothesize how various organizational structures of interest groups are linked with their potential ability to make politicians commit to this contract. When the interest groups organizational structures allow more effective monitoring and punishment of politicians who fail to commit (i.e., not to reelect or withdraw financial contribution), they are more likely to go to a political route. When their organizational structures do not allow effective monitoring and punishment of politicians who shirk or renege, they are more likely to go to a bureaucratic route to influence an earlier formulation or implementation stage of policy-making. Second, organized interests use various instruments to punish politicians who shirk or renege on the contracts: votes, candidate endorsement, and campaign contributions are the major

19 Who Lobbies Whom? P. 4 examples. We construct typologies of these punishment mechanisms into decentralized and centralized instruments and discuss how the effectiveness of various instruments changed due to the 1994 electoral reform. We expect that the electoral reform from a Single Non-Transferable Vote (SNTV) system with multi-member districts to a mixed SMD/PR system decreases the effectiveness of decentralized punishment mechanism (voting in a district or candidate endorsement) while increases the effectiveness of centralized punishment mechanism (voting for a party). Thus, the reform is expected to diminish the difference between decentralized and centralized groups lobbying strategies. We find, indeed, our organizational structure argument explains variations across interest groups as well as changes after the electoral reform after controlling for their organizational resources and for sectoral issue areas. THEORIES OF LOBBYING VENUE CHOICES Why do some interest groups lobby politicians, while others lobby bureaucrats? The question is at the heart of why interest groups lobby. Two school of thoughts have emerged in the literature on the United States: one that theorizes the interest groups goals of lobbying as changing or enhancing policy-makers preferences ( preference-centered approach ) and another that theorizes lobbying as a function of organizational resources of interest groups or legislators ( resource-centered approach ). Preference-Centered Approach: Exchange and Persuasion The preference-centered approach views organized interests lobbying as an effort to change or align their preferences with legislators to achieve policy goals. In this approach, the two mediums that organized interests use to shape legislators preferences are money and

20 Who Lobbies Whom? P. 5 information. One argument, which is referred to as exchange theories, focuses on the role of campaign contribution in buying support from legislators. While the debates regarding what organized interests seek to buy votes, access, or time of legislators have advanced the literature, the exchange theories still suffer from three major issues. First, without a third party enforcement or punishment, how do legislators and organized interests commit to this quid quo pro exchange? The literature has noticed this political market failure that exchanges between legislators and organized interests are tacit, supra-legal and inter-temporal in their nature and thus are susceptible to legislators reneging and shirking (McCarty and Rosenberg 1996; Snyder 1992; Weingast and Marshall 1988; Stokes 2005; Hall and Deardorff 2006). Yet empirical research on how legislators and organized interests seek to solve this political market failure is few and far between. Second, the empirical support to money-buys-votes argument is weak in light of the findings that interest groups are much more likely to lobby their allies than swing legislators or enemies (Bauer, Pool, and Dexter 1963; Hojnacki and Kimball 1998). This empirical regularity not only holds in the context of the U.S. but also in Japan (Sasaki et al. 1999; Muramatsu and Kume 2006). 1 Stratman (1998) also find that a surprising small amount of money is at stake to buy legislators vote in the United States. If the purpose of lobbying is to change legislators preferences, why would interest groups waste lobbying efforts on their core supporters? Finally, the exchange theory does not account for why interest groups extensively lobby bureaucrats: the 2003 survey done by Muramatsu-Kume group reveals that interest groups spend approximately 60% of their lobbying efforts (i.e., frequency of personal contacts) with bureaucrats only over 1 The 2003 Survey used in this paper (see the Data and Method section below for a detailed description of these surveys) also reveals that only 2.1% of interest group leaders choose campaign contribution as the most important reason for why legislators assist their groups which suggests that organized interests seem not buying legislators votes, access or time.

21 Who Lobbies Whom? P. 6 40% with politician. 51% of leaders say it is most effective to contact bureaucrats to express their opinions and to protect their interests while 36% says it is effective to contact politicians. With its focus on the role of money in changing legislators preferences, the exchange theories do not help us understand why interest groups lobby bureaucrats at all. Another argument, which is referred to as persuasion theories, theorizes information transmission from organized interests to policy-makers as a mechanism to enhance rather than to change policy-makers preferences (Hansen 1991; Wright 1996; Austen-Smith 1993, 1994). This approach views information asymmetry between legislators and organized interests as a key determinant of lobbying and explains several anomalies found in the exchange theories, such as why interest groups lobby allies more than swing legislators and why interest groups lobby bureaucrats. Moreover, unlike the exchange theories in which the money, not information or arguments of the lobbyist, is the variable doing the behavioral work (Hall and Deardorff 2006:71), the persuasion theories model the lobbying process more directly as information transmission via personal contacts and deliberations (Wright 1990). Its weakness, however, is two-fold. First, due to the difficulty in systematically studying the process of private information transmission, empirical tests of these theories have lagged far behind the theories (exceptions are Furlong 1998; Golden 1998; Yackee and Yackee 2006). Second, the persuasion theories assume that legislators face uncertainty about constituents positions on a given policy and this uncertainty makes them rely on information provided by organized interests. This logic, however, does not account for why interest groups often lobby bureaucrats who do not face this need to learn constituents positions. 2 2 The argument also does not account for why some legislators spend time meeting with interest groups that are not necessarily from their districts such as public interest environmental groups and international NGOs.

22 Who Lobbies Whom? P. 7 Finally, neither exchange or persuasion theories help us understand the prevailing patterns in Japan in which interest groups lobby legislators with whom they have a long-standing relationship. The 2003 Muramatsu-Kume Survey reveals what interest group leaders perceive to be a major reason why legislators assist their groups. The highest proportion of leaders (30%) says legislators assist their groups because they agrees with the group s goal or policy, 17% says because they have a long-term trustworthy relationship with a group, 13% says it is because they provide organized votes during elections. Only less than 10% says because the group provides information. What, then, is to be persuaded? Resource-Centered Approach Contrary to the preference-centered views discussed above, Hall and Deardorff (2006) propose a novel, resource-centered argument of lobbying. Direct lobbying, they argue, is a gift from organized interests to like-minded legislators in forms of information and subsidy to assist resource-scarce legislators to work at achieving a policy. The lobbying-as-subsidy theory solves many anomalies found in light of exchange and persuasion theories: why interest groups lobby allies more than swing legislators and why legislators often initiate contacts with organized interests. Yackee and Yackee (2006) demonstrate that not just elected representatives, but also bureaucrats are more likely to respond to big business interests than citizens as the former provides better quality information about complex policy issues. This line of research suggests that the goal of group s lobbying with bureaucrats and politicians might be more similar than thought (Niskanen 1975). Walker (1991) proposes another, resource-centered model of lobbying. He argues that interest groups access and proximity to elites in the Capitol vs. local political actors in a district

23 Who Lobbies Whom? P. 8 determines their lobbying strategies. He demonstrates that localized groups are more likely to use outside lobbying strategies (i.e., mobilize grassroots organizations and the public), while centralized groups are more likely to work though elites such as legislators and federal agencies ( inside strategies; Kollman 1998). Holyoke (2003) accounts for the venue choices by looking at the distribution of power between proponents and opponents of a given policy in venues. Figueiredo and Figueiredo (2002) theorize interest groups venue choices as strategic resource allocation to resource-responsive courts. These resource-centered approaches have an advantage over the preference-centered approach because they account for organized interests venue choices and intensity of lobbying. Although we echo Walker s approach to link centralized vs. decentralized interest groups with their venue choices, our contractual approach differs from his in four major ways. First, in general, electoral institutions are missing from the lobbying literature. We analyze how organizational structures of interest groups interact with electoral institutions to shape their venue choices. Second, groups ability to mobilize support not only depends on their access and proximity to political actors, but also on their ability to monitor, enforce and punish political actors whose efforts do not live up to the bargain. We demonstrate that even controlling for various measures of access and resources, our contractual approach better accounts for lobbying patterns. Accordingly, third, we define decentralized organizational structures as local organizations potential ability to make independent political decisions from the central headquarter such as candidate endorsement, vote switching and campaign contribution. Walker (1991) considers localized vs. centralized interest groups as interest groups organizational resources (pp ) as opposed to organization structures. Finally, Walker s argument does not account for variations within an inside strategy interest groups relative allocation of

24 Who Lobbies Whom? P. 9 lobbying efforts to politicians or bureaucrats. In sum, although the preference and resourcecentered approach have advanced our understanding of why interest groups lobby, they offer little help to understand why some interest groups lobby politicians and others lobby bureaucrats. These approaches also remain silent about varying levels of lobbying, i.e., why some interest groups contact policy-makers very often while others contact a few times. LOBBYING AS CONTRACT ENFORCEMENT We model interest group s lobbying venue choices and intensity as a function of their contract enforceability with policy-makers. Our contractual approach theorizes interest groups goal of lobbying as an effort to enforce a contract with legislators who are already sympathetic to their preferred policies. Our approach accounts for major puzzles about interest group lobbying in general and in Japan in particular: why do interest groups predominantly lobby legislators who are already sympathetic to their policy; why do they lobby legislators with a long-standing relationship; and why do some groups lobby politicians while others lobby bureaucrats? Policy-Making Environment: Electoral Institutions and Legislative Organizations To account for the major puzzles described above, we need to understand incentives of both supply (legislators) and demand-side (organized interests) of lobbying. There are two institutional characteristics that shape legislators incentives to form, renege on, or shirk the contract with organized interests in Japan: electoral institutions and legislative organization (i.e., a committee system). In Japan, under the SNTV system in which district magnitude generally ranged from one to five, same-party candidates of the largest ruling party, the Liberal Democrats, competed for a seat. The SNTV system encourages individual legislators to cultivate personal votes rather than to collectively pursue a coherent party label (Cox and Thies 1998; Ramseyer

25 Who Lobbies Whom? P. 10 and Rosenbluth, 1994 and many others). Moreover, since votes that individual legislators win in a district are not transferable to other same-party candidates, individual legislators have incentives to specialize either geographically (Hirano 2006; Tatebayashi 2004) or sectorally (Ramseyer and Rosenbluth 1994; McCubbins and Rosenbluth 1995) to differentiate themselves from other same-party candidates. Thus, under an SNTV system, legislators incentives were to target narrow constituents as opposed to building a majoritarian coalition with broader and more diverse constituents ( specialization ). Legislative organizations also mirror legislator s personal vote incentives shaped by electoral institutions. The LDP s policy-making organization under an SNTV system was decentralized and specialized. The party decided policies through an extensive committee system (Policy Affairs Research Council; PARC) each of which oversaw corresponding cabinet ministries and bureaucratic agencies. 3 The decentralized committee system also gave policymaking power to party backbenchers who needed to credit-claim to special interests to mobilize votes and campaign contribution. 4 Accordingly, organized interests have incentives to lobby individual legislators in a district, rather than lobbying through a party or party leaders. This decentralized political structure lowers the enforceability of contract between legislators and centralized organized interests. First, individual legislators have incentives to pursue their individual platform and the party does not function as an enforcer of a contract. Second, because the PARC employs unanimity rules and the meeting is not open to the interest groups or the public, 5 organized 3 Cox and Rosenbluth Because the LDP was the ruling party for all but 10 months since 1955, essentially the basic content of policies was decided within the LDP and only sometimes revised or slightly modified in the Diet (parliament) committees. Thus the LDP s PARC was the more important policymaking organ for the government. 4 Rosenbluth and McCubbins 1995:49. 5 Interest groups can only attend the meetings if they are officially invited for hearings.

26 Who Lobbies Whom? P. 11 interests have difficulty identifying who opposes vs. who supports a policy at the PARC meetings or any contribution made during the meetings. 6 This is one of the consequences of the one-party dominant, policy-making environment before 1994 in which major policy decisions were made within the ruling party rather than in the Diet. This means that centralized groups without local organizations have difficulty monitoring and acquiring information about the extent to which politicians commit and contribute to realizing their preferred policy. As a result, contract enforcement, monitoring and punishment mechanisms have to be decentralized and localized so that interest groups can hold individual politicians accountable to their promises. Decentralized Lobbying as Contract Enforcement under SNTV Under the SNTV system described above, politicians and organized interests exchanged promises during the election campaigns but these contracts were susceptible to legislators reneging and shirking. Organized interests used various instruments to reduce wasting their lobbying efforts, such as ex ante contracts and ex post punishment and rewards. Anecdotes suggest that legislators and special interests form such ex-ante contracts in Japan. Aurelia George Mulgan (2006:61) describes what a politician who specialized in the agricultural sector had to go through to get the endorsement of a major agricultural group: For Matsuoka to receive electoral support from the prefectural nouseiren, he had to demonstrate sympathy for, and understanding of, the organisation's agricultural policy campaigns (nousei undou) and to make a public promise of adherence to a position that would reflect the intentions of Nokyo along with farmers in politics. In exchange for recommendation and authorisation (kounin 6 Organized interests have two ways to deal with this lack of transparency at the PARC, one is to rely on media reporters who usually peep the PARC meetings outside the meeting rooms and another is to rely on self-reported stories about the meetings told by politicians (for instance, many politicians leak PARC debates on their blogs). These methods fall short, however, as the media only covers debates surrounding major legislations and politicians can exaggerate their contribution.