Weingast B.R., Wittman D. The Oxford Handbook of Political Economy

Подождите немного. Документ загружается.

232 legislative pivots

outcome, m =0,toallq in these intervals. Similarly, status quo points will also arise

in segments II or III with probability 1/3. These, however, will all elicit minority party

rolls because the ﬁnal policy m makes at least a majority of the minority party worse

oﬀ than q. The minority roll rate will not be 1/3, however, because in the remaining

two segments, IV and V, the majority party will exercise gatekeeping and thus no votes

will be observed or go into the denominator of calculating roll rates. The predicted

rollrates,then,are:minority=1/2 and majority = 0. That is, on half of the votes on

ﬁnal passage the minority party will be rolled, but on the other half large bipartisan

coalitions will be formed and neither party will be rolled.

With this example—and its corresponding assumption of a ﬁxed, uniform, ex-

ogenous distribution of status quo points—it is easy to see how an expansion of

the distance between the minority median and chamber median |d −m| increases

the minority party roll rate and, therefore, rationalizes the prediction in equation

(1)that‚ > 0. It is likewise easy to see that a change in the extremity of the ma-

jority party median |r − m| aﬀects the majority party’s incidence of gatekeeping

but not its constant-zero roll rate; therefore, the prediction in equation (2)isthat

„ = 0. But ﬁnally, it should also be abundantly clear that the distribution of sta-

tus quo points is absolutely essential to deriving these predictions. If the distribu-

tion were, for example, asymmetric rather than symmetric, variable over time, or

normal rather than uniform, knowledge about what the theory predicts declines

precipitously.

A few miscellaneous points have a bearing on comparative institutional analysis

using roll rate data. First, under the median voter theory, both ‚ and „ should

be positive. Second, the pivotal politics theory does not make predictions in the

regression framework with these variables because party medians play no role in that

theory. Third, cartel theory also implies that the constant terms, ·,intheequations

should not be signiﬁcantly diﬀerent from zero. Finally, the empirical plausibility of

the assumption of ﬁxed, exogenous, constant, and uniformly distributed status quo

points for each Congress is questionable. Among the many objections that might be

raised,themaincauseforconcernhereisthatthisframework,ineﬀect, deﬁes history.

That is, no matter what laws the Congress at time t enacts, the succeeding Congress at

time t + 1 is presumed to begin with a new history-independent distribution of status

quo points.

Two of Cox and McCubbins’s ﬁndings comport with the stated expectations for the

cartel agenda model. The estimate of ‚ is positive and signiﬁcant. The estimate of „ is

positive and insigniﬁcant. However, both constant terms are negative and signiﬁcantly

diﬀerent from zero, and the pseudo-R

measures are low (.037 and .052).

Overall, the

hypothesized preference variables fail to account for much variation in roll rates, and

the degree of discrimination between the two theories is not sharp. As a probable

cause for this the lack of discrimination in the tests, we next revisit the issue of how

status quo points are generated.

¹¹ Cox and McCubbins do not necessarily assume uniformly distributed status quo points, but their

analysis does not address the history-independence problem.

¹² See, for example, Cox and McCubbins 2005,table5.2, and the surrounding text.

keith krehbiel 233

5 Endogenous Status Quo Points

.............................................................................

The limitations of prior comparative institutional analyses underscore the need for

an explicit status-quo-generating framework. Such an approach not only can place

institutional models on equal footing but also may obtain sharper and more discrim-

inating predictions than prior approaches. The framework described below integrates

four components:

actual data on legislators’ preferences,

a presumption that a well-speciﬁed pivot theory is in eﬀect (namely, median

voter, pivotal politics, or procedural cartel),

a generally applicable stochastic model that relies on the focal theory to generate

endogenously the distribution of status quo points for the next Congress,

actual data on roll rates, that is, the proportion of roll calls on which a party is

on the losing side of a winning coalition.

The method performs calculations and evaluates data Congress by Congress. Its

distinctive feature is that when any given Congress commences, the distribution of

status quo points it addresses is not arbitrary, atheoretical, or exogenously imposed

as is customary in most prior tests. Rather, each Congress inherits its unique distribu-

tion of status quo points from the stochastically shocked equilibrium outcomes of the

previous Congress. Given an endogenously derived distribution of status quo points

along with preference measures of relevant medians and pivots, expected roll rates

compatible with the focal theory are calculated.

These expected roll rates are then

compared with actual roll rates using a minimum sum-of-squared-errors criterion.

Finally, diﬀerent theories are compared in terms of their ability to account for roll

rates. More precisely and formally, the procedure is as follows.

1. Postulate a model of stochastic status quo points as a function of prior equilib-

rium policies, namely, q

= x

∗

t−1

+ ε where q

is the status quo at time t, x

∗

t−1

is an equilibrium policy from the prior period, and ε is a random variable

distributed uniformly or normally with mean = 0 and variance to be speciﬁed

(and adjusted for robustness checks).

2. Begin with a neutral policy seed x

∗

centered on the historic average of chamber

preferences, and apply the status quo model to generate a distribution of status

quo points for period 1.

3. Incorporate real-world data (DW-NOMINATE scores

) to identify pivot points

for each congress in the time series. For roll rate analysis, record the majority

and minority party medians as well.

4. For each theory under consideration, indexed j ∈{M, C, P } for median voter,

procedural cartel, and pivotal politics, respectively, and for each Congress t in

the time series,

¹³ The calculations parallel those discussed above and illustrated in Figure 12.4.

¹⁴ Under the assumptions identiﬁed in Poole and Rosenthal (2000) these data are comparable over

time.

234 legislative pivots

Ta b le 12.1 Average observed roll

rates, 80th–99th Congresses

Majority Minority

House 0.02 0.26

Senate 0.04 0.20

(a) Use the theory to calculate equilibrium outcomes x

∗

) for each bin or

interval in the policy space for which q

has non-zero mass.

(b) Calculate the corresponding predicted proportions of minority party rolls

t, j

and majority party rolls

t, j

∗

to derive q

t+1

5. Conduct comparative institutional analysis by assessing the relative abilities of

theories to predict actual party rolls. Speciﬁcally, deﬁne the sum of squared roll

rate errors (SSRE) for theory j as:

SSRE



t=80



−

t, j

)

+(z

−

t, j

)



where y

and z

are the actual observed minority and majority party roll rates.

6 Findings

.............................................................................

Findings are summarized in three tables and a ﬁgure. Table 12.1 reports minority

and majority roll rates for both houses in the Congress. As noted elsewhere (Grose-

close and Snyder 2003; Cox and McCubbins 2005), proportions of minority rolls

exceed majority rolls by a large margin, and this asymmetry seems to support the

procedural cartel theory over the median voter and pivotal politics theories. Such

inferences are questionable, however, in the absence of a baseline model that clariﬁes

expectations. Consider two such models: symmetric bipartisanship and majority party

dominance. In the ﬁrst instance, assume that each party rolls the other exactly one-

half of the time. In the second instance, assume that the majority party always

rolls the minority party. Table 12.2 reports computations of the sum of squared roll

¹⁵ To simplify calculations signiﬁcantly, we move from a continuous choice space to a ﬁne-grained

discrete choice space by partitioning the preference and policy space into narrow bins. Speciﬁcally, for

measures of preferences such as DW-NOMINATE ratings in [−1, 1], a bin width of .05 produces a

41-bin partition (centered on 0,i.e.thecenterbinis(−.025,.025)) in which all ideal points lie, pivots

can be identiﬁed, roll rate predictions can be derived, etc.

¹⁶ The procedure is implemented in a MATLAB program written by Jon Woon. Code and

documentation are available upon request.

keith krehbiel 235

Ta b le 12.2 Sum of squared roll rate errors of

baseline models

Symmetric

bipartisanship

y =

z = .5

Majority

dominance

y =0;

z = .1

House 7.85 14.40

Senate 6.50 13.36

rate errors in actual data relative to these polar baseline models. The result is the

opposite of that suggested in Table 12.1. In the House and Senate alike, the symmetric-

bipartisanship baseline ﬁts the data much better than the majority-party dominance

baseline, with the latter generating approximately twice as much the error as the

former.

Ta bl es 12.1 and 12.2 are both somewhat misleading, however, for approximately

opposite reasons. Table 12.1 fails to specify baseline expectations and thereby gives

rise to inferences that overstate asymmetry. Table 12.2 is based on baseline expecta-

tions that are either perfectly symmetric or completely asymmetric and then aggre-

gates data in a way that masks underlying asymmetry. So, although the symmetric-

bipartisan baseline more closely approximates the data overall, the majority party

dominance baseline provides better account for cross party diﬀerences in roll

rates.

The more important aim of the present analysis, however, concerns the three

theoretically derived sets of expectations embedded in the endogenous status quo

procedure. The results are mixed, but some interesting, stark, and heretofore undis-

covered patterns emerge. Table 12.3 subdivides the ﬁndings along several dimensions:

House versus Senate, uniform versus binomial-normal distributions of the shock

parameter, median versus cartel versus pivot theories, and ﬁnally, the theories’ sum

of squared roll rate errors for the minority party, the majority party, and both parties

combined. Likewise, Figure 12.5 summarizes these data graphically.

Perhaps the most striking ﬁnding is the imbalanced performance of the procedural

cartel theory. For majority party rolls, it is invariably the best of the three models

(columns 4–6). The reason is transparent: the cartel theory predicts no majority

¹⁷ The horizontal axes in Figure 12.5 are denominated by the numbers of bins into which an

equilibrium outcome at time t can be shocked to generate a status quo at time t+1. For example, a

maximum range of 41 means that the equilibrium policy in the current period could be moved as far as

20 bins to the left or 20 bins to the right as the status quo for the next period, which is approximately half

the range of legislators’ DW-NOMINATE ratings. Notice, however, that in the binomial approximations

of the normal distribution, while the range in this scenario is the same as in the case of the uniform

distribution, the variance will be considerably smaller because of the unimodality. Consequently, for any

point x on the horizontal axis of the left-hand-side ﬁgures, the actual variance is greater than the

corresponding point x on the right-hand-side ﬁgures.

236 legislative pivots

Ta b le 12.3 Comparative analysis of three models using the endogenous status quo method

Minority Rolls Majority Rolls Combined

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Chamber Dist. Var. Median Cartel Pivot Median Cartel Pivot Median Cartel Pivot

House Uniform L 2.95 1 10.97 3 4.37 2 4.97 2 0.01 1 5.26 3 7.91 1 10.98 3 9.63 2

House Uniform M 1.91 1 6.75 3 1.91 2 4.10 3 0.01 1 3.57 2 6.01 2 6.76 3 5.49 2

House Uniform H 0.92 2 2.84 3 0.66 1 2.25 3 0.01 1 1.67 2 3.16 3 2.85 2 2.33 2

House Normal L 4.97 1 13.82 3 8.24 2 5.60 2 0.01 1 7.24 3 10.57 1 13.83 2 15.48 3

House Normal M 4.42 1 13.03 3 6.86 2 5.35 2 0.01 1 6.31 3 9.77 1 13.04 2 13.17 3

House Normal H 4.05 1 12.45 3 5.92 2 5.22 2 0.01 1 5.84 3 9.27 1 12.47 3 11.76 2

Senate Uniform L 2.63 1 10.40 3 5.27 2 3.55 3 0.12 1 2.24 2 6.18 1 10.53 3 7.51 2

Senate Uniform M 2.09 1 6.41 3 2.47 2 2.84 3 0.12 1 1.76 2 4.93 2 6.54 3 4.24 2

Senate Uniform H 1.25 2 2.91 3 1.19 1 1.38 3 0.12 1 0.70 2 2.63 2 3.03 3 1.89 1

Senate Normal L 4.11 1 12.85 3 10.54 2 3.92 3 0.12 1 1.14 2 3.14 1 12.97 3 11.68 2

Senate Normal M 3.56 1 12.09 3 8.55 2 3.62 3 0.12 1 1.44 2 3.44 1 12.21 3 9.99 2

Senate Normal H 3.29 1 11.56 3 7.30 2 3.53 3 0.12 1 1.66 2 3.66 1 11.68 3 8.96 2

Note: House data are from 80th–105th Congresses (Cox and McCubbins 2005); Senate data are from 80th–99th Congresses (Cox and McCubbins 2002). Normal

distributions are approximated by binomials. Range is measured in number of bin units (see n. 17): L = 21, M=41, H = 61. Pairs of cell entries are the sum of

squared roll rate errors and the rank order of performance across the three models within the chamber-distribution-variance combination.

keith krehbiel 237

Pivotal politics

Median voterProcedural cartel

25 50 75 100

Maximum range of status quo shock

0 5 10 15

Sum of squared roll rate errors

(House, Uniform)

25 50 75 100

Maximum range of status quo shock

0 5 10 15

Sum of squared roll rate errors

(Senate, Uniform)

25 50 75 100

Maximum range of status quo shock

0 5 10 15

Sum of squared roll rate errors

(Senate, Normal)

25 50 75 100

Maximum range of status quo shock

0 5 10 15

Sum of squared roll rate errors

(House, Normal)

Fig. 12.5 Comparison of predicted roll rate errors using endogenous-status-quo

method

party rolls, and few such rolls occur. Meanwhile, the median voter and pivotal politics

theories systematically over-predict such rolls.

But while the cartel theory is uniformly successful in predicting majority party

roll rates, it is uniformly abysmal in predicting those for the minority party. Without

exception, it predicts far more minority party rolls than in fact occur. Indeed, all of the

errors summarized in column 2 are due to over-prediction and these are considerably

greater than any other theory’s errors in columns 1–6. Consequently, when roll types

are combined and the overall predictive accuracy of the theories is compared in

columns 7–9, the cartel theory is the poorest performer.

For most assumptions about the distribution of the status quo shock term, the

median voter theory performs best of the three models. This is always the case under

the assumption of approximately normally distributed shocks. The pivotal politics

theory, in contrast, generates the fewest errors under the assumption of high variance

of uniformly distributed shocks.

The mixed nature of the ﬁndings—and especially their clear dependence on the

assumptions about the error term of the status quo shock—has two implications.

First, the ﬁndings underscore the importance of employing an explicit framework

that can accommodate a wide range of assumptions about the status quo. Second,

¹⁸ When the House data in the ﬁrst six rows of Table 12.3 are partitioned into Congresses in which a

given theory over- versus under-predicts, the pivot and the median voter theories never predict fewer

majority party rolls than occur. It is likewise rare in the Senate.

238 legislative pivots

because such assumptions clearly matter, future research should direct attention to

estimating real-world status quo points and incorporating them into frameworks

such as this one.

7 Conclusion

.............................................................................

Comparative institutional analysis of pivot models has only recently begun to make a

mark in the ﬁeld of political economy, and this research has taken only a small step in

illustrating its potential. An analytic, a methodological, and a substantive conclusion

can be drawn.

Analytically, pivot theories are remarkably ﬂexible, parsimonious, and powerful

tools of contemporary political economy. They can capture a wide range of variation

in institutional and behavioral features of collective choice. They can do so while re-

maining simple, transparent, and tractable. And, most importantly, their tractability

can be exploited to generate comparative, refutable hypotheses.

Methodologically, however, pivot theories and related developments in the litera-

ture have experienced growing pains. A long tradition of research in spatial models,

in legislative studies and elsewhere, has drawn substantial analytic leverage from the

concept of an exogenous status quo point, which, within a well-speciﬁed game form

in which voters have well-deﬁned preferences, leaves one such decision-maker pivotal

in equilibrium. With a rich assortment of such theories at our disposal, attention has

increasingly turned to empirical analysis that seeks to discriminate between models

that identify diﬀerent pivots depending upon which institutional features the models

encapsulate. Such analyses are more subtle and delicate than is often acknowledged

because of an important diﬀerence between the status quo as an analytic construct

and the status quo as an empirical phenomenon. Analytically, the status quo is

exogenous, pure and simple. Empirically, however, the status quo is typically an out-

come from a previous process of collective choice in the same institutional setting to

which the theory is intended to apply. Consequently, improvements in empirical tests

of pivot theories seem crucially dependent upon more careful attention to theory-

consistent distributions of status quo points. This study has demonstrated a new

method that is uniquely responsive to this concern.

Substantively, the combination of analytic explicitness and a methodological ap-

proach that accounts for status quo endogeneity seems quite important. Previous

studies tend not to be explicit about assumptions about the status quo and come

to conclusions of two types: either strong claims about the superiority of partisan

theories (e.g. Cox and McCubbins 2002, 2005;Campbell,Cox,andMcCubbins2002;

Chiou and Rothenberg 2002), or exasperation about the inability to discriminate

between competing theories (Krehbiel, Meirowitz, and Woon, 2004). This study

reachesmuchdiﬀerent conclusions from the ﬁrst type and somewhat sharper con-

clusions than the second. The cartel theory clearly provides the best account for the

keith krehbiel 239

paucity of majority party rolls, which it should: the distances between its majority

party gatekeeping assumption, its no-majority-roll prediction, and the ex ante known

paucity of such rolls are razor-thin. On the other hand, the party cartel theory is

clearly inadequate in predicting minority party rolls once status quo distributions are

endogenized. Meanwhile, the pivotal politics theory exhibits a higher mean and lower

variance in performance than the cartel theory. Yet, it is not the pivotal politics model

but rather the original and simplest pivot model—Black’s median voter theory—that

performs best in most circumstances.

References

Black,D.1958. The Theory of Committees and Elections. London: Cambridge University Press.

Brady,D.,andVolden,C.1998. Revolving Gridlock. New York: Westview Press.

Campbell,A.,Cox,G.,andMcCubbins,M.2002.AgendapowerintheU.S.Senate,1877–

1986.InParty, Process, and Political Change in Congress, ed. D. Brady and M. McCubbins.

Stanford, Calif.: Stanford University Press.

Chiou,F.Y.,andRothenberg,L.2003. When pivotal politics meets partisan politics. Ameri-

can Journal of Political Science, 47: 503–22.

Cox,G.,andMcCubbins,M.2002. Agenda power in the U.S. House of Representatives, 1877–

1986.InParty, Process, and Political Change in Congress, ed. D. Brady and M. McCubbins.

Stanford, Calif.: Stanford University Press.

—— —— 2005. Setting the Agenda. Cambridge: Cambridge University Press.

Crombez,C.,Krehbiel,K.,andGroseclose,T.2006.Gatekeeping.Journal of Politics, 68.

Denzau,A.,andMackay,R.1983. Gatekeeping and monopoly power of committees: an

analysis of sincere and sophisticated behavior. American Journal of Political Science, 27:

740–61.

roseclose,T.,andSnyder,J.M.,Jr. 2003.Interpretingthecoeﬃcient of party inﬂuence:

comment on Krehbiel. Political Analysis, 11: 104–7.

Krehbiel,K.1996. Institutional and partisan sources of gridlock: a theory of divided and

uniﬁed government. Journal of Theoretical Politics, 8: 7–40.

—— 1998. Pivotal Politics: A Theory of US Lawmaking. Chicago: University of Chicago

Press.

—— Meirowitz,A.,andWoon,J.2005. Testing theories of legislatures. In Essays in Honor of

Jeﬀrey S. Banks, ed. J. Duggan and D. Austen-Smith. Berlin: Springer-Verlag.

—— Shepsle,K.,andWeingast,B.1987. Why are congressional committees powerful? Amer-

ican Political Science Review, 81: 929–45.

Mayhew,D.R.1991. Divided We Govern. New Haven, Conn.: Yale University Press.

Mo,J.2001. Political culture and legislative gridlock: politics of economic reform in precrisis

Korea. Comparative Political Studies, 34: 467–92.

Nacif,B.2003. Policy making under divided government in Mexico. Working Paper #305,

Kellogg Institute, University of Notre Dame.

Poole,K.T.,andRosenthal,H.

2000. Congress: A Political-Economic History of Roll Call

Vot in g. Oxford: Oxford University Press.

Romer,T.,andRosenthal,H.1978. Political resource allocation, controlled agendas, and the

status quo. Public Choice, 33: 27–43.

240 legislative pivots

Shepsle,K.A.1979. Institutional arrangements and equilibrium in multidimensional voting

models. American Journal of Political Science, 23: 27–59.

—— and Weingast,B.R.1987. The institutional foundations of committee power. American

Political Science Review, 81: 85–104.

Wawro,G.J.,andShickler,E.2004. Where’s the pivot? Obstruction and lawmaking in the

pre-cloture senate. American Journal of Political Science, 48: 758–74.

chapter 13

..............................................................

THE POLITICAL

ECONOMY OF THE

US PRESIDENCY

..............................................................

charles m. cameron

By virtue of the extraordinary economic and military might of the United States, the

American president is the most powerful elected oﬃcial in the world. Not surpris-

ingly, the American president is also the most intensively studied of all political ex-

ecutives. Nonetheless, a distinctly political economic understanding of the presidency

is a recent undertaking. Despite this novelty, progress has been rapid and continues

apace.

As the political economy approach has developed, several features distinguish it

from the earlier legalistic, biographical, historical, and behavioral schools of presiden-

tial analysis. First, the political economy approach consciously suppresses the details

of presidents’ individual personalities in favor of an admittedly simplistic but highly

tractable psychology. Typically, analysts assume that straightforward goals, such as

advancing policy objectives or securing re-election, motivate presidents. In adopting

this view, the new approach to the presidency emulates modern work on Congress,

the bureaucracy, and the judiciary.

Second, the political economy approach self-consciously embeds the president

in the American separation-of-powers (SOP) system.

The SOP system, hardwired

into American government by the Constitution, compels the president to interact

repeatedly with Congress and, to a lesser degree, the federal courts. And, from the

time of the New Deal, the SOP system also compels the president to interact with huge

¹ De Figueiredo, Jacobi, and Weingast (this volume) survey the emerging literature on the new

separation-of-powers approach as applied to the major American national institutions.