Let's Go Fly a Kite: Correlates of Involvement in the House Bank Scandal

Download Let's Go Fly a Kite: Correlates of Involvement in the House Bank Scandal

Post on 17-Jan-2017

213 views

Category:

Documents

1 download

TRANSCRIPT

  • Let's Go Fly a Kite: Correlates of Involvement in the House Bank ScandalAuthor(s): Charles Stewart IIISource: Legislative Studies Quarterly, Vol. 19, No. 4 (Nov., 1994), pp. 521-535Published by: Comparative Legislative Research CenterStable URL: http://www.jstor.org/stable/440172 .Accessed: 14/06/2014 10:37

    Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

    .JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact support@jstor.org.

    .

    Comparative Legislative Research Center is collaborating with JSTOR to digitize, preserve and extend accessto Legislative Studies Quarterly.

    http://www.jstor.org

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/action/showPublisher?publisherCode=clrchttp://www.jstor.org/stable/440172?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp

  • CHARLES STEWART III Massachusetts Institute of Technology

    Let's Go Fly a Kite: Correlates of Involvement In the House Bank Scandal

    This research note examines some hypotheses about why members wrote overdrafts on the House bank, the central behavior of the House bank scandal. We use the negative binomial model, a type of event-count model, to compare the effects of vari- ables related to political power (party, seniority, and electoral security) with personal variables (age and wealth). The results are consistent both with partisan interpretations (Democrats bounced more checks, even when their lower average wealth is controlled for) and with more personalistic interpretations (younger members and members of modest means also wrote more overdrafts).

    The 1992 congressional election produced the greatest turn- over in the House since the one that immediately followed World War II. A central element of the election story was the House bank scandal, which used a broad brush to paint an unflattering picture of the institution.

    Although when viewed dispassionately the behavior that con- stituted the scandal was tame, it became a filter through which issues of democratic accountability were examined in 1992. Not surprisingly, the scandal was immediately subject to much press attention before and immediately after the election, an attention that has been mir- rored by scholars (see Groseclose and Krehbiel 1992; Jacobson and Dimock 1993).

    The scandal was viewed by many-partisans, scholars, and pundits alike-as an object lesson about power and corruption. Specif- ically, the fact that Democrats tended to bounce' more checks than Republicans (Table 1) was used to illustrate the possibility that long- time control of a legislature by one political party could tempt mem- bers of that party to use a public institution for private gain. This note explores the empirical foundation of this inference more dispassion- ately than did the partisans, who understandably were more interested in political gain than in careful social science analysis.

    LEGISLATIVE STUDIES QUARTERLY, XIX, 4, November 1994 521

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • Charles Stewart III

    TABLE 1 Overdrafts in the House of Representatives Bank,

    1 July 1988 to 3 October 1991 (in percentages of members)

    Number of Overdrafts All Members Democrats Republicans

    0 36.1 31.6 43.0 1-10 32.7 33.0 32.0 11-50 13.5 12.5 15.0 51-100 6.8 9.1 3.5 101-200 4.8 6.1 3.0 201-300 1.2 1.3 1.0 301-400 1.2 1.3 1.0 401-500 0.8 1.0 0.5 501-1000 2.8 4.0 1.0 Number of Cases 498 297 200

    Note: The data include all members who served during the period. Bernard Sanders, an independent, had five overdrafts. Source: Congressional Quarterly Weekly Report, 18 April 1992, 1006-7.

    Although Democrats were more likely to be implicated in the scandal than Republicans, the zero-order relationship between partisanship and scandal involvement may be spurious. Democrats may have bounced more checks than Republicans not because of their partisanship, per se, but because of the differences in individual char- acteristics associated with representatives of the two parties. For instance, if members of substantial means were less likely to write overdrafts and if Republicans were on average wealthier than Demo- crats, then the observed difference in the partisan tendency to write overdrafts could really be due to these differences in the average Dem- ocratic and Republican members, not in how the bank treated the par- ties themselves.

    This note shows that there is room for both group-level and individual-level explanations of involvement in the scandal. Senior members with fewer financial resources kited more checks than junior or wealthy members. Still, even with these measurable, individual- level characteristics controlled for in the equation, Democrats wrote more overdrafts than Republicans. These findings have implications for how House members were held accountable for their involvement in the scandal, which is the focus of the concluding discussion.

    522

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • House Bank Scandal 523

    FIGURE 1 Scandal Time Line

    Ethics Committee Report Coverage

    100th Congress 101st Congress

    102d Congress Ethics Committee

    Election Election GAOReport Report lection I I I I I

    I,....--I ---- ...-........ - -....-.-... .......,,, ....,.. . I.I

    1987 1988 1989 1990 1991 1992 1993

    Background

    The scandal was widely reported in the news media at the time it developed, so only the most basic facts will be rehearsed here. Offi- cial congressional and news reports made it clear that this scandal was a continuation of the bank's Keystone Kops history. (See U.S. Con- gress 1992; Kuntz 1992.) The bank-really the House's payroll office-allowed members of the House to write checks on insufficient funds, so long as the checks would be covered by the member's next pay check. The scandal broke when the press covered a General Accounting Office (GAO) audit in late 1991 that criticized the prac- tice. The House Ethics Committee issued two reports in April 1992 revealing the names of all members of the House (including three- fifths of the sitting members) who had written at least one overdraft with the bank sometime during a 39-month period from 1988 to 1991 (see U.S. Congress 1992; Congressional Quarterly Weekly Report 1992). The worst offender among sitting members was Robert Mrazek (D.-NY), who overdrew his account 920 times. Figure 1 sketches the timing of the key events as the scandal unfolded.

    One detail of the scandal bears on the empirical analysis that follows. According to the Ethics Committee report and press accounts that quoted members of Congress, they may have begun writing over- drafts in one of two ways. First, a few members generously interpreted House rules to mean that they had a right to draw their pay one month before it was actually placed in their accounts by the sergeant-at-arms.

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • Charles Stewart III

    Second, many members initially bounced checks through inadver- tence. However, the member might then have received a call from a bank employee saying that the check could be made good if the mem- ber just deposited enough funds so that the overdraft was less than the member's next monthly pay check. This call would alert the member to the same behavior followed by the first group, opening up the possibil- ity of further, purposeful overdrafts.

    Both paths suggest that check kiting was (to use the statistical term) contagious. That is, within the period covered by the Ethics Committee's report, if a member did not kite a check one day, she or he was unlikely to kite one the next. However, kiting one check increased the probability of kiting not just one more check the next day, but many more. This is a feature of the scandal that we will be able to address below in the statistical estimation. Unfortunately, there is no reliable way from official sources-or unofficial ones-to tell which avenue a member followed in getting caught up in the scandal.

    Data

    The data in this paper were drawn from numerous official sources. The dependent variable is the number of overdrafts by mem- bers during the period (July 1988 to October 1991) covered by the House Ethics Committee investigation into the scandal. This 39-month period spans parts of three congresses and encompasses two elections. Unfortunately, the committee's report does not subdivide the number of overdrafts by time periods. Because the Ethics Commit- tee report centered on the 101 st Congress, and because only the 101st was completely covered by the report, the analysis in this paper will consider only those members who were members for the entire 101st Congress.2 If members who died, resigned, or otherwise served for only part of the 101st Congress are excluded, the effective maximum sample size is 431.

    The explanatory variables can be grouped into two categories, one associated with institutional position and political power and one associated with personal circumstances. Four member characteristics examined here are associated with institutional position and power: electoral security, party, committee leadership, and chamber seniority. Electoral security is measured as the average two-party vote share received by the member in the 1988 and 1990 elections. Combining two elections should give a more stable estimate of the member's medium-run electoral security than would the vote in any single elec- tion. For the members in the analysis not running in the 1990 general

    524

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • House Bank Scandal

    election, I used only their 1988 vote share. If a member was unopposed in one of the elections, I used only the vote share from the other elec- tion. For the 15 members of the House who were unopposed in both 1988 and 1990, I set the average vote share at 100%.

    Party is a dummy variable coded as 1 if the member was a Democrat, 0 otherwise. Seniority is measured as the number of con- secutive years the member had served in the House at the start of the 101st Congress. Experimentation with functional forms revealed that the logarithm of seniority fit the data better, so that was the variable actually entered into the analysis. I include two dummy variables mea- suring committee leadership; one indicating committee chairs and one indicating ranking minority members.

    To the degree that involvement in the scandal was prompted by the arrogance of power, we should expect the number of bounced checks to be positively associated with electoral security, being a Dem- ocrat, seniority, and service as a committee leader.

    I examined two personal characteristics of members, their per- sonal wealth and age. Personal wealth was measured in two variables derived from the 1989 financial disclosure forms filed with the House Clerk. The first, the logarithm of estimated unearned income was used to measure members' investment income.3 The second income varia- ble was the logarithm of honorarium income, net of contributions to charities. Since the logarithm of zero is undefined, I added $100 to both of these variables before taking the logarithm. This method induces a small amount of bias into the estimated coefficients, but allows me to include in the analysis members with either no unearned income or no honoraria.

    Age is likely to be a surrogate for many personal factors that are difficult, if not impossible, to measure reliably. In the general pop- ulation younger workers, with families to raise and few accumulated assets, tend to be more in debt and to consume a higher fraction of their incomes than older workers; it is unlikely that members of the House are fundamentally different from the general population in this regard. Hence, older members (controlling for seniority) should bounce fewer checks than younger ones.

    If bouncing checks was sometimes associated with financial distress, then members with lower unearned incomes should have bounced more checks than wealthier members.4 The relationship between honoraria and check kiting is less intuitively obvious. How- ever, upon studying the financial disclosure forms it was clear that the wealthiest members rarely generated substantial net incomes from honoraria, while younger, less financially secure members participated

    525

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • Charles Stewart III

    TABLE 2 Descriptive Statistics for the Independent Variables

    Standard Variable Mean Deviation

    Overdrafts 53 146 Average Vote Share in 1988 and 1990 69 12 Democrat 0.60 Committee Chair 0.052 Ranking Member 0.052 Senioritya 12 8 Age 52 10 Net Honorariab $11,996 $ 15,651 Unearned Incomeb $70,915 $199,593 a Subsequently logged b Subsequently logged after adding $100

    in the honorarium circuit more vigorously. Hence, high honorarium income should be associated with greater check kiting.

    Table 2 reports the means and the standard deviations of the variables, and Table 3 reports the cross-correlation matrix of the inde- pendent variables.

    Estimation

    The dependent variable, checks kited, has a large number of observations at zero, is highly skewed, and is constrained to take inte- ger values greater than or equal to zero (see Table 1 again). Ordinary least squares is not the proper procedure for estimating the parameters of this model, in view of the distribution of the dependent variable and the behavior it attempts to capture. However, recent methodological work has adapted a suite of techniques to political science that are appropriate to address this problem. These techniques fall into the cat- egory of event-count models (Maddala 1983, pp. 51-54; King 1987, 1988, 1989a,b,c; King et al. 1990).5 The best-known event count tech- nique is Poisson regression, which is increasingly used in political sci- ence. The technique used in this paper is a generalization of Poisson regression called the negative binomial model.6

    Readers interested in investigating event-count models are referred to the literature cited above. For the purposes of this research note, one simply needs to know that the coefficients derived from the

    526

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • House Bank Scandal

    TABLE 3 Cross-correlation of Independent Variables

    Variable 1 2 3 4 5 6 7 8

    1. Average Vote Share 2. Democrat .10 3. Committee Chair .11 .19 4. Ranking Member .06 -.25 -.05 5. Seniority (logged) .14 .08 .33 .24 6. Age .09 .01 .29 .21 .56 7. Net Honoraria (logged) .07 .00 .01 .09 .26 .01 8. Unearned Income (logged) .00 -.10 .03 .01 .11 .24 -.11

    estimation can be interpreted as if produced by a linear regression of the form y = eBX, where B is the vector of coefficients and X is the vec- tor of independent variables. One useful extension of the negative binomial technique is that an added parameter, labeled y, describes the degree to which the dependent variable is over-dispersed, the sta- tistical property associated with the contagion phenomenon described above. If this coefficient is greater than 1, we can reject the null hypothesis that individual acts of check bouncing were independent.

    Analysis and Results

    The results of the analysis are found in Table 4. It shows mixed evidence for institutional position and the arrogance of power as explanations for involvement in the scandal.With individual charac- teristics that are weakly related to partisanship controlled for (Table 2), Democrats still bounced more checks than Republicans. The partisanship coefficient in Table 4 can be interpreted to mean that, other factors controlled for, a Democrat would still have had roughly one-quarter more overdrafts than an equivalent Republican. Whether this is a large or small difference is difficult to judge. Nonetheless, the finding of differences by party holds up in all alternative specifications of the model.

    The electoral safety coefficient is weak, both substantively and statistically. Finally, the variables measuring the effects of committee leadership are actually the opposite of what one would expect from an arrogance of power explanation: if anything, committee chairs may have been less likely to bounce checks than other Democrats, and rank- ing minority members clearly bounced fewer checks than otherwise similar rank-and-file Republicans.

    527

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • Charles Stewart III

    TABLE 4 Determinants of Overdrafts, 101st Congress

    (standard errors in parentheses)

    Variables Parameters

    Constant 4.21 (0.03)

    Average Vote Share, 1988 and 1990 0.010 (0.007)

    Democrat 0.25 (0.12)

    Committee chair -0.10 (0.07)

    Ranking minority member -0.16 (0.03)

    Seniority (logged) 0.33 (0.18)

    Age -0.25 (0.014)

    Honoraria (logged) 0.065 (0.021)

    Unearned income (logged) -0.12 (0.04)

    Overdispersion parameter (y) 5.57 (0.06)

    Note: Number of cases = 431.

    The next two independent variables in Table 4, seniority and age, represent the transition from more institutional to more personal factors. Figure 2 illustrates the estimated relationships between senior- ity and age and checks kited. It is true that, with age controlled for, greater seniority led to more check kiting and that, with seniority con- trolled for, greater age led to less check kiting. However, Figure 2 illus- trates how complicated the relationship actually was. For instance, Figure 2a shows that a representative 35 year old was likely to increase the number of checks kited in absolute terms at a rapid rate for each year of service. However, because a 35 year old could have served in the House for a maximum of 10 years, there is a quickly encountered limit to how many checks we would estimate any 35 year old would have bounced. While older members also increase the number of over- drafts as they acquired seniority, the absolute level of increase for each year of service decreased with age.

    528

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • House Bank Scandal 52'

    FIGURE 2 Relationship Between Overdrafts, Age, and Seniority

    a. Expected overdrafts given seniority, controlling for age

    250 - 35 years old

    200 -

    150 - 50 years old

    0 ? > / ...-**'

    U 100 ...

    x

    50 - 65 years old

    0- : :-.. ... I I I , I . I . | J | . | | | * t s | ^ 4~~~~~~~~~~~~~~ 0 5 10 15

    Seniority (in years)

    b. Expected overdrafts given age, controlling for seniority

    140 [ . 5 terms 120

    5 e

    , 100 : 10 terms t . ? .

    > 80- 0

    0. V * *. \

    u 60' Rookie

    40- - .

    20 - ...

    0 L - . , . . .

    20 30 40 50 60

    Age

    9

    20 25 30

    70 80

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • Charles Stewart III

    At the other end of the legislative career, the most senior mem- bers averaged more kited checks, ceteris paribus; however, it is difficult to be a very senior member unless one is quite old. Figure 2b shows that at the extremes of age and seniority together, the age effect dominated.

    Members with lower unearned incomes were significantly more likely to write bad checks than wealthier members. And mem- bers who generated larger incomes from honoraria tended to write more bad checks. Figure 3 illustrates these relationships, with all the other variables held at their means. It is apparent from these graphs that, across the entire ranges of these two income variables, unearned income had the greater total effect.

    To get a total picture of the relationship between a member's financial status and the tendency to write bad checks, one needs to keep in mind not only the two direct measures of income, but also the age variable. Age helps tap at least two financial factors that are not measured here: consumption and expectations about future income and consumption streams. Considered together, the coefficients in Table 4 constitute strong evidence that, whatever role partisanship may have played in check kiting, early- and mid-career members of both parties were the most susceptible to getting caught in the scandal.

    Without the unearned income variable in the equation, the party coefficient is even stronger, since Republicans had higher aver- age unearned incomes than Democrats. The geometric mean of unearned income was $10,823 for Democrats and $16,648 for Repub- licans. Nonetheless, because we were unable to drive away the party variable through the inclusion of the finance variables, it is likely that party per se was a factor in getting caught up in the scandal, too.

    There is one last, more subtle hypothesis to test with these data. If the strongest expressions of the partisanship hypothesis are true, then it could be that financially strapped Democrats took advan- tage of the bank's interest-free loans, while Republicans did not. We can test this hypothesis by re-running the analysis reported in Table 4, inserting a variable equal to Democrat X log(Unearned Income). The resulting coefficient was statistically indistinguishable from zero, sug- gesting that reliance on the bank among the financially strapped was a nonpartisan phenomenon.

    Discussion and Conclusion

    The 1992 election yielded a major electoral turnover in the House. Other social science research has shown that at least part of this

    530

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • House Bank Scandal 531

    FIGURE 3 Relationship Between Personal Financial Variables

    and Expected Number of Overdrafts

    a. Unearned income and overdrafts

    100- . ' '.--- Honorarium =$100

    90 - * Honorarium =$7,325 .'. 80- -- ' Honorarium =$26,850

    '~ 80 - "2 ' '*- '.

    o

    60 -.

    Unearned income (log scale)

    100 - r.

    20 .IA&.

    & I0.

    . . .... . . . . . . . '

    Unearned income (log scale)

    b. Net honorarium income and overdrafts

    100 -

    90-

    -: 80- - Unearned income=$100 ...... Unearned income =$13,625 o 70 - /---- Unearned income = $2,000,000 o a 60- a a 50- ....................

    a 40- '

    ._

    o 30 .

    20 - i . I i. . . . i. . I

    0 5000 10000 15000 20000 25000 30000

    Net honorarium income

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • Charles Stewart III

    turnover can be directly attributed to the bank scandal, since involve- ment in it increased the likelihood that one would retire from the House, be denied renomination, and perform more poorly in the gen- eral election (Groseclose and Krehbiel 1992; Jacobson and Dimock 1993). In addition to affecting individual participants, it is likely that the scandal damaged all members of the House, including those who were not caught up in the scandal at all.7 Because the electoral god of vengeance punished at least some of the check kiters, it is important to know what precisely they were punished for.

    The findings here suggest they may have been punished indi- rectly for their partisanship, their seniority, and their relative youth. However, without a full analysis of the election data itself, it is difficult to know how voters processed the information that their representa- tive was or was not involved in the scandal. For instance, voters may have been less forgiving of wealthy check kiters, such as Stephen Solarz ($52,500 unearned income, 743 kited checks, denied renomination) than of check kiters of more modest means, such as Ronald Dellums (no unearned income, 851 kited checks, reelected). But, as these exam- ples also demonstrate, more than just kited checks interfered with members' career plans in 1992. (Solarz especially was on the losing end of some difficult strategic choices that were induced by redistricting.) So the findings here are suggestive for further work on the electoral consequences of the scandal.

    Another empirical finding that points toward further research is the persistence of partisanship as an explanation for involvement in the scandal, even after controlling for member characteristics corre- lated with partisanship. Two classes of explanations suggest them- selves: first, that a spendthrift public ideology sometimes associated with the Democratic party permeated Democrats' personal behavior in Congress; second, that long-term control of the House by the Demo- crats led to an arrogance of power among some members, and this arrogance in turn led to institutional abuses. Regardless of the root of the partisan differences that emerged in the bank scandal, the findings in this article suggest that Republicans had at least some justification for ascribing blame to the Democratic party collectively.

    One final sidelight of the research here bears mentioning. Debate in recent years over congressional pay and perquisites has sometimes devolved into a battle between the relatively wealthy Sen- ate and the relatively less wealthy House and between members of rela- tively substantial or modest means within the two chambers. The findings here give a more precise glimpse of what may be at stake in these debates. The immediate impression one gets from poring

    532

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • House Bank Scandal

    through the financial disclosure forms is that members differ in their financial sophistication and in the way they handle the bulk of their own finances. Because members of more modest means were affected more by the scandal than wealthy members, one possible consequence of the scandal may be to heighten those tensions even more in future debates over pay, perquisites, and outside income.

    Charles Stewart III is Associate Professor of Political Science and Margaret MacVicar Faculty Fellow, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139.

    NOTES

    I acknowledge the help and inspiration of students in my seminar, Congress and the Policy Process. In particular, four participants of that seminar, Maryann Barakso, Robert Bird, Grant Emison, and Jay Youngclaus, provided critical data and insights for this paper. Gary King generously provided a copy of his computer program, COUNT, for the analysis of the overdraft data. Seminars in the political science depart- ments of Duke University and the Massachusetts Institute of Technology helped me clarify my thinking, as did the comments of the anonymous referees of this Quarterly. The standard disclaimers apply.

    1. A sidelight of the scandal was over what to call the behavior of House mem- bers who were writing overdrafts. The sticklers who insisted that the House bank was not technically a bank insisted that the behavior be called making overdrafts. Because we need some synonyms for this behavior, I will occasionally refer to this behavior by the more common expressions of check bouncing and check kiting.

    2. I experimented with either including all members who served during any part of the period covered by the report or including members from both the 101 st and 102d Congresses. The qualitative results that follow did not change at all when the sam- ple frame was changed. One possible solution to the time period problem would be to divide the number of checks bounced by the time the member served in the House dur- ing the 39-month period. However, the estimation technique used in the next section depends on integer values for the dependent variable. Developing a new statistical esti- mation procedure for noninteger data was well beyond the scope of this research note.

    3. The information in the financial disclosure forms (H.doc. 101-218) about wealth and unearned income is reported without great precision. Unearned income was reported in seven broad categories, ranging from category A, less than $1,000, to cate- gory G, over $100,000. I estimated unearned income by counting the number of items reported within each category and then multiplying that number by the midpoint of the range represented by the category. I counted the few responses in category G as $250,000. Hence, someone with two items in category A, three in category B ($1,000- $2,500), and one in category C ($2,500-$5,000) was estimated to have an unearned income of 2 x 500 + 3 x 1750 + 1 x 3850 = $10,100. The imprecision of the categories means that the estimated relationship between income and checks kited will be attenuated.

    4. One could argue that wealthier members may be more financially sophisti- cated than their impecunious colleagues and thus might recognize more readily the pos- sibilities offered by the interest-free loans that the overdrafts constituted. Hence,

    533

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • 534 Charles Stewart III

    wealthy members may have bounced more checks. Ultimately, the relationship between personal wealth and check kiting is an empirical question that is addressed below.

    5. The use of event-count techniques in this study violated the assumption that the micro-level behavior in question (i.e., the individual overdrafts) occurred dur- ing a uniform period of time across all cases. In fact, of the 431 people in the analysis of this paper, two-thirds served for the entire period covered by the Ethics Committee report and the remaining one-third served some subset of the period. (For instance, those who began their service with the 101 st Congress were absent for the six-month per- iod in the report from the 100th Congress.) Given all the other data-analytical hurdles to be jumped in this analysis, this violation ends up being a minor problem. Furthermore, explicitly controlling for length of service did not materially change the results reported below.

    6. Some commentators to this paper have suggested that Tobit analysis is the proper technique here, a suggestion that is precisely wrong. The Tobit technique is appropriate when the dependent variable is censored. The lower bound of zero for bounced checks does not represent a censoring of the data but rather an inherent charac- teristic of event counts. More appropriate for event counts is something like a hurdle Poisson regression, which separates the movement of the dependent variable from zero to one from the behavior of the dependent variable once it is strictly positive. Using such a technique would overly complicate the analysis here.

    7. However, members who had not bounced any checks and who ran in dis- tricts that were not substantially changed in the 1991-92 redistricting actually gained a small bonus from their noninvolvement (information available from the author). Some members who were not involved certainly tried to take advantage of that fact. Vic Fazio (D.-CA), sent out a campaign mailing boasting as his top achievement the following: "Bounced no checks. While checks bounced out of control in Washington, DC, Vic Fazio showed what fiscal responsibility means. He didn't bounce a single check. Not one." (Winneker 1992)

    REFERENCES

    Congressional Quarterly Weekly Report. 1992. "Overdrafts Listed from Most to Least According to House Ethics Committee" (chart). 18 April, 50:1006-07.

    Groseclose, Timothy, and Keith Krehbiel. 1992. "Golden Parachutes, Rubber Checks, and Strategic Retirements from the 102d House." Mimeo, 25 November, Carnegie Mellon University and Stanford University.

    Jacobson, Gary C., and Michael A. Dimock. 1993. "Checking Out: The Effects of Bank overdrafts on the 1992 House elections." Presented at the annual meeting of the Midwest Political Science Association, Chicago.

    King, Gary. 1987. "Presidential Appointments to the Supreme Court: Adding System- atic Explanation to Probabilistic Description." American Politics Quarterly 15:373-86.

    King, Gary. 1988. "Statistical Models for Political Science Event Counts: Bias in Con- ventional Procedures and Evidence for the Exponential Poisson Regression Model." American Journal of Political Science 32:838-63.

    King, Gary. 1989a. Unifying political methodology New York: Cambridge University Press.

    King, Gary. 1989b. "Event Count Models for International Relations: Generalizations and Applications." International Studies Quarterly 33:123-47.

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

  • House Bank Scandal 535

    King, Gary. 1989c. "A Seemingly Unrelated Poisson Regression Model." Sociological Methods and Research 17:253-55.

    King, Gary. 1989d. "Variance Specification in Event Count Models: From Restrictive Assumptions to a Generalized Estimator." American Journal of Political Sci- ence 33:762-84.

    King, Gary, James E. Alt, Nancy Elizabeth Burns, and Michael Laver. 1990. "A Unified Model of Cabinet Dissolution in Parliamentary Democracies." American Journal of Political Science 34:846-71.

    Kuntz, Phil. 1992. "The History of the House Bank: Scandal Waiting to Happen." Con- gressional Quarterly Weekly Report 8 February, 50:283-89.

    Maddala, G. S. 1983. Limited-dependent and Qualitative Variables in Econometrics. New York: Cambridge University Press.

    U.S. Congress, House. 1990. "Financial Disclosure Reports of Members of the U.S. House of Representatives of the 101 st Congress, for the Period Jan. 1, 1989 to Dec. 31, 1989" (2 vols.). 101st Congress, 2d session, H.doc. 101-218.

    U.S. Congress, House. 1992. "Inquiry into the Operation of the Bank of the Sergeant-at- Arms of the House of Representatives." 102d Congress, 2d session, H.Rpt. 102-452.

    Winneker, Craig. 1992. "Heard on the Hill." Roll Call 12 October, p. 22.

    This content downloaded from 188.72.96.115 on Sat, 14 Jun 2014 10:37:15 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsp

    Article Contentsp. 521p. 522p. 523p. 524p. 525p. 526p. 527p. 528p. 529p. 530p. 531p. 532p. 533p. 534p. 535

    Issue Table of ContentsLegislative Studies Quarterly, Vol. 19, No. 4 (Nov., 1994), pp. 453-596Volume Information [pp. 593 - 596]Front MatterEditors' Introduction [pp. 453 - 455]Formal Models of LegislaturesPrincipals, Goals, Dimensionality, and Congressional Committees [pp. 457 - 476]House Special Rules and the Institutional Design Controversy [pp. 477 - 494]

    Research ArticlesExplaining Contributions: Rational Contributors and the Elections for the 71st Texas House [pp. 495 - 506]Institutional Rules and State Legislator Compensation: Success for the Reform Movement? [pp. 507 - 520]Let's Go Fly a Kite: Correlates of Involvement in the House Bank Scandal [pp. 521 - 535]

    Continuities in Legislative ResearchOld Approaches and New Challenges in Legislative Election Research [pp. 537 - 582]

    Legislative Research Reports [pp. 583 - 591]Back Matter [pp. 536 - 592]

Recommended

View more >