The erosion of party hegemony, clientelism and portfolio

diversification: The Programa Nacional de Solidaridad

(Pronasol) in Mexico.1
Beatriz Magaloni, Alberto Diaz-Cayeros and Federico Estévez.2

1. Introduction

Interest in the study of clientelism has been renewed in the last few years.3 While

the theoretical frameworks produced predominantly by sociological and anthropological

studies of the 1960s and 1970s still provide important insights into the logic of the

exchanges involved in clientelism, a reckoning with the underlying political process that

makes those forms of political linkage so prevalent is in order.4 Much has happened in

the world and in social science scholarship since the early work. Clientelism was viewed

as a typical phenomenon of underdeveloped political systems, usually at early phases of

institutionalization, often under authoritarian or colonial regimes. In fact, the literature

1
Research for this paper was partially funded by the World Bank and the Academic Senate at UCLA.
Superb research assistance was provided by Lorena Becerra and Arianna Sánchez. The electoral database
was painstakingly compiled by Jacqueline Martínez. The database on Pronasol spending was compiled by
Marcela Gómez and Sandra Pineda. A previous version was presented at the 2002 Annual Meeting of the
American Political Science Association, Boston, August 29-September 1, 2002. We thank Robert Bates,
Herbert Kitschelt, John Londregan, Mona Lyne, Aaron Tornell and participants in workshops at Stanford,
Duke and Berkeley for insightful comments. The views expressed, as well as errors remaining are solely
the responsibility of the authors.
2
Estévez is Professor at ITAM, Instituto Tecnológico Autónomo de México, while Magaloni and Diaz-
Cayeros are Assistant Professors at Stanford University and also Adjunct professors at the Instituto
Tecnológico Autónomo de Mexico
3
To mention perhaps two of the most prominent examples, see Piattoni, Simona (2001) Clientelism,
Interests and Democratic Representation Cambridge: Cambridge University Press, with an emphasis on the
historical processes of European political development; and Kitschelt, Herbert (2000) “Linkages between
Citizens and Politicians in Democratic Politics” Comparative Political Studies 33(6/7) for an understanding
of clientelism in contrast to alternative linkage strategies.
4
The body of literature on clientelism is considerable. See Lemarchand, Rene and Legg, Keith (1972)
“Political Clientelism and Development: A Preliminary Analysis” Comparative Politics 4(2); Scott, James
(1972) “Patron-Client politics and Political Change in Southeast Asia” American Political Science Review
66(1); Lemarchand, Rene (1972) “Political Clientelism and Ethnicity in Tropical Africa: Competing
Solidarities in Nation Building” American Political Science Review 66(1); as well as the edited volumes by
Schmidt, S.W. et al. (1977) Friends, Followers and Factions: A Reader in Political Clientelism Berkeley:
University of California Press; Gellner, Ernest, and Waterbury, John (1977) Patrons and Clients London:
Duckworth; and Eisenstadt, S.N and Lemarchand, Rene (1981) Political Clientelism, Patronage and
Development London: SAGE Publications.

1
suggested that clientelism was the most characteristic form of political exchange

occurring in agrarian backward societies. Presumably, as societies became more

developed, social structures more differentiated, and political systems more

institutionalized, clientelism was bound to disappear. The world today is more

democratic than four decades ago, nascent political systems in the 1960s are today highly

institutionalized polities; and social, economic and political development has continued

apace around the world. Yet, clientelism has not disappeared throughout most of the

developing and large regions of the developed world.

A central feature of clientelism is that it involves direct exchanges between

patrons and clients, where votes or political support are traded for goods. Under what

conditions do politicians attempt to buy votes through the provision of particularistic,

excludable private goods, rather than through universalistic, non-excludable public

goods? This paper discusses what the literature has variously labeled as clientelism,

patronage, pork-barrel politics or spoils within the context of the development process.

All of these phenomena share a common element of what Kitschelt calls a “clientelist

linkage”, since politicians offer exchanges of goods for votes.5 However, we argue, that

the peculiarity of clientelism lies in the nature of the goods provided, the enforceability of

the implicit contract in the exchange, and its place within an overall strategy for

garnering electoral support from various voter groups (core, swing or opposition). We

suggest that the key to understand clientelism is to view it as one among a set of

alternative electoral investment choices politicians can make.

We propose a theoretical model and provide evidence for it in the context of the

erosion of hegemonic party rule in Mexico, employing municipal level data for the
5
Kitschelt (2000), op. cit.

2
Programa Nacional de Solidaridad (Pronasol) expenditures during the Carlos Salinas

presidency (1988-1994). Created in 1989 after one of the most contested and

controversial presidential races in the Partido Revolucionario Institucional (PRI)’s

history, Pronasol’s stated objective was poverty relief. The program was, however, meant

to halt the decline of the PRI’s electoral hegemony.6 The relative allocation of Pronasol

funds between private goods targeted to core clientelist supporters and public goods

benefiting a wider range of voters, including opposition backers, we argue, depended on

the specific conditions of party competition characterizing each municipality.

Clientelism was most prevalent in municipalities with high electoral risk and

multipartisan patterns of competition, controlling for levels of development. While

previous studies of clientelism have been unable to untangle the effects of party system

configuration from those of economic development, since both tend to be correlated

through time in the course of the modernization process, our dataset allows us to separate

the effect of socio-economic modernization from those generated by political influences.

In terms of modernization, we find that clientelism was most prevalent at intermediate

ranges of development.

6
See Molinar, Juan & Weldon, Jeffrey (1994) “Electoral Determinants and Consequences of National
Solidarity” in Cornelius, Wayne, Craig, Ann & Fox, Jonathan (eds.) Transforming State-Society Relations
in Mexico: The National Solidarity Strategy La Jolla: Center for U.S.-Mexican Studies, UCSD; Dresser,
Denise (1994) “Bringing the Poor Back In: National Solidarity as a Strategy of Regime Legitimation” in
Cornelius, Wayne, Craig, Ann & Fox, Jonathan (eds.) op. cit; Dresser, Denise (1991) Neopopulist Solutions
to Neoliberal Problems: Mexico’s National Solidarity Program. Current Issue Brief Series, no.3. La Jolla:
Center for U.S.-Mexican Studies, University of California, San Diego; Bruhn, Kathleen (1996) “Social
Spending and Political Support: The ‘Lessons’ of the National Solidarity Program in Mexico” Comparative
Politics 26(January):151-177; Gómez López Marcela & Pineda Antúnez, Sandra (1999) El Reparto
Municipal del Pronasol: Criterios de Asignación en Aguascalientes y Michoacán B.A. Thesis, ITAM,
Mexico; Hiskey, Jonathan (2000). “Does Democracy Matter? Electoral Competition and Local
Development in Mexico”, paper presented at the 2000 Latin American Studies Association Meeting; and
Magaloni, Beatriz, Estévez, Federico and Diaz-Cayeros, Alberto (2000) “Federalism, Redistributive
Politics and Poverty Relief Spending: The Programa Nacional de Solidaridad in Mexico (1989-1994)”
Paper delivered at the 2000 Annual Meeting of the American Political Science Association, Washington,
D.C.

3
 Employing local level data from 2417 municipalities in twenty programs

encompassed within Pronasol, we operationalize the definition of clientelism as

expenditures in divisible excludable goods delivered to individuals or to organized

groups of producers, indians and women. Examples of clientelistic transfers include

selective scholarships, credit, granaries, and livestock, among others. In our

operationalization, clientelism is distinguished from expenditures in what the literature

calls pork-barrel projects, namely local public goods, including public projects that were

limited in their impact to the local jurisdiction, such as electricity, drinking water, sewage

and the like. Public goods also include projects that span beyond the municipality,

including highway construction and regional hospitals.7

The theoretical model we propose builds upon the theory of portfolio

diversification from finance, in order to explain the supply of clientelism. Although

explaining the demand for clientelism is just as important, our framework takes that

demand as given, concentrating on clientelism as a strategic choice made by politicians.8

The gist of the model is the following: an incumbent party must decide how to allocate a

basket of discretionary transfers to voters in an attempt to get reelected. The basket may

include transfers that range from private, excludable outlays that can be individually

targeted; to non-excludable public goods that are targeted to a jurisdiction or to be

consumed by voters across several jurisdictions. As different instruments of electoral

investments, these transfers differ in a) their relative budgetary cost; b) their expected

7
We do not include public employment in our classification, which the literature discusses as patronage or
spoils, given that Pronasol incorporated community work in its projects, rather than being a job creation
program. In keeping with our definition of clientelism, had Pronasol encompassed job programs, they
would be classified as clientelist, since they provide private, indivisible benefits.
8
We thank Bob Bates for clarifying the supply and demand aspects of clientelism. For a discussion of
clientelism as supply and demand based, see Shefter, Martin (1994) Political Parties and the State: The
American Historical Experience Princeton: Princeton University Press; and Piattoni (2001) op. cit.

4
electoral return, which is defined as the expected number of votes from a unit of transfer;

and c) their electoral risk. The model assumes a positive correlation between expected

return and electoral risk: risky investments yield higher expected electoral returns.

Risk varies according to the “publicness” of the electoral investment instrument.

Private, excludable transfers that can only be consumed by a party’s supporters are

riskless; while public, non-excludable goods that can be consumed by all voters,

regardless of party sympathy, are the most risky. The intuition behind this notion is that

the easier it is to target a transfer, the more an incumbent party can screen between

opponents and supporters, so as to increase the probability that benefits will be received

by those most likely to vote for it. Resource transfers that go all the way down to the

individual are also the most effective at securing low risk electoral support. Private,

material discretionary outlays such as land titles, access to credit, jobs, construction

materials, or direct income transfers, for example, generally tend to favor core supporters,

because the party possesses more information and organizational links to that core.

Although riskless, private outlays are not the ones that yield the highest electoral

return, since fewer voters can normally be targeted. This is because governments face

budget constraints. Private transfers can be extremely expensive, both due to the

transaction costs that must be overcome in order to be effectively targeted, and due to the

number of people that need to be included in order to garner a sufficiently high number of

votes to win an election. Public goods reduce transaction costs, and are more cost

effective per beneficiary. However, they are far riskier than private goods, because the

incumbent can invest in them without receiving, ex post, any electoral pay-off.

5
 The supply of clientelism is modeled as a budget allocation decision by an

incumbent party which seeks to minimize its electoral risk, subject to attaining an

expected threshold of electoral return.9 This leads the incumbent to diversify its portfolio

among the two types of transfers, simultaneously allocating private outlays to its “core”

constituency, and public goods to “swing” and opposition voters. In contrast to extant

literature on distributive politics, which predict that politicians will either allocate

discretionary transfers to core10 or swing voters,11 the model predicts “mixed policy

baskets” by which both core and swing voters are favored.12

When a political monopoly comes under threat, the model predicts that

incumbents will diversify their portfolio, employing other, more universalistic modes of

exchange, in an attempt to catch more votes. Hence, political competition is virtuous,

because it leads politicians to introduce universalistic forms of political exchange, in the

form of public goods.

However, as a corollary of the first prediction in the model, clientelism is not

likely to disappear with political competition, but will persist in a risk hedging role. An

incumbent confronting electoral competition seeks investments with higher yields; but in

9
We construct our theoretical model under the assumption that the incumbent party seeks to minimize
electoral risk. The assumption is quite realistic for understanding the behavior of hegemonic parties, which
normally win elections with margins that far exceed a majority of 50% plus one. In its heyday, the PRI in
Mexico used to win elections with margins as high as 80 or 90%, and the same has been true of hegemonic
parties elsewhere.
10
Cox, Gary, McCubbins, Mathew D. and Sullivan, Terry (1984) “Policy Choice as an Electoral
Investment” Social Choice and Welfare 1:231-242 and Cox, Gary & McCubbins, Mathew D. (1986)
“Electoral Politics as a Redistributive Game” Journal of Politics 48(May):370-89.
11
Lindbeck, Assar & Weibull, Jorgen (1987) “Balanced Budget Redistribution as the Outcome of Political
Competition” Public Choice 52(3):273-97 and Dixit, Avinash & Londregan, John (1996) “The
Determinants of Success of Special Interests in Redistributive Politics” Journal of Politics
58(November):1132-55.
12
For empirical evidence regarding core vs. swing voters see Magaloni et al, 2000 op. cit.; Schady, Norbert
R. (2000) “The Political Economy of Expenditures by the Peruvian Social Fund (FONCODES), 1991-95”
American Political Science Review 94(2):289-304 and Dahlberg, Matz and Johansson, Eva (2002) “On the
Vote-Purchasing Behavior of Incumbent Governments” American Political Science Review 96(1).

6
order to mitigate risks, it will retain part of its diversified electoral investment portfolio in

clientelistic transfers. A “safety first” principle prevails: incumbents take risks because of

the higher electoral returns they expect; but when risk increases, ceteris paribus -that is,

without changing the expected levels of electoral return-, incumbents should resort to

more clientelism. Individual risk aversion will determine the degree to which hedging

will be more or less prevalent.

The chapter is organized as follows. The next section discusses clientelism in the

context of party hegemony in Mexico, reviewing some of the insights that the literature

on clientelism can offer to understand the distribution of public funds to political

supporters. The section after that provides the portfolio model of political investment,

where clientelism is clearly distinguished from the provision of public goods. The fourth

section discusses the case of Pronasol in Mexico, in the light of the theoretical model.

The final section provides some concluding comments and an agenda for future research.

2. Sustaining Hegemony through clientelism: deterring exit by targeting

Clientelism is characterized by dyadic personal relationships that are asymmetric

but reciprocal: the patron delivers desired material benefits to its clients in exchange for

services and loyalty to the patron.13 In electoral politics, this form of personalistic linkage

translates into a direct exchange of private benefits and favors for votes. James Scott

argued that patron-client links were based on inequality, which arises from the fact that

“the patron is in a position to supply unilaterally goods and services which the potential

13
Scott, 1972 op. cit.; Lemarchand, 1972 op. cit.; Archer, Ronald (1990) “The Transition from Traditional
to Broker Clientelism in Colombia: Political Stability and Social Unrest” Kellog Institute for International
Studies, Working Paper # 140.

7
client and his family need for their survival and well being”.14 As a monopolist for

critical resources, such as protection, access to arable land, fertilizers, water and

irrigation, or education, the patron is in a position to exploit his market power and

demand compliance from those who wish a share of those goods. However, if the client

did not need these goods so badly, or if she had savings and alternative sources of income

to finance his needs, or if she could incur in the costs of exiting to another jurisdiction to

secure the needed services, the client would not succumb to the patron’s domination.15

The relationship in patron-client links is considered to be asymmetrical because

there is normally one patron and a multiplicity of clients, but reciprocal in that politicians

must deliver in order to sustain the support of their clienteles. In our view, clientelism

prevails in monopolistic political markets because it allows politicians to deter exit. By

delivering private as opposed to public goods, politicians can screen between supporters

and opponents, excluding from the stream of benefits those who throw their support to

another patron. Incumbents can thus deter voters from exiting to the opposition. To deter

exit, political monopolists must attempt to rely as much as possible on the provision of

excludable material benefits that allow screening. The easier it is to target voters with

excludable benefits, and the more closely a party can monitor the voters’ choices, the

more the deterrence logic applies.

To see why clientelism serves to maintain a political monopoly, imagine a voter

who faces the following option: support the incumbent party and receive transfers in the

14
Ibid. p. 125; emphasis added.
15
See Diaz-Cayeros, Alberto, Magaloni, Beatriz and Weingast, Barry (2002) “Democratization and the
Economy in Mexico: Equilibrium (PRI) Hegemony and its Demise” Stanford University (typescript).; and
Medina, Luis Fernando and Stokes, Susan (2002) "Clientelism as Political Monopoly." Paper prepared
delivered at the 2002 Annual Meeting of the American Political Science Association, Boston, August 29-
September 1.

8
form of jobs, income supplements, credit and the like, or opt for the opposition and

receive none of these desirable benefits. Unless the voter possesses alternative high

sources of income and is indifferent to these benefits, her rational strategy is to support

the incumbent, even if reluctantly. If most voters reason likewise, the political monopolist

remains in power. The dilemma voters face is that of coordination: if all could agree

simultaneously to vote against the incumbent, they could defeat it. But if voters can’t

coordinate, each one will fear to be the first one to defect and face punishment in the

form of lack of access to vital resources.

Diaz-Cayeros, Magaloni and Weingast provide such a deterrence game to explain

how Mexico’s PRI sustained its hegemony for so long. The deterrence game, they argue,

is the secret of what they call the “tragic brilliance” of the PRI: tragic in that it forced

voters to accept high levels of corruption and poor governance; brilliant in that it led

them to play an active role in sustaining the PRI’s hegemony.16

The basic logic of the deterrence game is the following: consider a locality in

isolation that prefers the opposition to the PRI. The locality must elect a local leader,

either the PRI candidate or an opposition rival. The national PRI, in control of the federal

government, then decides whether to reward or punish the locality through the provision

or withdrawal of budgetary funds. Diaz-Cayeros, Magaloni and Weingast show that, in

equilibrium, citizens are induced to support the PRI. The reason reflects the centralized

fiscal system where localities receive the lion’s share of their funds from the national

government controlled by the PRI. Although voters in the locality may prefer the

opposition, the PRI’s credible threat to punish implies that voters face the choice between

an opposition without funds and the local PRI with funds. In equilibrium, the PRI
16
Diaz Cayeros, Magaloni and Weingast (2002) op. cit.

9
maintains its monopoly at the local level, but not because people prefer it to the

alternatives.17

This type of hegemonic equilibrium is maintained through a clientelistic form of

political exchange. The nature of the clientelistic exchange is based, first, on the

monopolistic control of fiscal resources by the national PRI, and second, on the PRI’s

ability to target transfers so as to screen between supporters and opponents. This power

constitutes a credible threat of punishment for those who defect to the opposition.

Defecting to the opposition implies being excluded from the stream of benefits that the

PRI qua political monopolist controls.18

What is the difference between clientelism and other forms of democratic

exchange where politicians trade policies for votes? As in Medina and Stokes, we believe

that the main difference lies in that clientelism implies a credible threat of being excluded

from a transfer or stream of benefits if the voter chooses to vote for the opposition.19 It is

for this reason that we tend to correlate clientelism with the trade of excludable benefits

for political support. Public goods that are not divisible imply that a voter can support

whichever politician he chooses, and still benefit from such policies.

17
A second issue that Diaz-Cayeros, Magaloni and Weingast (2002) op. cit. model moves the voter’s
choice from the local representative to the national government. Why do districts, that is, choose to keep
the PRI in control of the national government? In the face of the PRI’s credible punishment, no district
wants to be the first to defect to the opposition. Without a mechanism to coordinate with a majority of
districts, voters in each district have an incentive to acquiesce and accept the PRI. The reason is that
districts face a common pool problem: voters in one district alone cannot affect the system. Only if voters
across a majority of electoral districts coordinate to elect the opposition to national office can they affect
the national system.
18
See also Stokes and Medina (2002) op. cit. for a similar argument on the nature of clientelistic exchange.
19
In a somewhat different argument, but where credibility figures prominently, Phil Keefer (2003)
“Clientelism, Development and Democracy” (Typescript: World Bank), suggests that clientelism emerges
because politicians fail to credibly commit to a promise of delivering goods equally to all voters.

10
 Robinson and Verdier20 provide a model where clientelism represents a solution

to this commitment problem. “By its very nature, since the law cannot be used to enforce

[clientelistic] political exchanges, they must be self-enforcing. The problem of credibility

is two-sided. Citizen/voters must indeed deliver their support, and politicians, once in

power, must pay for their support with the policies they promised” (p. 1) In their model,

the solution to the commitment problem is given by trading employment in the public

sector for political support.21 We agree with Robinson and Verdier (2001) in that the

commitment problem is central for understanding clientelistic ties. We believe, however,

that public jobs are only one of the possible instruments that politicians can use to deal

with this problem. In the analysis that follows, we implicitly assume that the more a party

can target transfers, the better it can solve the commitment problem.22

The PRI in Mexico could choose to target transfers to the individual, the local

jurisdiction, or not to target at all, by investing in public goods extending beyond the

locality. Public goods do not solve the commitment problem, because they generate non-

excludable and irreversible benefits. Local public works, however, are less risky than

public goods spanning beyond a single political jurisdiction. Local public goods allow the

20
Robinson, James and Verdier, Thierry (2002) “The Political Economy of Clientelism” Paper presented at
the Conference The Political Economy of Clientelism, Stanford University (May).
21
“We argue that the appeals of offers of employment in the bureaucracy is precisely that a job is a
credible, excludable and reversible method of redistribution which ties the continuation utility of a voter to
the political success of a particular politician” (Robinson and Verdier, op. cit. p. 2). The commitment
problem runs both ways: voters prefer employment offers because politicians can’t credibly commit to
deliver other polices once in office. In their model employment is a credible offer because politicians have
incentives to expand the public sector when they taxing the private sector is too costly. And politicians
prefer to offer jobs as opposed to other public policies because they can tie the continuation of employment
to their political success, creating incentives for bureaucrats to continue to support them.
22
Voters also face a commitment problem: suppose that a voter chooses a party on ideological grounds on
the basis of its campaign promises. Once in office, the party can renege, implementing a different set of
polices. See Stokes, Susan (2001) Mandates and Democracy: Neoliberalism by Surprise in Latin America
Cambridge: Cambridge University Press. Phil Keefer (2003) op. cit. models clientelism as a response to
this type of commitment problem: when politicians can’t credibly commit to deliver universalities
programmatic polices due to, for example, high levels of political instability or low institutionalization,
voters will demand private outlays (clientelism) before elections.

11
PRI to employ geographic targeting according to the landscape of political units as in

Diaz-Cayeros, Magaloni and Weingast (2002). Nonetheless, as opposed to particularistic

transfers, public works do not fully solve the commitment problem: once the party

transfers a public good to a locality, how can it make sure that all voters, including those

who prefer the opposition on ideological grounds, comply with their part of the

exchange? And once delivered, a public good cannot be withdrawn, as is clearly the case

with private resource transfers.23This is the reason why we believe public goods are

riskier than private outlays.

Private outlays such as jobs and other transfers can better solve the commitment

problem. A party can identify voters individually, screen between supporters and

opponents, and invest only in those core constituencies that will support it with certainty.

A party requires a dense organizational network to successfully deliver these transfers

and identify true partisans from all non-partisans who have incentives to misrepresent

their type. Historically, the organizational network that the PRI employed to deliver

private transfers ranged from party unions and local party bosses, to schoolteachers,

caciques (local bosses) and presidentes ejidales (the heads of the ejidos, a form of

communal landholding). The goods that the party distributed through these networks

ranged from land, subsidized credit and fertilizers, to subsidized food, scholarships, and

government-built houses, among others.

An alternative, much cheaper way a party can solve the commitment problem is

by violating the secrecy of the ballot. This illegal practice seems to be quite common in

the developing world. For years, the PRI openly violated the secrecy of the ballot. It

23
Diaz-Cayeros, Magaloni and Weingast (2002) do not discuss the commitment problem since they assume
that the locality coordinates in some way to reelect the PRI.

12
could get away with this practice because the opposition did not possess a sufficiently

dense organizational network to monitor elections. During the years of solid party

hegemony, the PRI ran uncontested in the overwhelming majority of municipalities -the

opposition did not even bother to send party activists to monitor the ballot. These

practices were more common in the countryside than in the cities, where the opposition

had more presence and voters were better educated.

However, voting is not the only behavior that is involved in the clientelistic

exchange. To fulfill their part of the agreement, citizens must support the party, which in

the context of the PRI hegemony sometimes meant participating in rallies and other

public events organized by the party, but more important, citizens had to refrain from

assisting to the rallies and meetings organized by the opposition. The party’s dense

organizational network in the countryside through the caciques allowed the PRI to

monitor these forms of political behavior.

Thus, the PRI did not coerce voters into choosing the ruling party over the

opposition. In smaller and isolated localities often there was not even a menu of choices,

only the PRI. And even if there was some opposition presence, often all that the PRI

really needed to do to get peasants to cast a vote for the party was to pay for their

transportation, because peasants’ “sincere” preference was the PRI. Peasants freely chose

the PRI, although their choice was constrained. On the one hand, the PRI could use its

monopolistic control over key resources such as land, fertilizers and credit to buy their

support; 24 and on the other, its network of party organizations and government agencies

24
Land reform in Mexico was implemented since after the revolution until 1992, when the constitution was
modified to end the process and grant peasants the possibility to legally own their land. Before 1992,
peasants had to petition the government to acquire land and once it was granted, they did not have the title
over it. This meant that they could not rent or sell their land. See Diaz-Cayeros, Alberto, Magaloni, Beatriz

13
permitted it to monitor the political behavior of its clients in the countryside and in small

cities. By threatening to withdraw the transfers that the peasants needed, the PRI thus

managed to deter rural voters from supporting another party or engaging in any form of

open confrontation with the regime.

The story of the larger and wealthier localities is different. Since the early 1950s,

the opposition had important presence in the larger cities. Mexico City, for example, was

the earliest opposition bastion25 Until the 1980s, the PRI enjoyed political support among

the working class affiliated with the official unions. It also attempted to build clientelistic

links among the migrant poor in the city slums and among informal sector workers.26 To

these groups, the party offered land titles, subsidized housing and food, and work permits

and licenses for selling merchandise in the numerous flea markets of the cities, among

other inducements. However, with the onset of the debt crisis, the urban poor were not

loyal to the PRI, as it became clear in the 1988 presidential elections, when they defected

en masse to Cuahutémoc Cárdenas. Even some sectors of organized labor, traditionally

unconditional in their support to the party, including, for example the powerful oil

workers union, supported Cárdenas in that election. Pronasol was designed largely as an

attempt to recover the support of these voters.

Why was the PRI much weaker in the larger and wealthier localities? One key

difference between the city and the countryside, we argue, is voter heterogeneity. The

and Weingast, Barry (2002) “Perverse Hegemony: Land, electoral support and growth in Mexico”
(typescript), for an analysis of the welfare implications of this political arrangement.
25
Ames, Barry (1970) “Bases of Support for Mexico’s Dominant Party” American Political Science
Review 64(1):153-167; Molinar Horcasitas, Juan (1991) El Tiempo de la Legitimidad. Elecciones,
Autoritarismo y Democracia en Mexico Mexico: Cal y Arena; Klesner, Joseph (1996) “¿Realineación of
Desalineación? Consecuencias de la Crisis y la Reestructuración Economica para el Sistema Partidiario
Mexicano in Maria Lorena Cook, Kevin Middlebrook and Juan Molinar (eds.), México: Cal y Arena.
26
Cornelius, Wayne (1975) Poverty and Politics of the Migrant Poor in MexicoCity Stanford: Stanford
University Press.

14
overwhelming majority of the so-called urban poor that abandoned the PRI in the 1988,

for example, came from the low-skilled service sector of the economy –taxi and other

public transportation drivers, domestic employees, low-level bureaucrats, nurses, etc.

There was no functional party organization that could encompass such dissimilar groups

largely because these groups had little goals in common. Without an efficient party

organization, the commitment problem became pervasive.

The second difference between city and country is related to income levels.

Richer voters are much less susceptible to vote buying.27 Diaz-Cayeros, Magaloni and

Weingast provide a rationale for the correlation between affluence and the emergence of

political competition in the context of the deterrence game of party hegemony. In their

view, modernization contributes to undermine party hegemony because it makes

clientelism more expensive, as it is much harder to continue to deter wealthier localities

from voting for the opposition.28

Thus, highly marginalized rural localities kept the PRI’s hold of a political

monopoly because they were cheap to deter through clientelism. In the absence of

political competition, the PRI faced no incentives to invest in growth-enhancing public

goods. Political competition, we posit, generates incentives for politicians to diversify

their portfolio in an attempt to simultaneously halt the deterioration of their core and

capture swing and opposition voters, even while economic development might remain

unchanged. The model below seeks to provide an understanding of how a political

monopoly under threat might respond to electoral competition.

27
See Dixit and Londregan (1996) op. cit.
28
Modernization is not a sufficient condition for bringing hegemony to an end. In Diaz-Cayeros, Magaloni
and Weingast (2002), op. cit., modernization promotes unilateral defections to the PRI among wealthier
localities. Democratization requires, however, massive coordination among rich and poor localities in the
form of a tipping game.

15
3. Monopoly under threat: a portfolio diversification strategy

The choice of clientelistic strategies is driven by both demand and supply factors.

The most important demand factor stressed by the literature is the economic status of

citizens, which might make them more or less willing to accept this type of exchange. If

voters have an income elasticity of public good demands larger than one, they will prefer

to have less clientelism delivered by governments as they become richer.29 Other factors

on the demand side might be highly correlated with economic status: cognitive

capabilities depend on literacy rates; while organizational capabilities depend on

membership in voluntary and independent associations, which might be highly correlated

with income levels. Thus, a socioeconomic theory of clientelism is primarily a demand

based account.

A supply based account stresses “circumstances that may induce party leaders to

adopt clientelism or patronage as a strategy for attracting voters, supporters, and activists

to their side”.30 The existing literature stresses the lack of an independent bureaucracy

and the motivations that historically led politicians or parties to mobilize voters through

clientelist inducements. In formal models of clientelism, the supply side is determined by

monopoly over the control of a valuable resource, such as a riskless job or a technology.31

This paper models the supply side as a budgetary decision by a risk averse politician

seeking to achieve a given level of electoral support.

29
We thank Robert Bates for this simple formulation.
30
Piattoni (2001); op. cit. p.17
31
See Robinson and Verdier (2002) op. cit.; and Medina and Stokes (2002) op. cit.

16
 The model is derived from a portfolio diversification approach to electoral

investment.32 Politicians buy votes in order to stay in office. Incumbents, especially

entrenched hegemonic parties, are risk averse. They would rather invest resources in

private transfers targeted to core voters, than on uncertain public goods. In their quest for

a high electoral return, however, they shift their electoral investments into public goods

that offer high electoral returns, in spite of the risks involved.

A diversification logic of “safety first” suggests, however, that a party will not

devote all of its financial resources to the provision of public goods. Since public goods

are non-excludable, the benefits of their provision impact both supporters as well as

adversaries of the party. The electoral return of public goods can be high, given that they

benefit a larger group of voters. However, that return is highly uncertain (i.e. risky).33

Private goods, in contrast, are safe bets: they assure, through their compliance

mechanisms, that the recipients will support the incumbent party. 34 In our model private

good provision is equated with clientelism.

The problem of finding the optimal political allocation of public funds from the

incumbent’s point of view can be conceived as a decision over the relative allocation of

32
Federico Estevez, Beatriz Magaloni and Alberto Diaz-Cayeros (2002) “A Portfolio Diversification
Model of Electoral Investment” (typescript, Stanford University); Alberto Diaz-Cayeros (2001) “The
Geography of Electoral Risk: A Portfolio Diversification Model of Fiscal Transfers” (typescript, UCLA).
33
One discussion of clientelism and public goods provision in Italy along these lines, in a speculative rather
than a theoretical vein, can be found in Piattoni (1999) “Politica locale e sviluppo economico nell
Mezzogiorno” Stato e Mercato 55(April):117-149.
34
In both cases there are credibility issues involved. The delivery of private or public goods is subject to
opportunism, since the incumbent might not comply with its promise. If the vote follows the provision of
the good, the problem of credibility is on the voter’s side: there might be incentives to not deliver the vote
once the good has been delivered, specially if monitoring of electoral support is relatively difficult (for
example, with an enforced secret ballot or in communities that are large enough or anonymous enough so
that it is difficult to know who voted in what way). In what follows we set credibility issues aside,
assuming that there is a mechanism through which the promises can stick on both sides of the electoral
exchange (including repeated interaction, non-reversibility or the threat of withholding of resources
controlled by the patron). For discussions of the credibility issue see James Robinson and Thierry Verdier
op. cit. and Stokes and Medina op. cit.

17
funds among particularistic (private) and collective (public) goods projects. Figure 1

depicts such choice, in terms of a public good which yields an uncertain electoral return,

described by the expected vote value E[X] with known variance σ2; and a private good,

with a smaller, but certain, electoral return denoted Y<E[X]. 35 The model is silent as to

the developmental impact or social welfare generated by these two forms of expenditure.

What matters to politicians is how many votes they can get from alternative uses of

public funds. A crucial assumption in this framework is that the electoral return of

clientelism is not as high as public good provision. The problem with public goods is that

they do not embody monitoring and compliance mechanisms that ensure voter’s support,

the way clientelism does. Hence, while the vote return of public good investments is a

random variable, with uncertain realizations (although an ex ante known mean and a

variance) the vote return from clientelism is assured. The two values are expressed in

terms of the votes that an allocation of all the available budget (B) to each type of good

would provide.

Hence, the electoral returns Y and E[X] incorporate both a budget constraint and

the relative cost of public vs. private good provision. Given a fixed budget B, that can

finance ny private transfers with a unitary cost cy for each beneficiary if there are no

transaction costs and commitment problems in the clientelist exchange, ny voters would

support the party for sure, so that when B=nycy; Y=ny=B/cy. This means that the vote

return of clientelism is given by the budget divided by a fixed unitary cost of each private

transfer. For example, the unitary cost could represent the market price of a sack of grain,

and the assumption in this framework is that the incumbent can be sure of receiving as

35
We thank Aaron Tornell for suggesting this depiction of the portfolio problem.

18
many votes as sacks of grains it distributes, given its budget. Those sacks would be

distributed through the clientelistic networks already in place.

Of course that the effectiveness of private transfers depends on the relative

propensity of voters to exchange their votes for money. If voters are very poor, it is likely

that a small private transfer, such as a sack of grain, is enough to tilt their voting choice.

When voters are relatively rich, it might be that voters need more expensive private

transfers, such as agricultural credits or title deeds. The size of the budget limits the

provision of such private goods.

The budget could alternatively be allocated to a public good with total cost Cx=B. The

number of voters that support the party with this strategy is uncertain (and might depend,

for example, on a complex relationship given by heterogeneous public good demand

functions). We depict a reduced form expression of this relationship, which simply says

that the electoral yield of public goods, X is a random variable that depends on an

aggregate propensity to support a party that provides a public goods, with a known

variance. Figure 1

Public E[X], σ2
Good

Y
Private
Good

19
 When budgets (or public goods) are not divisible, the choice theoretic problem is

then to decide whether to provide public or private goods. This choice would depend on

the degree of risk aversion characterizing an incumbent: even when public good

provision has an expected value greater than that of the private good, an incumbent

might still choose to go for the safety of private good provision. Of course, the return of

private goods should be large enough to satisfy the minimum votes necessary to keep the

incumbent in office. Such condition depends on how cheap it is to buy votes from core

constituencies. If voters are not very poor, the available budget might not be enough to

buy enough core voters to stay in office, and the incumbent will be compelled to take

some risks. Hence, the most crucial feature of the socioeconomic theory of clientelism,

namely the association between poverty and clientelism, is accounted in this model by the

demand side assumption that it is cheap to buy votes from poor people.36

A dichotomic choice presumes the impossibility of dividing up budgets among

alternative uses. However, budgets (and the choice of different public goods) are

divisible, and incumbents can consequently be better off by combining both clientelism

and public good provision provided they have a preference over risk. If incumbents were

risk neutral, they would obtain no advantage from diversification, because they have no

use for risk hedging.

We assume that incumbents seek to obtain a given vote level that ensures their

permanence in power, with the least possible risk. Their optimal strategy is thus to find a

36
In the depiction of the problem as a dichotomic choice, public goods would be provided, even when
clientelism keeps the incumbent in office for sure, provided that the incumbent is not too risk averse and
prefers more, rather than less votes. This feature means that even if clientelism might ensure electoral
success, the gamble of public good provision might be preferred by the incumbent, to the extent that the
public good provision is more cost effective, per expected vote, than clientelism.

20
diversified allocation of funds between public and private goods, where they devote a

proportion α of the budget to public goods, and the remainder 1-α to private ones. This

strategy yields a higher overall return, taking advantage of the electoral opportunities

afforded by public good provision, while hedging risks through an optimal combination

with the risk free investment.37

The problem for the incumbent can then be reformulated into that of finding a

combination of public and private goods (α) that minimizes risks (the variance of votes),

given the constraint of a desired level of expected electoral support. The vote constraint is

given by:

V= α E[X] +(1- α)Y (1)

Where V is the exogenously desired level of votes. This desired level of votes

could well be beyond 50 percent.38 Risk is measured by the variance of total vote:

S= α2σy2+2α(1- α)σxy +(1- α)2σx2 (2)

Since private goods are assumed to provide a constant electoral return, the

variance of private goods and its covariance with public ones is zero, σx=σxy=0, which

means that the variance of total vote is only the first term in (2), the variance of the public

goods, discounted by its (squared) share in the portfolio.

A constrained maximization of (2) given (1) yields:

α*=λ(E[X]-Y)/2σ2 (3)

37
A mixed portfolio always involves less risk (except when the covariance of both goods is 1).
38
For example, in the hegemonic era of the PRI in Mexico, governors who could bring in more votes from
their states could signal to the president that they were more “competent” in their jobs. The PRI rewarded
such competence with longer and more attractive political careers –to governors, the reward was a
“promotion” to the federal cabinet after their terms ended or even before that, as there is no immediate
reelection in Mexico for any of the elective office.

21
where λ is a Lagrange multiplier, denoting the degree of risk that is acceptable.39 This

expression means that the optimal portfolio of electoral investment depends on individual

risk aversion, the variance of public good returns and on how large the gap between the

electoral return of public vs. private goods is.

Hence, ceteris paribus, the comparative statics of this expression suggest that the

proportion of private good allocations (clientelism) will be higher:

1. The smaller the difference in yield between the two goods

2. The greater the risk of the public good

3. The higher the politician’s risk aversion

Figure 2 depicts the solution to the problem in a standard mean-variance space,

showing a specific allocation of the electoral investment portfolio, given by an

indifference curve.40 The space depicts two goods, labeled as clientelism and public

good, according to their electoral yield and variance. The variance of clientelism is zero,

but the votes it can provide are less than those of the risky investment in a public good.

The difference between E[X] and Y in the vertical axis represent result 1 in the

comparative statics. As that gap becomes greater, clientelism is less attractive. The

variance in the horizontal axis represents result 2. As the variance is higher, clientelism is

more attractive as a “safety first” resource. The line linking the public good and

39
Since λ measures the sensitivity of the objective function to changes in the right hand side of the
constraint, that is, how much risk (standard deviation) the incumbent is willing to trade-off with changes in
expected votes. Such a preference is embedded in a utility function.
40
To simplify the exposition we have not introduced a utility function, which gives closure to the formal
model. The mean variance space and risk aversion interpretation of the utility function is a well known
result, dating back at least to Roy, A.D. (1952) “Safety First and the Holding of Assets”, Econometrica 20
(3):431-449. The simplest utility function that yields a mean-variance space like the one depicted in Figure
2 is a quadratic one. While economists dislike this functional form, in politics its properties are rather
reasonable: at some point the marginal utility of some extra votes is negative, an assumption which is not
very reasonable for money, but it is for votes. See Hirschleifer, Jack and Riley, John (1992) The Analytics
of Uncertainty and Information Cambridge: Cambridge University Press.

22
clientelism denotes all the possible combinations that produce intermediate risks and

returns. Every point in the line yields a higher return in expected value terms than

clientelism. In that sense, if politicians care about risk, diversification is always better

than solely distributing private goods. Every point in the line yields a lower risk than the

public good, so diversification in attractive on the grounds of risk hedging. The specific

solution to the composition of the expenditure portfolio depends on the curvature of the

indifference curve (risk acceptance, related to λ in the comparative statics), and the slope

of the line (which depicts the relationship between risks and returns).

Figure 2

Public Good
E[X]

α*

Y
Clientelism

Σ σ2

To sum up, the model suggests that in the process of modernization clientelism

might be the exclusive form of political linkage only in very poor environments, where it

is very cheap to buy votes and it is possible to win an election purely on the basis of core

voters. As the cost of buying votes through clientelism increases, diversified portfolios of

23
budgetary allocations should be observed, where public goods are combined with private

provisions to core supporters. Hence, when political competition is more stringent, so that

core support does not assure victory, one should observe combinations of public and

private good provision. Clientelism should not be abandoned altogether, however, due to

the risk-hedging role it plays. The relative importance of clientelism would depend, in

this context, on poverty, which would make it more prevalent; on political competition,

which would work at increasing public good provision; and electoral risk, which would,

other things equal, make clientelism more attractive. If we were able to measure

politicians attitudes towards risk, our model suggests that more risk averse politicians

would be more likely to provide higher shares of clientelism in their portfolio mix. The

next section provides some evidence regarding the choice of clientelism as response to

electoral risk in the case of Pronasol in Mexico.

4. Clientelism and public goods in Mexico: The case of Pronasol

Pronasol was organized around 20 programs, each of them geared towards various

provisions of private or public goods.41 The coverage of Pronasol was so extensive that

all municipalities in Mexico received some program every year, although the composition

by programs varied widely from year to year and among municipalities. If one conceives

of a continuum of the types of projects where Pronasol funds could be invested, at one

extreme we find private goods, which are targeted to specific individuals; while at the

41
Analysis of individual programs can be found in Gershberg, Alec Ian (1994) “Distributing Resources in
the Education Sector”in Cornelius, Wayne, Craig, Ann & Fox, Jonathan (eds.) Transforming State-Society
Relations in Mexico: The National Solidarity Strategy La Jolla: Center for U.S.-Mexican Studies, UCSD;
Jonathan Fox (1994) “The Difficult Transition From Clientelism to Citizenship: Lessons from Mexico”
World Politics 46(2), for Indian communities; Magaloni, Estévez and Diaz-Cayeros (2000), op. cit. for,
sewage, highway construction and funds for urban infrastructure improvements. The source for the
expenditure data is the Secretaría de Desarrollo Social (1994) Hechos en Solidaridad CD-ROM.

24
other extreme there are non-excludable public goods, spanning large regions of the

country. Among the most private of the goods the program selectively provided, one

might mention scholarships to children and adults, title deeds for land regularization, or

housing improvements. Among the most public, one can think of hospitals. Most of the

programs fell in between, with various mixes of more private or public good provision,

according to the specific issue area they were geared towards, and their target population.

By breaking down each program into the specific goods provided, we were able to

classify the money spent according to two categories, consonant with the portfolio

allocation model. The first are private goods, which we identify with clientelism, where

we included strictly excludable goods delivered to individuals and organized groups of

producers, indians and women. For public goods we included both projects that were

limited in their impact to the local jurisdiction as well as projects that spanned across the

municipality and beyond. Clientelism in Pronasol spending is measured through the share

of private goods in total spending and the per capita allocation of private goods. 42

Throughout the life of the program, clientelism constituted 28 percent of the funds

received by an average municipality. As the Program became consolidated throughout the

years, it became more clientelistic: when the Program started in 1989, 25 percent of the

funds distributed to the average municipality constituted private transfers, and by 1994

this percentage increased to 35. Figure 3 maps the average share of private goods in total

expenditure during the six years of the program, for all the municipalities in Mexico. The

42
The classification of each program is as follows: 1) private goods: Support to social service, housing,
children in solidarity, women in solidarity, infrastructure for productive support, solidarity for production,
productive ecology and regional development programs (indian communities); and 2) public goods:
drinking water and sewage, food and distribution, electricity, sports infrastructure, dignified school,
urbanization, educational infrastructure, IMSS solidaridad (health clinics in remote areas), health support,
municipal funds, regional hospitals and highway infrastructure.

25
darkest areas represent private good shares of more than 40%. It is obvious that

clientelism was practiced throughout the country. However, notwithstanding a common

view that characterizes all of Pronasol spending, and generally government spending in

Mexico during the era of PRI hegemony, as clientelist, the map demonstrates that vast

areas of the country were not primarily clientelist in the orientation of public good

provision. Clientelism was concentrated geographically in the Yucatán peninsula (the

states of Yucatan, Campeche and Quintana Roo), and a strip of states in the North lying

around the tropic of Capricorn (Sinaloa, Durango, Zacatecas and Tamaulipas). The

border with the US was not particularly clientelistic. High clientelism allocations were

also scattered in the center of the country around Mexico City, in the state of Chihuahua

to the North, in the oil producing state of Tabasco, and in the Western highlands (what is

also known as the Bajío region). Low shares of clientelism are observed in municipalities

close to the US border, the coastline, and some (though not all) of the poor southern

states (i.e. Chiapas). The pattern is thus not a North South divide in which rich places

receive larger shares of public goods, while poorer ones are characterized by clientelism.

26
Figure 3. The geographic distribution of clientelism in Mexico.

The overwhelming majority of the municipalities were provided shares of private

goods below 40 percent. This suggests that the PRI was providing public goods through

Pronasol as a strategic effort to reach out to voters beyond its core clienteles. Most

municipalities received combinations of goods, consonant with a portfolio diversification

logic.

Diversification strategies can be distinguished according to two dimensions of

interest: the poverty levels of municipalities as a measure of socioeconomic

modernization, and the competitiveness of their party systems, as a measure of electoral

considerations. Below we explore both of these issues.

27
 Figure 4 shows the average per capita allocation of PRONASOL in private goods

(clientelism) depending on economic development levels and partisan configurations.

Development is measured through the deprivation index from the Consejo Nacional de

Poblacion (CONAPO), which is constructed with a factor analysis of census variables

commonly associated with deprivation (illiteracy, no elementary school, dwellings

lacking access to drinking water, sewage and electricity, density of inhabitants and

quality of housing construction, population living in rural localities, and workers earning

less than two minimum wages).43 The figure reports how clientelism varies at different

levels of development in municipalities characterized by various partisan configurations.

We distinguish between municipalities where there was no electoral competition (the PRI

received 100% of the vote); hegemonic municipalities, where there was some opposition

presence, but the effective number of parties (using the Laakso-Taagepera index44) was

less than 1.7; and competitive localities where there were two or more effective political

parties (an index N of more than 1.7).

In terms of economic development, the graph suggests that clientelism exhibits an

inverted J-shape relationship, which is striking from the point of view of a modernization

account. Regardless of party configurations, clientelism tends to be greatly eroded at the

highest level of development (localities showing a deprivation index of 1, which

represent only 5% of our observations). This suggests that, consonant with the

socioeconomic theory of linkage building, rich voters much prefer public good provision

over private transfers, which makes it too expensive for a party to attempt to buy them off

43
Consejo Nacional de Población (1993) Indicadores Socioeconómicos e Indice de Marginación
Municipal, 1990 Mexico: CONAPO-CNA
44
Laakso, Markku and Taagepera, Rein (1979) “Effective Number of Parties: A Measure with Application
to West Europe” Comparative Political Studies 12:3-27.

28
through particularism. However, clientelism is most prevalent in middle-range levels of

development (deprivation index of 4, 3, and 2, which represent close to 80% of our

observations). In the poorest localities (14% of our observations at deprivation index 5)

clientelism is higher than in the riches ones, but lower than in the intermediate ones. The

figure thus suggests that voters in semi-urban localities and smaller cities are highly

susceptible to vote buying, and that modernization does not erode clientelism unless it

surpasses a sufficiently high threshold.

Figure 4

Effects of modernization and electoral competition on clientelism

100

90

80

70

60
per capita pesos

no competition
50 hegemonic
competitive
40

30

20

10

0
5 4 3 2 1
Level of development (inverse poverty scale, 5 is poorest)

The figure also shows how political competition impacts clientelism. As we

expected, holding development levels constant, political competition induces politicians

to invest more in public good provision in an attempt to cater a wider and more

29
heterogeneous public. constituencies. Note that at high levels of electoral competition,

clientelism is abandoned very quickly: the inverted J almost becomes a downward

sloping curve. Figure 4 also highlights that non-competitive municipalities receive more

clientelism. This might indicate, as argued above, that clientelism helps politicians keep

their electoral monopolies and clienteles in place.

Electoral competition induces investment in public goods. However, a question

remains as to whether electoral competition is responding to economic development,

lacking an independent effect. Our dataset allows us to separate the socioeconomic from

the political processes that influence clientelism. Development is correlated with political

competition. However, the correlation between the deprivation index and the effective

number of parties is negative, but not too strong (-0.37), which means that we can find

poor localities with significant party competition and rich localities with no competition.

Table 1 shows how party system configurations relate to the CONAPO

deprivation index. Among the multiparty configurations, the table distinguishes between

bipartisan (N between 1.7 and 2.3) and multipartisan ones (N greater than 2.3)Indeed, the

richest municipalities (CONAPO index of 1) tend to have more bipartisan and

multipartisan configurations, but there are quite a few highly developed municipalities

which are hegemonic. By the same token, the poorest municipalities (CONAPO index of

5) tend to have less competition, but there are many very poor localities that are

competitive. Municipalities at middle-range levels of development exhibit almost an

equal chance of being hegemonic or bipartisan. Multipartisan configurations are the least

likely at all levels of development, although they do tend to concentrate in the richest

municipalities.

30
Table 1: Party System Configurations and Socioeconomic development (percentages)
Marginality Nonompetitive Hegemonic Bipartisan Multipartisan Total
Index N=1 1>N>1.7 1.7>N>2.3 N>2.3
Very High 27 12 7 7 14
(5)
High 44 37 26 15 34
(4)
Medium 16 22 21 21 20
(3)
Low 13 25 38 44 27
(2)
Very low 1 3 8 14 5
(1)
Total 100 100 100 100 100
NOTE: CONAPO index measures marginality, so 5 is poorest, 1 is richest.

Thus, electoral competition and development, although correlated, are two clearly

distinct variables. Holding political competition constant, there is more clientelism at

middle-range levels of development; and holding development constant, there is less

clientelism as political competition increases. An additional political variable that our

portfolio model stresses is electoral risk. Our expectation is that political competition

should lead politicians to diversify their portfolios, introducing more public good

provision in an attempt to cater to a wider and more heterogeneous electorate.

Nonetheless, as public good provision is accompanied by higher risk, we expect

politicians to attempt to hedge these risks by disproportionately investing in clientelism in

the riskiest localities, holding levels of support constant.

Measuring risk in each municipality is not straightforward. One possibility is to

measure the standard deviation of PRI support. However, such measure is highest in

31
those municipalities where PRI support has been highest. Given the general trend for

convergence in levels of electoral support, a standard deviation measure would make the

politically most backward municipalities seem to be the most risky. The measure of risk

we use instead, drawing from the finance literature, is the systematic risk for each

municipality, controlling for electoral risk at the national level. Systematic risk is

calculated through what the finance literature calls a beta coefficient,45 for each

municipality, regressing municipal PRI vote share on the national PRI vote share.

We measure risk as the coefficient of the independent variable in a linear

regression of the form Y=α+βX, where X is the national support for the PRI since 1970,

and Y is the support in each municipality. Depending on the staggered electoral calendar

of municipal elections, the number of observations is 6 or 7. National PRI support is

calculated for each year according to the elections taking place in that particular year.

This means that the national vote trend for the party depends on the specific states that

held elections that year. Such calculation allows for a comparison that discounts the

shocks that might occur to the support of the party nationally, from the risks which are

specific to the locality. It also separates a component of vote volatility, the non-

systematic risk (which would be measured by the variance of the error term in the

regression). Politicians cannot predict non-systematic risk, since it depends on random

events that are statistically speaking, “noise”. Hence, they should not concentrate simply
45
For the seminal work introducing this concept see Sharpe, W. (1964) “Capital Asset Prices: a Theory of
Market Equilibrium Under Conditions of Risk” Journal of Finance 19:425-42. A huge discussion emerged
from the empirical work. See in particular, Fama, Eugene and French, Kenneth (1996) “Multifactor
Explanations of Asset Pricing Anomalies” Journal of Finance 51:55-86. A good textbook discussion is
Bodie, Zvi, Kane, Alex and Marcus, Alan (2001) Investments Fifth Edition. Boston: Mc-Graw-Hill. For
one of the few applications of beta coefficients as risk measures in political science, see Crain,
Messenheimer and Tollison (1993) “The Probability of Being President” The Review of Economics and
Statistics. 75(4):683-689. For an application to Mexico see Alberto Diaz-Cayeros, Beatriz Magaloni and
Federico Estevez (2003) “Electoral Risk and Redistributive Politics in Mexico and the United States” Paper
presented at the Midwest Political Science Association Meeting, Chicago, April 3-6.

32
on how volatile vote shares are, but rather on their systematic behavior in comparison

with national trends.

Table 2 reports the average beta coefficients calculated for Mexican

municipalities according to the CONAPO classification of level of development and the

partisan configuration given by the effective number of parties. Any coefficient above 1

implies that the municipality is more risky than the national electoral trends. Places with

risk below 1 compensate the national trends. Risky places, instead, would constitute

attractive places to channel financial resources to the extent that they have high expected

vote shares, or are pivotal to win an election.

Table 2. Beta Coefficients According to Party System and Development

Development No Two
index competition Hegemonic Party Multiparty Total
Very high
marginality 0.182 1.011 1.855 2.212 0.729
High
Marginality 0.336 1.096 1.854 2.193 1.038
Medium
marginality 0.639 1.122 1.841 2.244 1.303
Low
Marginality 0.780 1.191 1.626 1.845 1.356
Very low
marginality 0.393 1.081 1.216 1.497 1.178

On average, municipalities were more risky than the national level, since the

average beta coefficient for all municipalities was slightly above 1 (β=1.15).

Furthermore, risk is linked to partisan configurations: more competitive municipalities

are riskier. Yet, as can be inferred from the table, risk and party configurations are

different measures as there is high variance in the level of risk even among municipalities

of same party configuration. Starting from the last column, which shows the average risk

33
regardless of party system, only the very poor municipalities show a behavior that can

hedge against the national trends: in those places, given that PRI support has remained

high, even as it falls elsewhere, the beta coefficient is less than one. High marginality

areas show the same trend as the country as a whole. The next least risky municipalities

are the richer areas of the country. This is probably due to the fact that competition has

become more stabilized in two party and three party systems, where the PRI is capable of

reversing some of the national trends. The biggest collapses, and highest risks, were faced

in bipartisan and multipartisan races, at intermediate and low levels of development.

Thus, electoral risk cannot be conceived as a consequence of economic

development. Developed regions are more risky than the poorest areas; but when partisan

configurations are taken into account, developed areas in the country with stable

bipartisan and multiparty electoral configurations are less risky, from the incumbent’s

perspective, than poorer areas where electoral competition just emerges. Hegemonic

party configurations at the national level are slightly more risky than the national level

risk; and it is only in the non-competitive municipalities that electoral risk is less than in

the country as a whole. The rich modern large cities and the poorest rural municipalities

are not the riskiest arenas of competition for the PRI. This provides a rationale for the

greater emphasis on clientelism on intermediate levels of development.

To see that the beta coefficient as a measure of risk is not the consequence of

political modernization, table 3 groups municipalities according to whether their beta

coefficient is above or below the municipal average. It also separates the outlier cases,

namely, the coefficients which are outside one standard deviation on either side of the

distribution. The average beta coefficient was 1.15, with a standard deviation of 1.30.

34
hence, we define as very low beta coefficient those that fall outside of the range on the

mean minus one standard deviation (b<-0.15); below average those that are within one

standard deviation under the mean (-0.15<b<1.15); above average, those within one

standard deviation above the mean (1.15<b<2.45); and very high, those that are outside

of the one standard deviation range (b>2.45).

Table 3. Risk distribution by level of development

Beta very high High Low very low
coefficient marginality marginality Medium marginality marginality
Very Low 9% 10% 8% 9% 7%
Below
average 61% 48% 41% 35% 39%
Above
average 19% 26% 34% 36% 47%
Very High 11% 16% 17% 19% 7%
Total 100% 100% 100% 100% 100%

The risk distribution is skewed to the left in very poor places, reflecting a

modernization effect, in that poorer places are less risky. However, past an intermediate

level of development the risk distributions are quite symmetric, and even slightly skewed

to the right, suggesting greater than average risk in poorer areas. What this means in

terms of allocation of clientelism is that poor places might be given resources because

they are poor, or because they are riskier, places where the PRI is losing support at faster

rates than the national trend. . If both poor and rich places are allocated more clientelism

when they have high levels of risk, we can be relatively confident that the overriding

consideration is electoral risk, rather than economic or social development.

Table 4 provides the final evidence that this is the case, by showing average

clientelism according to level of development and risk category. The final column in the

table reveals that clientelism increases as risk is higher: while in low risk areas

clientelism constitutes about a fifth of Pronasol funds, in the highest risk municipalities

35
this share increases to one third. In places with very low marginality, clientelism is less

prevalent, but the share still increases as risk increases. At intermediate levels of

development clientelism is most prevalent, and still responds to risk. This general pattern

is confirmed to be statistically significant in an unreported test of means as well as in a

multivariate regression including controls for level of development.

Table 4. Clientelism by Risk and Development

Percentage of Funds Channelled to Clientelism (Private Good Provision)
very high High Low very low
Betacat marginality marginality medium marginality marginality Total
min/-.15 22.8% 24.5% 24.1% 20.2% 15.1% 22.6%
-0.15/1.15 23.1% 26.7% 31.6% 24.9% 16.3% 26.0%
1.15/2.30 28.3% 32.3% 33.5% 27.7% 21.8% 29.9%
2.30/max 29.2% 32.7% 33.7% 35.3% 22.9% 33.2%

The final piece of evidence we provide is how clientelism was distributed

according to the partisan identity of the municipal governor. Table 6 presents the

information. In the first year opposition municipalities received very little clientelism

projects, but in subsequent years, when a municipal president belonged to a party

different from the PRI, there was a tendency for that municipality to be given more,

rather than less, clientelist projects. Part of the explanation for such pattern is that the

control and credit claiming of public works projects normally goes to municipal

presidents, while clientelism could be targeted more effectively by PRI officials in charge

of Pronasol distributions, agreed in conjunction with the state governor, to individuals

and producer organizations closer to the hegemonic party, bypassing municipal

authorities.

Opposition governed municipalities posses two characteristics, one is that they are

characterized by high risk, and according to our account they should receive greater

36
shares of clientelism. The second is that the municipal governor does not belong to the

PRI’s clientelistic network, and can’t be trusted with public goods to distribute according

to political imperatives different from the preferences of the national PRI. The

information presented in table 5 does not mean that opposition municipalities receive

more funds overall –they do not, consonant with the punishment regime discussed by

Diaz-Cayeros, Magaloni and Weingast.46 Neither does it mean that municipal presidents

receive more per capita funds. Quite the contrary, it means that a greater share of

clientelism means that more resources are being targeted individually to core supporters

by the PRI, in order to curtail local executives belonging to an opposition party.

Table 5. Average clientelism by partisanship

(private goods/total expenditures)

PRI PAN PRD Others

1989 25.7 9.8 0.0 23.1

1990 22.2 22.6 23.2 23.6
1991 26.2 21.5 35.8 26.2
1992 24.9 23.4 28.5 22.2
1993 35.4 36.5 45.6 23.1
1994 36.2 41.7 44.4 37.1

5. Conclusions

This chapter employs a portfolio diversification model to make predictions about

politicians’ choice of clientelism as an electoral investment strategy. Employing the

assumption that politicians seek both to maximize the number of votes (or obtain certain

vote threshold), and minimize electoral risk, we argue that politicians will generally seek

46
Diaz-Cayeros, Magaloni and Weingast (2002) op. cit. provide evidence of the punishment regime in
revenue sharing transfers; Magaloni, Estévez and Diaz-Cayeros (2000) op. cit. show its prevalence in
Pronasol spending.

37
to diversify their portfolios among less risky particularistic transfers (clientelism) and

public good provision, for which an electoral return is more uncertain.

Clientelism, we argue, minimizes electoral risk because politicians can employ

preexisting clientelistic networks to target transfers to core constituencies or true

partisans whose electoral support is certain. In addition, clientelism has the advantage of

allowing incumbents to cement for the future their electoral clienteles because it allows

them to more effectively deter exit. By targeting benefits to supporters and punishing

opponents, politicians can deter opposition-leaning voters from defecting. The voter is

then confronted with a choice of the incumbent with funds, or the opposition without

funds. Unless the voter possesses alternative sources of income or can cheaply exit to

another political jurisdiction, the logic compels her to support the incumbent, even if

reluctantly. This deterrence logic applies as long as the incumbent possesses a

monopolistic control of resources and it can effectively target transfers according to the

recipient’s electoral behavior or political identity.

Budget constraints and transaction costs in targeting do not allow politicians to

rely on clientelism as the sole investment strategy, however. Only where voters are really

poor, we have argued, is the exclusive reliance on clientelism optimal. As a country

modernizes and the pivotal voter becomes wealthier, politicians will be compelled to

relay less on clientelism and introduce public good provision as a dominant form of

political exchange.

From the incumbent’s point of view, public goods have the advantage of lowering

transaction costs, and reaching a potentially larger and more heterogeneous electorate. In

our view, the expected electoral return of public goods is higher than that of private

38
goods, yet public goods show the disadvantage of being more risky precisely because all

voter groups can consume them regardless of their expected voting behavior.

Clientelism thus differs from pork-barreling and other similar forms of vote

buying in two main respects: first, it is targeted to individuals or clearly specified groups;

and second, it is delivered through a party’s clientelistic network such that screening

between true loyalists and opponents takes place. Our approach yields three main

empirical predictions, the first two related to the impact of development and the erosion

of party hegemony over time; and the last one with prevailing configurations of party

competition and electoral risk. First, as a country develops and the pivotal voter becomes

wealthier, clientelism should erode as a dominant form of political exchange simply

because it becomes too costly. Second, as the size of the incumbent’s electoral monopoly

shrinks over time, the party should attempt to buy-off the increasingly more

heterogeneous electorate through public good provision. This means that there should be

less clientelism as political competition gets established. Third, consistent with the logic

of portfolio diversification, the incumbent should attempt to hedge the higher risks

involved in public good provision by devoting more resources in clientelistic transfers to

the most risky localities, holding party system configurations and the incumbent’s level

of electoral support constant.

Our theoretical approach was then applied to the analysis of the erosion of the

PRI’s electoral monopoly and how it responded to higher levels of electoral risk. Our

empirical focus was Pronasol, a poverty relief program implemented between 1989 and

1994. Unlike preexisting analyses of the Program, in this paper we focused on the share

of private to public good provision, and on total per capita private good provision as

39
opposed to the total per capita funds allocated to each political jurisdiction. By

reclassifying the more than 20 programs into these two categories, we were able to assess

some of the main implications of our model. We also depart from most preexisting work

in that our data set is municipal level, not state-level, and it encompasses the whole set of

municipalities, instead of a small sample of municipalities situated in a few states.

Around a third of the Program’s funds were devoted to clientelist transfers, and

the rest of the funds were invested in programs commonly associated with pork-barrel

projects –e.g., road construction, pavement of streets, installment of electricity and

sewage, construction of regional hospitals, and the like. The overwhelming majority of

the municipalities were provided shares of private goods below 40 percent. Thus, the PRI

was mostly providing public goods through Pronasol as a strategic effort to reach out to

voters beyond its core clienteles. Yet, consonant with our portfolio diversification

approach, most municipalities received combinations of private and public goods.

The most developed municipalities received the smallest percentage of

clientelistic transfers. This might indicate either that rich voters do not demand

particularistic goods, or that given a budget constraint, it would be too expensive to

attempt to buy them off with clientelistic transfers. These types of municipalities

constitute only around 5 percent of the observations in our data, however. In addition,

contrary to a modernization theory of linkage building, we found more clientelism at

middle-range levels of development instead of the least developed municipalities.

These findings indicate, first, that modernization will not necessarily erode

clientelistic ties, and second, that other political variables are also relevant for

understanding clientelism. Our approach highlights political competition and prevailing

40
levels of electoral risk. With respect to political competition, we found that non-

competitive municipalities, namely places were the PRI received 100 percent of the vote,

received more per capita funds in clientelism, than places with some form of political

competition - either places where the PRI was still hegemonic or where it confronted

more serious competition. This finding is consonant with our deterrence account of

clientelism. In our view, to keep an electoral monopoly within a municipality in place,

the PRI employed individual targeting, giving benefits to supporters and threatening to

withdraw them from opponents, in order to deter opposition-leaning voters from

defecting.

Our results thus indicate that maintaining an electoral monopoly was hence costly,

from the incumbent’s point of view, as it needed to deliver some form of benefits to its

clienteles even when the party received 100 percent of the vote. Once opposition entry

took place, as in hegemonic and competitive municipalities, the PRI invested more in

public goods as a way to attract the support of a more heterogeneous electorate. Yet, not

all hegemonic and competitive municipalities received less clientelism. Consistent with

the second prediction of our portfolio diversification approach, holding levels of political

competition constant, the PRI employed risk hedging, increasing the share of clientelistic

transfers to the most risky municipalities.

Another component of our approach was to provide an alternative measure of

electoral risk, which builds on the finance literature. We measured risk as the coefficient

of the independent variable in a linear regression where the independent variable is the

national support for the PRI since 1970, and the dependent variable is the support in each

municipality. The “beta” coefficient provides a sense of the systematic electoral risk in

41
each municipality, namely how much the PRI’s vote varies with respect to the national

trend. The national trend in this period is downward sloping: the PRI experienced a

secular decline of support. Thus, if the beta coefficient is larger than one, this means that

the municipality experienced a sharper rate of PRI decline than the national trend. These

are places where the PRI experienced sharp and possibly unexpected vote loses.

Our results indicate that risk in not correlated with development. Indeed, from the

PRI’s point of view, the least risky places were the least developed municipalities, and

the richest areas of the country, where local patterns of electoral competition had

probably been stabilized. The most risky places were situated at middle-levels of

development. It is there where the PRI systematically assigned a higher share of

clientelistic transfers to total transfers.

Finally, we showed that the PRI employed clientelism to go around municipal

opposition governments, and instead use its own clientelistic networks to directly target

the party’s own clientele. The data indicate that opposition governed municipalities

received a higher share of clientelistic transfers to total transfers. In our view, these types

of municipalities posses two characteristics, one is that they are characterized by high

risk, and according to our account they should receive greater shares of clientelism. The

second is that the municipal governor does not belong to the PRI’s clientelistic network,

and can’t be trusted with public goods to distribute according to political imperatives

different from the preferences of the national PRI.

This result does not mean that opposition municipalities receive more funds

overall. As we have demonstrated elsewhere, opposition municipalities received less per

capita funds. What the result means is that the PRI increased the share of clientelisitc

42
transfers to total transfers as a way to more appropriately target resources individually to

core supporters in order to curtail local executives belonging to an opposition party.

43