Perverse Accountability: A Formal Model of Machine Politics with Evidence from Argentina

Author(s): Susan C. Stokes

Source: The American Political Science Review, Vol. 99, No. 3 (Aug., 2005), pp. 315-325
Published by: American Political Science Association
Stable URL: http://www.jstor.org/stable/30038942
Accessed: 05-01-2016 18:02 UTC

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 of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact [email protected].

American Political Science Association and Cambridge University Press are collaborating with JSTOR to digitize, preserve
and extend access to The American Political Science Review.

http://www.jstor.org

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions
American Political Science Review Vol. 99, No. 3 August 2005

PerverseAccountability:A FormalModelof Machine Politics

with Evidencefrom Argentina
SUSAN C. STOKES Yale University
Political machines (or clientelistparties) mobilize electoral support by trading particularistic benefits
to voters in exchange for their votes. But if the secret ballot hides voters' actions from the machine,
voters are able to renege, accepting benefits and then voting as they choose. To explain how
machine politics works, I observe that machines use their deep insertion into voters' social networks
to try to circumvent the secret ballot and infer individuals' votes. When parties influence how people
vote by threatening to punish them for voting for another party, I call this perverse accountability. I
analyze the strategic interaction between machines and voters as an iterated prisoners' dilemma game
with one-sided uncertainty. The game generates hypotheses about the impact of the machine's capacity to
monitor voters, and of voters' incomes and ideological stances, on the effectiveness of machine politics. I
test these hypotheses with data from Argentina.

Thirty-five years ago, James Scott (1969) observed machines also have the secret ballot. Political machines
that political life of contemporary new nations did not disappear in the United States after the intro-
bore a strong resemblance to the machine poli- duction of the Australian ballot in most U.S. states at
tics of the United States in earlier eras. The patronage, the end of the nineteenth century.1 And clientelism
particularism, and graft endemic to the Philippines or flourishes in countries from Mexico (Fox 1994) to Italy
Malaysia in the postwar decades recalled, for Scott, (Chubb 1982) to Bulgaria (Kitschelt et al. 1999), all of
the Tweed machine in nineteenth-century New York which have the ballot.
or the Dawson machine in twentieth-century Chicago. Assuming that machines can overcome the problem
Much has happened in the third of a century since Scott of their clients' reneging, what kinds of voters will they
outlined "the contours and dynamics of the 'machine target? Scattered through the qualitative literature is
model' in comparative perspective" (1143). Many of evidence that poor voters are the targets of machines
the new nations that occupied his analysis have under- (see, e.g., Chubb 1982; Wilson and Banfield 1963). For-
gone transitions to electoral democracy; yet politics in mal treatments agree, citing diminishing marginal util-
these systems often remains particularistic, clientelis- ity of income as the reason why particularlistic ben-
tic, and corrupt. We therefore have a larger sample of efits generate more votes among the poor than the
countries, and a richer experience on which to draw, rich (Calvo and Murillo 2004; Dixit and Londregan
to understand the contours and dynamics of the ma- 1996).
chine. The historiography of the U.S. political machine Yet in the societies where clientelistic parties or ma-
has also grown, as have historical studies of patronage chines are active, not all poor voters receive benefits.
and vote buying in the history of today's advanced Limited resources force political machines to choose
European democracies (see, e.g., Piattoni 2001). Fi- among poor voters. Machine operatives everywhere
nally, a formal literature on redistributive politics has face a version of the dilemma that an Argentine Pero-
developed, one in which the political machine plays a nist explains. About 40 voters live in her neighborhood,
central role. and her responsibility is to get them to the polls and
Yet the formal literature on the political ma- get them to vote for her party. But the party gives her
chine leaves some crucial questions unanswered. Chief only 10 bags of food to distribute, "ten little bags," she
among them: How does the machine keep voters from laments, "nothing more."2 How does she, and machine
reneging on the implicit deal whereby the machine dis- operatives like her in systems around the world, de-
tributes goods and the recipient votes for the machine? cide who among her neighbors shall and who shall not
If voters can renege, then machines should not waste receive handouts?
scarce resources on them and clientelist politics breaks The formal literature answers this question by saying
down. The question is the more pressing, given that that machines target core constituents. But if these con-
many of the societies in which we find active political stituents are ideologically committed to the machine,
is it not wasting resources if it distributes rewards to
them? Would it not do better by distributing rewards to
Susan C. Stokes is the John S. Saden Professor of Political Science, the uncommitted or even to those who, on ideological
Department of Political Science, P.O. Box 208301, Yale University,
New Haven, Ct 06520-8301. ([email protected]).
I thank Valeria Brusco, John Carey, Matt Cleary, Avinash Dixit, 1 The Australian ballot is one in that is produced by governments
Jeff Grynaviski, John Londregan, Scott Mainwaring, Roger Myerson, or neutral election authorities (rather than by political parties), dis-
Luis Fernando Medina, Mario Navarro, Marcelo Nazareno, Steve tributed through guarded channels on or close to election day, and
Pincus, Duncan Snidal, and three anonymous APSR reviewers for that lists all parties or candidates for an office in a single format.
excellent comments. I am especially indebted to Michael Wallerstein. 2 Interview conducted by Valeria Brusco, Susan Stokes, and Gloria
Research was supported by National Science Foundation research Trocello, in Villa Mercedes, Argentina, July 2003; my translation.
Grant SES-0241958 and by the John Simon Guggenheim Memorial This and all subsequent translations from the Spanish are by the
Foundation. author.

315

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions
Perverse Accountability August 2005

grounds, oppose the party? The selection method of every group, the parties deploy tactical rewards to com-
our Argentine operative is to help "the people who pete for the same groups of swing voters-groups with
complain the most, the ones who say, 'What are you a relatively large number of moderate voters who are
going to give me?' I pick them up [to take them to ideologically indifferent between the two parties. But
the polls] and after I take them they say, 'Aren't there when one party has an especially close link to a group of
any bags of food?"' Her words hint at a logic in which voters, then the party will target this core constituency.
machines give private handouts not to die-hard sup- Dixit and Londregan write that core constituents are
porters but to people whose future support is in doubt. ones
The analysis in this article helps make sense of her
whom [the party]understandswell... A party'score con-
explanation.
stituencies need not prefer its issue positions. It is the
Far from being just a Latin American problem, or a party'sadvantageover its competitors at swaying voters
problem that advanced democracies have completely in a groupwith offers of particularisticbenefitsthat makes
overcome, vote buying, clientelism, and machine poli- the group core (1986, 1134). . .The key to the electoral
tics are blights on many democracies around the world, strategiesof the urbanpolitical machineswas their ability
even today. Prosecutors in 2004 accused a candidate to provide"personalservices"to their core constituentsat
for a district judgeship in Eastern Kentucky of giving a lowercost than could their competitors.They did this by
$50 checks to voters, implicitly in return for their sup- knowing their constituents (1147).
port.3 Journalists reported, also in 2004, that an elderly
For Cox and McCubbins (1986), the crucial feature
hospital patient in Ukraine confessed to his son that
of the machine-core constituent link is that the party
he had voted for the official presidential candidate, is more certain about how core groups will respond
Viktor Yanukovych, rather than for the opposition can-
to rewards than it is about other groups. The party
didate, Viktor Yuschenko. He had planned to support
is more certain because "core supporters... are well-
Yuschenko but switched his vote after a nurse at the
known quantities. The candidate is in frequent and
hospital promised him a wheelchair if he switched.4 intensive contact with them and has relatively precise
These practices make a mockery of democratic ac- and accurate ideas about how they will react" (1986,
countability. Democratic accountability usually means 378-9).
that voters know, or can make good inferences about,
The problem with both pairs of authors' models
what parties have done in office and reward or punish is that they don't deal adequately with commitment
them conditional on these actions. But when parties problems. Both assume by caveat that the party won't
know, or can make good inferences about, what indi- renege on its offer of particularistic rewards once it's
vidual voters have done in the voting booth and reward
won the election.5 And they don't deal adequately
or punish them conditional on these actions, this is per-
with the fact that a voter, once in the voting booth,
verse accountability. We usually think of accountability can also renege by voting his or her conscience or
in democratic systems as a good thing: it means that preference, ignoring the reward he or she received.
voters can keep elected officials from misbehaving and When we translate these authors' models into one-
pressure governments to be more responsive to voters.
shot strategic interactions between party operatives
But perverse accountability is bad for democracy: it and voters, redistributive politics does not happen. (For
reduces the pressure on governments to perform well
reasons of space, I do not analyze such games here.)
and to provide public goods, keeps voters from using The operative doesn't give a reward, and the possibil-
elections to express their policy preferences, and under-
ity of a reward doesn't change the voter's vote. This
mines voter autonomy (see Karlan 1994; Kochin and
commitment problem looms not only over the relation
Kochin 1998; O'Donnell 1996; Stokes 2004). To over- between machines and core constituents but also over
come perverse accountability, we need first to under-
the one between parties and swing voters: the party's
stand how machine politics works. This article begins
dominant strategy is to renege, and the voter's is to vote
to build such an understanding. for the party it prefers on ideological or programmatic
grounds, not the one that deployed tactical rewards.
STATIC MODELS OF REDISTRIBUTIVE
POLITICS AND THE COMMITMENT A DYNAMIC MODEL OF MACHINE POLITICS
PROBLEM
In some of our leading formal models of redistributive Assumptions
politics, the political machine plays a large role. Dixit A way to deal with these commitment problems is
and Londregan (1996) model the strategies of two par- to place the machine-voter interaction in a dynamic
ties as they attempt to mobilize groups of voters, who context. To model the interaction between machine
care both about consumption and about ideology. Par- operatives and voters as a repeated game, we have to
ties tax some voters and redistribute to others. When make certain assumptions. First, we have to assume
both parties are equally able to deliver resources to that parties can monitor individual voters' actions and

3 The New York Times, August 29, 2004. 5 Aware that parties in their model suffer from a commitment prob-
4 "Ukrainian Campaigns Gear Up for Presidential Re-Vote," Emily lem, Cox and McCubbins simply add an assumption "that candidates,
Harris, December 7, 2004, www.npr.org. once elected, carry out their promises" (1986, 373).

316

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions
American Political Science Review Vol. 99, No. 3

condition rewards on their inferred votes. Second, we (2004) notes that parties designate polling agents to ob-
have to assume that both sides perceive the interaction serve the progress of voting. Polling agents are "usually
as ongoing indefinitely into the future. men from the village itself, or from close by, who know
The assumption that machines can hold voters ac- the identity of each voter. While they do not witness
countable, that they can monitor individuals' votes the actual vote, they know who shows up to vote and
(even if imperfectly) and make rewards contingent can report on turnout figures" (139). Chandra reports
on the voter's support, departs from the implicit as- that Indian parties could undermine voters' anonymity
sumption of redistributive theorists. They assume that a by emptying boxes and counting the returns at fre-
member of a favored group will receive private rewards quent intervals over the course of an election day. (An
whether or not he votes for the party; individual vot- electoral reform in 1994 outlawed the practice.) To cite
ers are anonymous and therefore free from the party's another example, in the 2003 Russian Duma elections,
retribution should they defect. The premise that voting international observers reported "significant problems
is a private and anonymous act may have discouraged relating to the secrecy of the vote, with open voting in
formal theorists from modeling these interactions as 30% ... of polling stations... polling officials and party
repeated games; repeated games generally rely on each observers were seen to be actively encouraging persons
player being able to observe the actions of the other to vote outside of polling booths" (Organization for
in the previous round. The assumption that voting is Security and Cooperation in Europe 2003).
anonymous is appropriate in most advanced democ- Certain party-organizational structures allow par-
racies, but not necessarily in the historical context of ties to discern individual voters' types-their predis-
political machines or in many new democracies today.6 position for or against the machine. The typical po-
There are two kinds of private information about the litical machine (or clientelist party) is bottom-heavy,
voter that are useful to the party: his actions-which decentralized, and relies on an army of grassroots mil-
party he votes for-and his type-his partisan predis- itants. Voters in today's democracies in the developing
position in relation to the two parties. Machines are world are frequently geographically immobile, living in
good at gathering information about voters' actions neighborhoods where they grew up and where family
and types. Indeed, formal theorists have identified fea- members and close acquaintances live. Some of these
tures of the machine that makes it good at discerning familiar neighbors work as operatives for political par-
what people need and delivering it to them efficiently, ties. They therefore know much about an individual
but these same features also make it good at discerning that shapes his partisan attachments: his job, associ-
individuals' likely votes. ational membership, parents' ideological inclinations,
Certain voting technologies allow parties to monitor and public statements about parties and policies. It is
individuals' votes. The recent historiography of U.S. also hard for voters to dissemble before people they've
machines deepens our appreciation of these technolo- known all their lives: as one grassroots party organizer
gies. Until the introduction of the Australian ballot in Argentina explained, you know if a neighbor voted
in the United States, in most states in 1891, parties against your party if he can't look you in the eye on
produced "ticket" or "coupon" ballots, ones that listed election day.
only their candidates. To monitor which party's ballot Information about individual voters' partisan pre-
the voter was using, parties printed ballots on paper dispositions helps the machine make inferences about
of different weights or colors. Voters deposited the how individuals vote and whether they are good can-
ballot directly in the ballot box, under the watchful didates for vote buying. For instance, the model in the
eye of party operatives, without first concealing it in an next section shows that voters who are predisposed
envelope (for descriptions, see Keyssar 2000; Reynolds in favor of the machine on partisan or programmatic
1988). Reynolds (1980, 193) reports that New Jersey's grounds cannot credibly threaten to punish their fa-
early automatic voting machines, introduced in 1890, vored party if it withholds rewards. Therefore the party
made clicking noises that allowed party officials stand- should not waste rewards on them. The model also
ing nearby to detect the voter's selection. And oper- shows that voters who are strongly opposed to the ma-
atives from the Philadelphia Republican party in the chine will not trade their votes for rewards. A machine
late nineteenth and early twentieth century offered to can compensate, to some degree, for an effective secret
fill out ballots on voters' behalf (McCaffery 1993). ballot if it can distinguish strong opponents from peo-
Voting practices and technologies undermine the ple who oppose it more moderately, or strong loyalists
anonymity of the vote in contemporary developing from people who are indifferent about whom to vote
democracies, as they did in U.S. machine cities, even for.
where the Australian ballot is in use and where voting Argentina, the country from which I present evi-
is, in a narrow sense, secret. In her description of con- dence, combines a balloting system that gives parties
temporary India as a "patronage democracy," Chandra greater control over voters than does the Australian
ballot, a social structure of reduced anonymity, espe-
6 It is not always appropriate for advanced democracies. In contem- cially among the poor, and party organizations that
porary Spain, voters retrieve sheets containing party lists from an help parties monitor voters. These features contribute
open table at their polling place. They can retreat into an enclosed to a widespread perception among Argentine voters
booth to cast their ballot. But they are not required to vote in secret
and many vote in the open.
and party operatives that voting is a less than fully
7 In the models that follow, I assume two-party competition, as do anonymous act. As one grassroots party organizer ex-
the theorists of redistributive politics discussed earlier. plained, "Anyone who's militating in the streets, you

317

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions
Perverse Accountability August 2005

know who's with you and who's not with you."8 In a how the party used the ballots. "The most important
survey conducted in four Argentine provinces in July- thing is to go look for people and give them the ballot.
August 2003, respondents were asked, "Even though You give them the ballot in the taxi [which the party
the vote is secret, do you believe that party operatives has hired to transport them to the polls]. Then no one
can find out how a person in your neighborhood has has time to change their ballots for them [i.e., give
voted?" Despite a technically secret ballot, 37% of the them a different ballot. After taking voters into the
sample responded that party operatives can find out, polling place] you put them on line to vote... Then
51% that they cannot, and the remaining 12% didn't they don't have a chance to change the ballot. Only
know (total sample size: 2,000).9 This perception was if they're really sneaky and they change it inside the
echoed in an interview with a couple from a small city voting booth."12
in the Argentine province of C6rdoba: In sum, my first assumption is that machines can
effectively, if imperfectly, monitor the actions of their
Husband: Here it's different than in C6rdoba [the near- constituents.
est big city]. Here they know everyone. And they A second assumption needed to model machine pol-
know whom everyone is going to vote for. itics as a repeated game is that all players foresee the
Author: When people come and give things out during game continuing into the future. It is entirely appropri-
the campaign, are they people whom you know? ate to think of the interactions between machine oper-
Husband: Yes, they're people from here, they're neigh- atives and their constituents as repeated over many
bors. Here everyone knows each other. "Small town, iterations; the more artificial assumption would be
big hell." (Pueblo chico, infierno grande.) that these are one-shot or short-lived interactions. Ma-
Author: Do they know how you voted? chines and clientelist parties are effective to the extent
Husband: For many years we've seen, people will say, that they insert themselves into the social networks
"So-and-so voted for so-and-so." And he wins, and of constituents. The grassroots party operative is a
they come and say, "You voted for so-and-so." I don't long-time neighbor of the people she tries to mobilize.
know how they do it, but they know. In Latin America, clientelist parties of renown have
Wife: We were at the unidad bdsica [a neighborhood been long-standing organizations, deeply enmeshed
Peronist locale] and they say to me, "[Your cousin] in working-class communities: Peru's Partido Aprista
voted for Eloy" [the given name of a Radical-party Peruano (APRA), founded in the 1920s, Mexico's In-
candidate]. And I asked my cousin, "did you vote stitutional Revolutionary Party (PRI), founded in the
for Eloy?" And she said "yes"! They knew that my 1930s, Argentina's Peronists, founded in the in the
cousin had voted for Eloy!10 1940s.
Voting technologies in Argentina also reduce the The repeated-play assumption may be most ap-
anonymity of the vote. Argentina has the secret but not propriate in countries where parties are old, even if
the Australian ballot.11 Argentines vote with slips of the democracies in which they compete are new. The
paper that carry the names only of a given party's can- three democracies just mentioned are new: Peru and
didates, like the coupon ballots used in the nineteenth- Argentina redemocratized in 1980 and 1983, respec-
century United States. People can vote with ballots that tively, and Mexico democratized for the first time in
they receive directly from party operatives. Or they 2000. Yet clientelist parties in all of them are old. The
can vote with ballots supplied inside the voting booth. repeated-play assumption may be less appropriate in
People tend to receive ballots as part of a process of new democracies where the major political parties are
direct, face-to-face mobilization. also young and hence less enmeshed in social networks.
The practice of handing out ballots basically serves When parties that are not enmeshed in social net-
as a method of monitoring and influencing how people works try to buy votes with private inducements, voters
vote. One Peronist organizer explained in an interview greet their efforts with skepticism. In connection with
research I conducted in Lima, Peru, I observed the
reactions of people in a working-class neighborhood
to a soup kitchen that a political party established
8 Interview conducted January 2003, in the city of C6rdoba, by in 1985, shortly before national elections (see Stokes,
Valeria Brusco, Marcelo Nazareno, and Susan Stokes. 1995). Soup kitchens were familiar in the neighbor-
9 We used multistage cluster sampling techniques, based on census hood: Catholic activists and women's organizations ran
tracks, to select 500 adults each in the provinces of Buenos Aires,
C6rdoba, Misiones, and San Luis. The margin of error was plus or some and the local mayor's office supported them.
minus 4.5%. But when residents saw an outsider party set up a
10 Interview conducted by Valeria Brusco, Lucas Lizaro, and Susan soup kitchen they predicted that it would disappear
Stokes, July 2003. after election day. They were unmoved by the sponsor-
11 Scholars often fail to distinguish between the two. Argentina, party's implicit appeal for electoral support. And they
Panama, and Uruguay are examples of developing democracies that
don't use Australian ballots but where balloting is secret. Voting were right: the soup kitchen did disappear right after
takes place in enclosed booths, and ballots are placed in opaque the election.
envelopes before being returned to election officials. But the ballots In the Argentine case, furthermore, it is appropriate
are produced by political parties and contain only a given party's to assume that parties and voters see their interaction
list of candidates. Furthermore, I have cited two other developing
democracies, India and Russia, where the Australian ballot is used
but where experts claim that the secrecy of the ballot is informally 12 Interview conducted in June 2002 in the city of C6rdoba, by Valeria
violated. Brusco, Marcelo Nazareno, and Susan Stokes.

318

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions
American Political Science Review Vol. 99, No. 3

FIGURE 1. The Location on a Spatial Dimension of a Political Machine (xi), Its Opponent (x2), the
Median Voter (x*), and a Hypothetical Voter (x;)

4- Favor machine Oppose machine -+

I I I I "

x1 X* Xi X2

as extending into the indefinite future; even if they The Model

could imagine hypothetical circumstances in which it
might end (in the event, e.g., of a military coup), at the I begin with a one-shot game in which a person's vote is
time of any given election since the return to democracy assumed to be perfectly observable by political parties.
in that country, few would have anticipated a particu- Let the ideological position of the machine in a one-
lar moment when it would end. The perception of an dimensional policy space be represented by xl, the ide-
interaction with no identifiable stopping point makes ological position of the opposition by x2, and xl < x2.
Let x* = (xl + x2)/2 be the midpoint between the two
it reasonable to model this as an infinitely repeated
game. parties (see Figure 1). Let the voters' preferences be
To capture the repeated-play dynamic of machine given by
politics, it is necessary to depart in a third way from Ui = -'(vi - Xi)2 + bi,
received models of redistributive politics. These mod-
els assume that the machine's ability to reward voters where vi = {x, x2 } represents a vote for either the ma-
for their support depends on its winning elections.13 chine or the opposition, xi represents voter i's position
A voter whose support will only be rewarded if the on the ideological spectrum, and bi = {0, b} represents
machine wins anticipates that the game in effect ends the value to the voter of the reward offered by the
each time the machine loses. Many machines, such as machine in exchange for votes, relative to the value
Mexico's PRI (Diaz-Cayeros, Magaloni, and Weingast of voting according to the voter's preferences. Thus
2001), Singapore's People's Action Party (PAP; Tam -(1/2)(vi - xi)2 represents the expressive value of vot-
2003), or, for many decades, Italy's Christian Demo- ing for one of the two parties. If the machine does not
cratic Party as it operated in the south (Chubb 1982), offer a gift, then bi = 0 and the voter votes for the ma-
face negligible competition. Because the machine ef- chine if -(xi - xi)2 > -(xi - X2)2, or if xi > x*. That is,
fectively cannot lose, voters anticipate that the game if there is no gift the voter supports the party that falls
will continue. But other machines operate in settings closest to the voter on the ideological or programmatic
where they can lose. Even in competitive settings, the dimension. If the machine offers a gift of b > 0, the
game between machine and voter need not end when voter will vote for it if
the machine finds itself in opposition. It does not end
-1/2(xi - xi)2 + b > -1/2(xi - X2)
if the machine can carry over public funds from the
party's time in power, or if it can make use of resources or
donated by private actors, private actors who expect
policy concessions from the machine when it is back in b 1[(xi - xl)2 - (Xi - X2)2] = (X2 -- X)(Xi - X*),
power (Stigler 1975). Note that two of the three long-
or
term clientelist Latin American parties mentioned ear-
lier, the Peronists and APRA, were more often in op- - X1)).
xi < x* + (b/(x2
position than in power.
To summarize, my key assumptions are that ma- The normal form of the stage game is depicted in
chines can monitor voters' actions and that both sides Table 1. In the Table, the machine is represented as
foresee their interaction extending indefinitely into the expending b when it pays a reward, and gaining v when
future. The latter assumption implies that machines it receives a vote.
don't lose their ability to distribute goods when they Define voters for whom x < x* as Loyal voters (see
find themselves in opposition. Figure 2). Loyal voters' dominant strategy is to vote
for the machine. Define voters for whom x > x +

TABLE 1. Normal Form of a Game Between

13 In static models of clientelism in which the party only pays a the Machine Operative and a Voter
reward if it wins, a voter's actions depend on his or her beliefs about Machine
the likely actions of other voters. A collective-action problem arises
when voters prefer, on programmatic grounds, to vote against the Voter Reward No Reward
machine. Then defeating the machine is a public good, but individual
voters pay a cost for attempting to unseat it if the attempt fails. See Comply -1/2 (x - x)2 + b, v- b -1/2 (x - x1)2, v
Medina and Stokes 2003, and Diaz-Cayeros, Magaloni, and Weingast Defect -1/2(x, - x2)2+ b, -b -1/2 (x, - x2)2,O
2001.

319

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions
Perverse Accountability August 2005

FIGURE 2. Types of Voters by Their Location on a One-Dimensional Policy Space

Loyal Weakly opposed Opposition ^

Xi X* x* + b/(x2 - Xi) X2

reward and votes for the machine, supported by a grim

TABLE 2. Normal Form of the Game trigger strategy should the voter be observed to renege,
Between the Machine Operative and is
the Weakly Opposed Voter with
Simplified Payoffs 1/(1 - /)[b - (xi - x)2/2]
Machine
> [b - (xi - x2)2/2] + [/(1 - /)]{(1 -p)
Voter Reward No Reward
Comply 3, 3 1,4 x [b - (xi - x1)2/2] -p(xi - x2)2/2}. (1)
Defect 4, 1 2, 2
In other words, to sustain cooperation, the value to
the voter in the current and all subsequent periods of
voting for the machine and receiving a reward must
b/(x2 - X1) as Opposition voters. Opposition voters will equal or exceed the sum of the payoff from defecting
oppose the machine even if offered b to change their in the current period plus (1) avoiding detection and
votes. Define voters for whom x* < x < x* + b/(x2 - returning to cooperation in the next and subsequent
xl) as Weakly opposed voters. Weakly opposed voters periods (with probability p), or (2) being caught and,
prefer to vote against the machine in the absence of in all subsequent periods, voting against the machine
a reward, but prefer to vote for the machine if doing but foregoing rewards (with probability 1 - p).
so brings them a reward. If the value of the vote to Inequality [1] simplifies to
the machine exceeds b, the machine and the Weakly
opposed voter are in a prisoners' dilemma. Table 2 xi < x* + .(b/x2 - xi),
gives the game between a Weakly opposed voter and
a machine, with simplified payoffs that make clear the where
prisoners'-dilemma structure of the game.
Next, I assume an infinite sequence of elections X = pl/(1 - , + p/).
and model the interaction between the machine and
a Weakly opposed voter as an iterated prisoners' Hence, the set of voters who would sell their votes in
dilemma with one-sided uncertainty.14 I also assume exchange for a private benefit is the set whose ideal
that the two are playing a grim-trigger strategy, point, xi, satisfies
whereby when one player defects, the other defects in
all subsequent rounds. Aside from theoretical reasons X* < xi Ix*±x (bX2 Xi). (2)
in favor of the grim trigger, interviews with Argen-
tine party operatives suggest that they in fact follow Lambda falls on the [0, 1] interval. Lambda is an
a strategy of this sort. For instance, we asked a Pero- increasing function of the discount rate (/3) and of the
nist organizer how she responded when she suspected probability of a defector being caught (p). If p = 0
that a person to whom she had extended favors voted (there is no possibility that the machine would observe
for another party. She answered, "He's dead. He died, a defection by the voter), or if /3= 0 (the voter cares
forever."15 nothing about future consumption), then inequality [2]
Returning to the model, if the voter votes against reduces to xi = x*. In these cases the machine can buy
the machine, I now assume, the machine observes the the votes only of voters who are indifferent, on ideo-
negative vote with a probability p. Voters discount the logical grounds, between the parties.
future by a discount factor /, which falls on the interval Loyal voters do not meet the condition in [2]. As
[0, 1]. The condition for a subgame-perfect equilibrium illustrated in Figure 2, for Loyal voters xL < x*. In-
(SPE) in which the Weakly opposed voter receives the tuitively, Loyal voters who want to extract private
rewards from their preferred party would, under the
grim trigger, have to threaten to vote against the party
14 In a sense there is uncertainty on both sides, about whether forever if the machine denied them a reward once.
the other will cooperate or defect in the future. This uncertainty Such a threat would lack credibility: the party knows
characterizes all iterated prisoners' dilemmas-indeed, all repeated
games-in which there is more than one equilibrium. I model this that the Loyal voter, even without rewards, is better
game as one of one-sided uncertainty because only the machine is off cooperating forever than defecting forever.16 Nor
uncertain about whether the voter has cooperated or defected. The
voter, by contrast, observes perfectly whether the machine gives him
a reward. 16 The loyal voter's diehard ideological commitment to the party
15 Interview conducted in January 2003 in the city of C6rdoba by allows the machine, in a sense, to exploit him, garnering his vote
Valeria Brusco, Marcelo Nazareno, and Susan Stokes. without having to spend scarce resources on him. Loyalists would

320

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions
American Political Science Review Vol. 99, No. 3

do Opposition voters, those who oppose the machine MACHINE POLITICS AND VOTE BUYING
on programmatic grounds more strongly than do the IN ARGENTINA
Weakly opposed, satisfy condition [2] (for Opposition
voters, xo > x* + X(b/x2 - X1)). The reason is that even The comparative statics from my formal model gener-
though the Opposition voter would like to receive a re- ate hypotheses about the causes of machine or clien-
ward, the machine cannot use the threat of withholding telist politics. In this section, I test these hypothe-
a reward to secure this voter's compliance: he is always ses with evidence from one developing democracy,
better off forgoing the reward and voting against the Argentina.18 The evidence I present comes mainly
machine. The machine knows this and does not offer from a survey of 1,920 voters, conducted in December
him a reward. 2001 and January 2002 in three Argentine provinces.19
Weakly opposed voters (and indifferent voters, The survey allows us to explore the strategies of clien-
where xi = x*) are the only types whose policy ideal telist parties indirectly, by revealing what kinds of
points make them potential vote sellers.17The intuition voters these parties target and who among the vot-
behind this result is that, in contrast to the Opposition ers are responsive to private rewards.20 Respondents
voter, Weakly opposed voters can credibly commit were asked whether they had received any goods from
to voting for the machine in exchange for a gift; the a political party during the election campaign that
machine knows that the voter is better off cooperat- had taken place two months earlier (variable name,
ing forever than defecting forever. In contrast to the Reward). Of low-income respondents in the sample,
Loyal voter, the threat to punish the machine by voting 12% (89 out of 734) reported having received goods.
against it in the future by the Weakly opposed voters is Most of them said that they had received food; other
credible: left to their own devices, this is their preferred items mentioned frequently were building materials,
course of action. mattresses, and clothing. In an open-ended question
Inequality [2] implies four comparative statics: about whether receiving goods influenced their vote
(Influence), about one in five of the low-income voters,
* As the ideological distance between the two par- and one-quarter of low-income Peronist voters, said it
ties (x2 - x1) shrinks, the potential for vote buying did. We asked other questions meant to detect clien-
grows. Intuitively, when the two parties are ideologi- telism, such as whether the person had turned to a
cally or programmatically close, there is less at stake locally important political actor for help during the
for the voter in the decision of which to vote for, past year (Patron) and whether, if the head of their
and the value of the private reward becomes more household lost his or her job, the family would turn to
salient. a party operative for help (Job).
* As the value of the private reward (b) relative to the
value of voting in accordance to one's policy or ide-
ological preference increases, the potential for vote Poverty and Vote Buying
buying increases. The reward must be worth a lot to I discuss five pieces of evidence from the survey that
the voter. But its value to the machine must be less lend support to my theory of machine politics. The
than the value of a single vote-not very much. This first has to do with the effect of poverty on a voter's
suggests that, given decreasing marginal utility from willingness to sell his or her vote. The formal model
income, machines will target poor voters. analyzed earlier predicts that vote buying is more easily
* The more accurately the machine can monitor voters,
the greater the potential for vote buying (X is an
increasing function of p). This accuracy is a function 18 The one comparative static from the model that I do not test is
of the technology for monitoring voters' actions and that ideological proximity between the parties encourages vote buy-
ing. The surveys did not elicit respondents' views of the ideological
of the machine's organizational structure. distance between Argentina's two major parties.
* Among its core constituents--those whom it can 19 As in the 2003 survey reported on earlier, we used multistage
observe well-the machine is most effective when cluster sampling techniques, based on census tracks. In this earlier
it targets Weakly opposed voters (for whom x* < survey we selected 480 adults each in the provinces of Buenos Aires,
xi 5 x* + X(b/x2 - X1)), rather than Loyal (xi < x*) C6rdoba, and Misiones, and from the area of Mar del Plata. The
margin of error was plus or minus 4.5%.
or Opposition voters (xi > x* + b/(x2 - x1)) voters. 20 Students of political clientelism and redistributive politics have
typically observed the distribution of resources and their effects on
voting at aggregated levels, such as the district or the county (see,
e.g., Ansolabehere and Snyder, 2002, or Diaz-Cayeros, Magaloni,
and Weingast, 2001). The problem of ecological inference can mar
therefore have an incentive to masquerade as indifferent voters, a this approach. In contrast, the main problem with the survey ap-
possibility that I do not model here. It might, however, be psycholog- proach used here is that people may be reluctant to acknowledge
ically difficult for party enthusiasts to feign indifference. Note also receiving handouts, in the Argentine case probably as much because
that any ideological shift by the machine runs the risk of turning the of the implication that they are poor enough to sell their votes as
loyalist into an indifferent or even an opposition voter. Machines out of concern about the illegality or immorality of their actions. It
would then have to consider the distribution of loyal voters and the is probably evidence of this reluctance that only 7% of our sample
additional resources that might be needed to retain their support, acknowledged having received goods, whereas 44% said goods were
were it to consider a change in its ideological stance. distributed in their neighborhood, 39% could mention exactly what
17 Their minmax payoffs are, for the machine, 0, and, for WO, items were distributed, and 35% could name the party that gave them
-1/2(xwo - x2)2. Hence, the feasible and individually rational pay- out. The effect of underreporting of clientelism is, in estimations
offs they will accept in repeated play include the cooperation payoffs where it is the dependent variable, to bias coefficients downward
of (v - b, -1/2(xwo - X1)2 + b). and make statistically significant associations appear insignificant.

321

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions
Perverse Accountability August 2005

sustained, all else equal, when the voter values the

TABLE 3. Model Estimations of Vote Buying
private reward relatively highly but the party values
it relatively little. The picture this paints is of parties (1) (2) (3) (4)
giving minor benefits to voters who are poor enough Dependent Patron Job Reward Influence
Variable
to value them highly-a picture consistent with much Logit Logit
Model Logit Ordered
of the qualitative literature on machine and clientelist
Estimated Logit
parties. To cite just one of many examples, Wilson and Income -0.126 -0.054 -0.194
-0.195
Banfield (1963) explain that U.S. machines operated (0.058) (0.037) (0.074) (0.070)
in a city's "river wards," where working-class residents Education -0.005 -0.197 -0.212 -0.223
lived, but not in the "newspaper wards," where middle- (0.058) (0.035) (0.079) (0.073)
class residents lived. Housing -0.215 -0.133 -0.212 -0.310
Table 3 reports regression estimates of the likeli- quality (0.114) (0.073) (0.131) (0.022)
hood of a clientelistic response to the set of questions Log -0.361 -0.035 -0.135 -0.139
discussed earlier, including whether the respondent population (0.044) (0.029) (0.050) (0.045)
received a private reward from a party. The negative Ballot 0.578 0.572
and significant coefficients on Income, Education, and (0.225) (0.211)
Peronist 0.594 0.735 0.550 0.549
Housing quality variables show that poverty predicts sympathizer (0.192) (0.119) (0.220) (0.207)
clientelism. To illustrate the effect, the simulated ex- Age -0.005 -0.022 -0.016 -0.017
pected probability that a wealthy person (one with the (0.006) (0.003) (0.007) (0.006)
highest income, education, and housing-quality level) Gender -0.178 0.208 -0.158 0.092
would have received a reward and acknowledged that it (0.166) (0.103) (0.195) (0.180)
influenced her vote is 0.2%. The probability that a poor Radical 0.357 0.146 -0.455 0.026
person (one with the lowest income, education, and sympathizer (0.243) (0.158) (0.371) (0.299)
housing-quality level) would have received a reward Constant 3.254 1.879 1.580
and allowed his or her vote to be influenced by it is (0.643) (0.397) (0.746)
65 times greater: 13 %.21 N 1114 1920 1618 1619
In sum, political machines buy the votes of poor peo- observations
ple in Argentina. Note: Cell entries are coefficients, and standard errors are in
parentheses. Boldface indicates significance at the p= 0.05
level or smaller.
Explanation of dependent variables: Patron: "Inthe past year,
Monitoring Voters have you turned to [the person the respondent previously identi-
fied as the most important local political figure) for help?" Coded
Machine Organizational Structure. In the presence yes = 1. Job: "Ifthe head of your household lost his or her job,
of the secret ballot, parties make inferences about how would you turn to a party operative for help?" Coded yes = 1.
people vote by observing their type-where they fall Reward: "Didyou receive goods distributed by a party in the last
on the dimension of programmatic support for the campaign?" Coded yes = 1. Influence: "Didthe fact of having re-
ceived goods influence your vote?" Coded 1 = Did not receive
parties. A tentacle-like organizational structure is a goods; 2 =received goods, no influence; 3= received goods,
great asset to parties in this regard. We know from acknowledged influence. Based on responses to open-ended
a large secondary literature that the Argentine party question.
with the organizational structure most like that of the Explanation of independent variables: Log population: natural
machine is the Peronist party (see, e.g., Auyero 2000, log of population of respondent's municipality (2001 census).
Ballot coded 1 for people who reported voting with a ballot given
and Levitsky 2003). And our surveys indicate that the to them by a party operative, 0 for people who voted with a ballot
Peronist party was by far the most active in distributing they acquired in the voting booth. Peronist sympathizer, coded
private rewards. Eight hundred thirty-nine of our re- 1 for respondents who said they liked the Peronist Party more
spondents said that a party distributed private rewards than others, 0 otherwise. Income: Self-reported by respondent,
9-level scale. Education: 9-level scale, from no formal educa-
in their neighborhoods during the campaign; of these, tion to postgraduate. Housing quality. Assessed by interviewer,
423 (50%) said that the Peronists distributed them. 5-level scale (1 = poorest quality, 5 = highest quality). Gender
The next most frequently mentioned party, the Radical female = 1. Radical sympathizer, coded 1 for respondents who
Party, was mentioned by only 49 respondents. said they liked the Radical Party more than others, 0 otherwise.

Community Structure. The ease of monitoring is also

21 All simulations reported in this section were executed with the influenced by the structure of communities where ma-
Clarify program (King, Tomz, and Wittenberg, 2000, and Tomz,
Wittenberg, and King, 2001). Clarify draws simulations of param-
chines operate. We expect voters to be less anonymous,
eters of statistical models (in this case, ordered logit regressions) their partisan predispositions or types more a matter of
from their sampling distribution and then converts these simulated public knowledge, in smaller towns and cities, where so-
parameters into expected values, such as expected probabilities of an cial relations are multifaceted and where, as one person
answer to a survey question, given hypothetical values of explanatory we interviewed put it, "everyone knows each other."22
variables. Clarify software and documentation are available from
Gary King's web site at http://gking.harvard.edu. For this simulation These are places where it is easier for parties to know
I assumed a female Peronist supporter whose age and municipality
size were average for the sample. Confidence intervals around the
0.2% expected probability were 0.05% and 0.5%, and around the 22 Interview conducted by Valeria Brusco, Lucas Lazaro, and Susan
13% probability, 7% and 22%. Stokes, July 2003.

322

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions
American Political Science Review Vol. 99, No. 3

who's who, who is inclined toward one party or another, have seen, is 13%. This assumes that the voter received
and how people are likely to have voted. And they are his or her ballot from a party operative. If we assume
places where, as the same person explained, parties the same hypothetical poor voter voted with a ballot
can use this information to "discriminate a little" when he or she finds in the voting booth, the probability is
a defecting voter comes and "asks for a favor." It is cut almost in half, to 7%.
reasonable, then, to treat community size as a proxy In sum, in Argentina the more able a party is to
for observability of residents' votes. monitor its constituents, the more effective its efforts
In our surveys, the smaller the population size of at vote buying. The party with the most decentralized
the respondent's municipality, the more likely she or and tentacle-like organizational structure, hence the
he was to have received rewards and to be responsive one best able to monitor the actions and types of its
to them. These two effects are revealed in Table 3 by constituents, was the party that most actively attempted
the negative and significant coefficients relating logged to buy votes. And the more observable the vote, either
population size (as measured in the 2001 census) to because the voter lives in a small community or be-
Patron Reward and to Influence variables.23 cause he or she receives the ballot directly from a party
operative, the more likely he or she is to be the target
The Technology of Voting. The last two findings- of vote buying.
that rewards are distributed by the party with the most
machine-like structure, and that people in small towns
and cities are more likely to receive, and to be re- Types of Voters and Vote Buying
sponsive to, rewards-might be interpreted as simply
A fifth piece of evidence speaks to the question, What
showing that parties hand out rewards preferentially
types of voters do machines pursue? My theoretical
to people whom they can reach most efficiently. But I
prediction was that machines focus their vote-buying
have argued that efficiency of distribution is just one
efforts on people in the middle of the distribution
side of the link between political machines and their
of partisan predispositions: ones who are indifferent
constituents. The other side is perverse accountability:
about whether to vote for or against the machine
the machine's ability to hold voters accountable for
(XM= x*), and ones with a weak predisposition against
their votes.
it (x* < x < x* + b/(x2 - X1)). Machines will avoid vot-
The fourth piece of evidence that I report goes di-
ers who are loyalists or strong opponents.
rectly to a party's ability to discern people's votes and
We asked respondents their opinions of the Pero-
to condition rewards on compliance. This evidence
nist party, Argentina's preeminent political machine.
has to do with the technology of voting. Recall that
We asked them to choose among "very good," "good,"
Argentines vote with party-produced ballots, which "bad," and "very bad" as their answers. Figure 3 dis-
they can acquire either directly from party operatives,
plays the percentages of people who received or did
as part of the process of face-to-face mobilization, or not receive handouts by opinions of the Peronist party.
anonymously, in the voting booth. Ballot in Table 3 A striking finding, and one that conforms to the
is a dummy variable for people who voted with ballots
theoretical prediction, is the small proportion of those
given to them by party operatives (15% of our sample).
who rated the party "very good" who received rewards.
The positive and significant coefficient relating Ballot Three times as many people who did not receive re-
to Reward shows that people who vote with person-
wards as those who did receive them rated the Peronists
ally distributed ballots are more likely than others to "very good" (31% vs. 10%). The Peronist party turned
receive rewards from parties, such as food or clothing. away from its strongest loyalists when it gave out pri-
The positive and significant coefficient relating Ballot
vate rewards. (The difference is all the more striking
to Influence shows that people who receive person- given that one might anticipate some endogeneity of
ally distributed ballots are also more responsive to
perceptions of the party: people who receive rewards
rewards.24
from it might be more prone, because of the gift, to rate
To give a sense of the magnitude of this effect, the
the party "very good.") Another aspect of the findings
simulated expected probability that a poor voter would
that accords with my model's predictions is that many
allow his or her vote to be influenced by a reward, as we more people who rated the party "bad" received re-
wards than those who rated it "very bad."
23 Note that 90% of our interviews were with people who lived in
In some ways, however, the findings do not accord
cities with more than 10 thousand inhabitants. Thus, we interviewed with the predictions. Recipients of rewards were con-
few people who could be said to live in rural communities, and our centrated in the "Good" category: nearly 60% of those
population variable is best interpreted as distinguishing people ac- who received handouts from the Peronists saw it as
cording to the size of the urban area in which they lived. a good party. These findings are inconsistent with the
24 The confidence intervals around the 7% figure are 4% and 12%.
theory if we think of people who called the machine
An alternative interpretation is that parties, as a service, deliver bal-
lots to the loyal partisans, who are more likely to vote for them any- "good" as falling somewhat to the left of the median
way. In this case partisanship would "cause" both the hand delivery ideal point (x*) in Figure 2 and hence as being weakly
of the ballot and support for the party, and the apparent link between predisposed in the machine's favor. Recall that, in
ballot delivery and support would be spurious. Yet this alternative theory, even voters just mildly predisposed in the ma-
explanation is inconsistent with the testimony of party operatives,
who, like the one cited earlier, focus their ballot-delivery efforts
chine's favor would not be able to credibly threaten to
on uncommitted or indifferent voters, ones who-they fear-might punish the machine if it defected and therefore would
change the ballot in the voting booth. not, in repeated play, be able to induce the machine

323

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions
Perverse Accountability August 2005

FIGURE 3. Opinions of Peronists Among Recipients and Nonrecipients of Rewards

100%
90%
80%
70%
60% 0 Very Good
OGood
50%
* Bad
40% E Very Bad
30%
20%
10%
0%
No Reward Reward

to pay them rewards. And we might have expected a politics. For reasons of space, I haven't addressed the
relatively larger proportion of voters who rated the question, If two parties compete by offering private
Peronists as "bad," and hence who were weakly op- rewards, what determines a voter's choice? One can
posed to it, to receive rewards. Similarly, note that the imagine a bidding-war dynamic, where the value of
regression models in Table 3 show that, controlling for private rewards escalates rapidly. If two parties offered
other factors, Peronist sympathizers were significantly private rewards of the same value, one would expect
more likely to receive rewards. the machines to compete for the same set of (ideo-
One explanation for the slippage between the theory logically) marginal voters. But competition between
and the evidence is that our survey did not offer people "dueling machines" seems, empirically, unusual. It is
the option of indicating true indifference about the more common that, even in settings where politics is
Peronists. Some people who chose the "good" option competitive at the macro level, parties have especially
might in fact be closer to indifferent. And some people close links to particular groups of voters. And often
who were close to indifferent, prerewards, might have one party specializes in machine-style politics, whereas
called the party "bad" but, because of the reward, been another focuses on programmatic mobilization.
nudged into seeing it as "good." This last point raises the question, If parties that
The finding may also suggest a dynamic that goes be- are organized as machines can use minor payoffs to
yond the model. Political machines organize by neigh- sway voters, why don't all parties organize themselves
borhood and district, and they do more than just give this way? A tentative answer is that parties face un-
out tactical rewards. They also proselytize. Although equal costs of monitoring voters. Monitors are most
their proselytizing, in a competitive setting such as effective when they live among the voters they are
Argentina's, is not perfectly successful, to the extent observing. Given residential segregation by income,
that it is successful at all we expect the distribution of parties with a middle-class base would have to em-
voter types in areas of machine organizational pene- ploy middle-class monitors, who would require greater
tration to be skewed toward machine supporters, weak compensation than do the working-class operatives.
and strong. In other words, we expect organizational Parties with middle-class constituencies therefore are
penetration by the party to increase not only the effi- more effective when they advertise their programs,
ciency with which it distributes rewards and its ability focusing resources on "air," rather than "ground,"
to monitor voters, but also its partisan support (as Cox campaigns.
and McCubbins 1986 assume). If organizational pen- These limitations notwithstanding, we have made
etration increases partisan support, then the machine some headway. I have returned to Scott's insight that
will target its supporters more than its opponents sim- machine politics of old is a lot like clientelist politics
ply because it has greater access to them. Whatever of new. I have argued that the dynamics of machine or
the explanation for this anomaly, the evidence from clientelist redistribution has only been half-understood
Argentina does show unambiguously that, among core in the literature, which has captured the delivery-of-
constituents, the machine discriminates against its most services but not the monitoring-of-voters side of the
ardent supporters. story. The literature thus misses the fact that machines
are able to use their social proximity to voters to mon-
itor their actions and types and hence to enforce the
CONCLUSIONS
implicit redistributive contract. This insight allows us
The dynamic model I analyze and test here by no to model the strategic interactions between machines
means answers all of our questions about machine and constituents as repeated games.

324

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions
American Political Science Review Vol. 99, No. 3

Casting voter-machine interactions as repeated Karlan, Pamela. 1994. "Not by Virtue but by Money Won? Vote
games allows us to overcome commitment problems, Trafficking and the Voting Rights Systems." Virginia Law Review
80 (October): 1455-75.
over which the formal literature has stumbled, and to Keyssar, Alexander. 2000. The Right to Vote: The Contested History
identify equalibria in which vote buying actually takes of Democracy in the United States. New York: Basic Books.
place. I have shown formally that when voters see par- King, Gary, Michael Tomz, and Jason Wittenberg. 2000. "Making
ties as ideologically close to one another, vote buying is the Most of Statistical Analysis: Improving Interpretation and
more likely to occur. I have shown, formally and empir- Presentation." American Journal of Political Science 94: 347-
61.
ically, that machines target poor people, for whom the Kitschelt, Herbert, Zdenka Mansfeldova, Radoslaw Markowski,
payoff of even a small reward outweighs the expres- and Gibor T6ka. 1999. Post-Communist Party Systems: Compe-
sive value of voting for one's preferred party. Empiri- tition, Representation, and Inter-Party Cooperation. Cambridge:
cal evidence also supports the theoretical finding that Cambridge University Press.
the more accurately the machine monitors individual Kochin, Michael S., and Levis A. Kochin. 1998. "When Is Vote Buying
Wrong?" Public Choice 97: 645-62.
voters, either through a tentacle-like party structure or Levitsky, Steven. 2003. Transforming Labor-Based Parties in Latin
through voting technologies that reduce the anonymity America: Argentine Peronism in Comparative Perspective. New
of the vote, the more successful are its efforts at vote York: Cambridge University Press.
buying. And evidence supports (though with some nu- McCaffery, Peter. 1993. When Bosses Ruled Philadelphia: The Emer-
gence of the Republican Machine, 1867-1933. University Park, PA:
ances) the theoretical finding that machines avoid ex- Penn State Press.
tending largesse to diehard loyalists and focus their Medina, Luis Fernando, and Susan C. Stokes. 2003. Clientelism as Po-
rewards on voters in the middle of the distribution of litical Monopoly. Chicago Center on Democracy Working Paper
partisanship. #25.
The Argentine evidence, then, on the whole sup- O'Donnell, Guillermo. 1996. Another Institutionalization: Latin
America and Elsewhere. Kellogg Institute Working Paper
ports the theoretical finding that perverse account- #222. Notre Dame, IN: Kellogg Institute for International
ability-the ability of parties to monitor constituents' Studies.
votes, reward them for their support and punish them Organization for Security and Cooperation in Europe. 2003. Russian
for defection-is what sustains machine politics. Federation State Duma Elections, 7 December 2003: Preliminary
Findings of the OSCE International Election Observation Mission.
Vienna, Austria: OSCE.
Piattoni, Simona, ed. 2001. Clientelism, Interests, and Democratic
REFERENCES Representation: The European Experience in Historical Perspec-
tive. Cambridge, England: Cambridge University Press.
Ansolabehere, Stephen, and James M. Snyder, Jr. 2002. Party Control Reynolds, John. 1988. Testing Democracy: Electoral Behavior and
of State Government and the Distribution of Public Expenditures. Progressive Reform in New Jersey, 1880-1920. Chapel Hill, NC:
Massachusetts Institute of Technology. Typescript. University of North Carolina Press.
Auyero, Javier. 2000. Poor People's Politics: Peronist Survival Net- Reynolds, John. 1980. "'The Silent Dollar': Vote Buying in New
works and the Legacy of Evita. Durham, NC: Duke University Jersey." New Jersey History 98: 191-211.
Press. Scott, James C. 1969. "Corruption, Machine Politics, and Political
Calvo, Ernesto, and Maria Victoria Murillo. 2004. "Who Delivers? Change." American Political Science Review 63 (December): 1142-
Partisan Clients in the Argentine Electoral Market." American 58.
Journal of Political Science (October): 742-57. Stigler, George J. 1975. The Citizen and the State: Essays on Regula-
Chandra, Kanchan. 2004. Why Ethnic Parties Succeed: Patronage and tion. Chicago: University of Chicago Press.
Ethnic Head Counts in India. New York: Cambridge University Stokes, Susan C. 1995. Cultures in Conflict: Social Movements
Press. and the State in Peru. Berkeley, CA: University of California
Chubb, Judith. 1982. Patronage, Power, and Poverty in Southern Italy. Press.
New York: Cambridge University Press. Stokes, Susan C. 2004. Is Vote Buying Undemocratic? Paper pre-
Cox, Gary W., and Matthew D. McCubbins. 1986. "Electoral Politics sented annual meetings of the American Political Science Associ-
as a Redistributive Game." Journal of Politics 48 (May): 370-89. ation, Chicago.
Diaz-Cayeros, Alberto, Beatriz Magaloni, and Barry Weingast. 2001. Tam, Waikeung. 2003. Clientelist Politics in Singapore: Selective Pro-
Democratization and the Economy in Mexico: Equilibrium (PRI) vision of Housing Services as an Electoral Mobilization Strategy.
Hegemony and its Demise. Stanford University. Typescript. University of Chicago. Typescript.
Dixit, Avinash, and John Londregan. 1996. "The Determinants of Tomz Michael, Jason Wittenberg, and Gary King. 2003. Clar-
Success of Special Interests in Redistributive Politics." Journal of ify: Software for Interpreting and Presenting Statistical Results.
Politics 58 (November): 1132-55. Version 2.1. Cambridge, MA: Harvard University, January 5.
Fox, Jonathan. 1994. "The Difficult Transition from Clientelism to Http://gking.harvard.edu.
Citizenship: Lessons from Mexico." World Politics 46 (January): Wilson, James Q., and Edward Banfield. 1963. City Politics.
151-84. Cambridge, MA: Harvard University Press.

325

This content downloaded from 129.96.252.188 on Tue, 05 Jan 2016 18:02:11 UTC
All use subject to JSTOR Terms and Conditions