From Open to Secret Ballot

Abstract

The secret ballot is one of the cornerstones of democracy. We contend that the historical process of modernization caused the switch from open to secret ballot with the underlying mechanism being that income growth, urbanization, and rising education standards undermined vote markets. We undertake event history studies of ballot reform in Western Europe and the U.S. states during the 19th and 20th centuries to establish that modernization was systematically related to ballot reform. We study electoral turnout before and after ballot reform among the U.S. states and British parliamentary constituencies to substantiate the hypothesis that modernization reduced the volume of trade in the vote market.

Keywords

secret ballot modernization electoral turnout democratization

Introduction

One of the most striking facts in comparative political economy is the positive correlation between national income and democracy. Since the first statistical evidence was unearthed in the 1950s by Seymour M. Lipset in his influential article “Social Requisites of Democracy: Economic Development and Political Legitimacy” (Lipset, 1959), a lively debate among political scientists, sociologists, and economists regarding the correct interpretation of this correlation has raged. Lipset (1959) himself interprets, in what has subsequently become known as modernization theory, the correlation as a unidirectional causal relationship from economic development to democracy, or as he puts it “economic development involving industrialization, urbanization, higher educational standards and a steady increase in the overall wealth of the society is a basic condition sustaining democracy” (p. 86). This interpretation is controversial.¹

We propose a new perspective on the modernization debate that we believe will help explore the boundaries of the theory in a more constructive way. We begin by observing that democracy is a package of institutions. This observation is neither new nor novel, and most writers follow Dahl (1971) and make a distinction between different aspects of democracy. Yet, the modernization debate centers on the causal relationship between GDP per capita and composite indices of democracy. It is, therefore, either assumed that democratization is an all or nothing choice—a view that is contradicted by the historical record—or that all the sub-components of the overall package are equally likely (or unlikely) to be causally driven by modernization. The alternative we propose is that modernization may be causally linked to specific sub-components of the overall package of democratic institutions without necessarily governing, in a causal sense, the evolution of the overall package or all of its parts.

We focus on the secret ballot for three related reasons.² First, the secret ballot is regarded as one of the cornerstones of fair elections (e.g., Alvarez, Cheibub, Limongi, & Przeworski, 1996; Elklit, 2000; Rokkan, 1961). Baland and Robinson (2007) note that “the introduction of political institutions that stop corruption and vote buying, such as the Australian ballot, appear to be as significant a step in the process of political development as the construction of electoral democracy itself” (p. 140). Against this background, gaining a better understanding of how the secret ballot came about is important in itself (Przeworski, 2015). Second, there is a straightforward causal mechanism that links modernization to the secret ballot. The mechanism operates through the vote market. Vote markets thrive under open or semi-open voting because this allows the buyer of a vote to verify that the seller kept his part of the bargain. Modernization tends to erode social control and the scope for economic sanction, to improve outside options for ordinary voters, or to undermine old norms of social deference. Income growth also tends to increase the price of a vote. All these forces combine to make vote buying less economical, and the importance of vote markets is diminished. The defenders of the open ballot then become less stout defenders, and ballot reform becomes more likely. Third, as we discuss in more detail below, a large body of research suggests that it is modernization and development that destroy clientelism in modern democracies. Historically, similar processes were in operation (e.g., Stokes, Dunning, Nazareno, & Brusco, 2013), and the demise of the open vote and the introduction of secrecy in voting were a potentially important part of this process.

We begin the analysis by formalizing our reasoning in a novel rational choice model of ballot reform. The model demonstrates why modernization undermines the vote market and how this triggers the secret ballot. The logic is that modernization, while not necessarily eliminating vote buying altogether, reduces the importance of vote markets for those groups who benefit from their existence. This happens because the price of a vote is pushed up as voters become richer or become less willing to conform with norms of social deference. As a consequence of this, the hand of those groups which stand to gain from secret ballot is strengthened, whereas the hand of the groups that benefit from vote buying under open ballot is weakened. This makes ballot reform more likely, and when the secret ballot is introduced, the scope for vote buying is further reduced (in the model eliminated entirely). The model, therefore, predicts (a) that modernization makes secret ballot more likely and (b) that it gradually undermines (but does not eliminate) vote markets in the lead up to the introduction of secret ballot at which point vote markets are further undermined. To support these predictions empirically, we marshal two types of evidence. The first type is based on event history studies of the adoption of the secret ballot. This research design enables us to ask whether modernization—higher income levels, urbanization, and higher education standards—predicts the timing of ballot reform in two different historical samples: Western Europe plus English-speaking off-shoots (1820-1913) and U.S. states (1840-1950). In both cases, we find strong evidence that modernization affected the timing of the secret ballot. In contrast, the same modernization variables cannot predict the timing of reforms that extended the suffrage to broader segments of the male population (Aidt & Jensen, 2014; Przeworski, 2009). The second type of evidence delves deeper into the underlying causal mechanism. With the secret ballot, the vote loses (much of) its value as a tradable commodity, and voters have one less reason to vote. The implied drop in turnout can, therefore, be taken as an indicator of the importance of the vote market under open ballot. We can, then, ask whether the fall in turnout is smaller in places where modernization has progressed more, as one would expect if modernization encourages ballot reform by making electoral corruption less economical. We investigate this in a sample of U.S. states (1870-1950) and in a sample of parliamentary constituencies of Great Britain in the election before and after the Ballot Act of 1872. In both contexts, we find evidence consistent with the proposed causal mechanism.

Our analysis speaks to a large literature on clientelism. A common theme of this literature is that structural forces, such as economic growth, industrialization, urbanization, and expanding electorates, tend to undermine politics based on the exchange of a citizen’s vote for payments or continuing access to employment, goods, and services and tend to encourage political parties to establish other modes of distributive politics or to develop programmatic policy platforms. Kitschelt and Wilkinson (2007) emphasize how such structural forces interact with party competition to generate complex nonlinear dynamics. Stokes et al. (2013) emphasize monitoring and incentive problems in the relationship between party leaders and the intermediaries or brokers they need to employ to deliver private benefits to particular voters and to keep track of the holding-voters-to-account side of the vote market. Keefer and Vlaicu (2008) emphasize that vote buying emerges when political parties are unable to commit to programmatic policy platforms or lack the technologies to communicate them to large groups of voters. In all cases, modernization tends to make vote buying uneconomical and to increase the relative return to alternative electoral strategies. As in the broader literature on clientelism, our new theory of ballot reform emphasizes that modernization—income growth and education in particular—and franchise extension undermine vote markets. We argue that this generates a causal link running from modernization to ballot reform. This, in turn, generates particular patterns in electoral turnout that enable a more refined test of the theory. Our empirical finding—modernization was important for the demise of open voting—not only provides systematic, statistical evidence consistent with our particular model but also speaks to the broader literature on modernization and clientelism. The statistical analysis serves as a complement to the many insightful case studies presented in the existing literature to illuminate this link.

Our analysis is also related to the important work by Leemann and Mares (2011) and Mares (2015) on the politics of secret ballot in Imperial Germany. In Imperial Germany, electoral corruption manifested itself mostly through intimidation of voters by employers and public officials. Mares (2015, Chap. 7) shows that the demand for ballot reform (aimed at increasing the degree of secrecy in voting) in Prussia systematically came from politicians elected in constituencies with a high degree of economic diversification, presumably because electoral strategies based on intimidation were more costly in places with a more diversified employment structure. Our theory emphasizes how other aspects of modernization—income growth, urbanization, and education—increase the monetary cost of vote buying thereby reducing opposition to ballot reform. Although the in-depth analysis of a single case, such as that of Prussia, has many advantages, the external validity will always be an issue. Our approach is to study the relationship between modernization and ballot reform (and the consequences thereof) using a variety of research designs, ranging from cross-country and cross-state analyses to constituency-level analysis within a given country. This, we believe, helps with external validity.

The rest of the paper is organized as follows. In the next section, we present the theoretical framework. This is followed by the results of the two event history studies. The next sections examines the effects of the secret ballot on turnout and discusses alternative mechanism and theories. The last section concludes. The supplementary material includes mathematical proofs and an evaluation of key assumptions, information on our coding of ballot reforms, definitions of all the variables used in the empirical investigation and their sources, and information on empirical robustness checks, including instrumental variable estimations.

A Theory of Ballot Reform

We propose a new theory of ballot reform that formalizes the logic behind the three particular hypotheses that we test empirically.³

Assumptions

We consider a society with regular elections. The suffrage is not universal and voting is, initially, open. While we take the suffrage as given, the choice of ballot system is endogenous. In each period, two parties—party E and party R—compete in an election. Party E represents the old elite, while party R represents the (enfranchised) middle or working classes, henceforth called the radicals.⁴ All $N_{E}$ core supporters of party E can vote, while only some $(N_{R})$ of party R’s core supporters are enfranchised. An (exogenous) increase in $N_{R}$ represents a franchise extension. The old elite is outnumbered $N_{R} > N_{E} .$ Voting is costly, and some enfranchised voters may decide not to vote. The party allegiance of a voter is observable, and the parties share policy R preferences with their core voters. The party that gains the support of the majority of those who turn out to vote wins office. The historical context of our overall study does not suggest that the parties had the capacity (or incentive) to contest elections based on broad programmatic policy platforms. As in Keefer and Vlaicu (2008), the inability of parties to commit to programmatic policies is one of the factors that make vote buying attractive. Yet, voters would care, at least to a small extent, about which party is in power.⁵ To capture this, we assume that the policy with party R—policy R—in power is better for voters of type R than the policy associated with party E—policy E—and vice versa. With this in mind, we can write the utility gain for a voter of type R or E of having “their” party in power as

Δ_{R} \equiv u_{R} (R) - u_{R} (E) > 0,

Δ_{E} \equiv u_{E} (E) - u_{E} (R) > 0,

where $u_{i} (j)$ is the utility that a voter of type $i \in {E, R}$ derives from policy $j \in {E, R}$ . It is possible that the utility loss per capita to the elite associated with a switch in power from party E to party R is larger than the gain per capita to the radicals or vice versa. A neutral assumption is that gains and losses are of equal size, and, as nothing substantial depends on it, we let $Δ \equiv Δ_{R} = Δ_{E}$ . We observe that $Δ$ —the utility difference between the policy platforms of the two parties—is likely to increase over time as parties acquire the technologies needed to commit to more programmatic policy platforms.

The society has an infinitely long-time horizon. Time is indexed by t, and the common discount factor is $β \in (0, 1)$ . Because calendar time plays no important role, we omit time index t when it is not strictly needed. There are two possible ballot regimes: open or secret ballot. Under secret ballot, there is no electoral corruption; under open ballot, votes can be bought and sold in a vote market.⁶ Initially, the ballot is open, but the secret ballot may be adopted if the majority party proposes such a reform and the opposition party is unwilling to call a costly veto. Once secret ballot is introduced, it is not possible to go back to open ballot again.

Within a given period, five stages (A to E) evolve sequentially. What happens within each stage depends on the ballot regime.

A. Planning

Under open ballot, the two parties (simultaneously) decide on how many opposition voters to target with a bribe in the upcoming election. We denote these targets by $n_{~ j j}^{b}$ for $j \in {E, R}$ .⁷ The associated cost is $p_{j} υ_{~ j} n_{~ j j}^{b}$ where $p_{j}$ is the money offered to a voter of type $j$ in exchange for his vote (to be determined below), and $υ_{~ j}$ is the marginal cost of raising funds for party $~ j$ . The parties care about policy, the cost of electoral bribery, and being in power. Their per-period expected payoffs are

V_{E} = f_{E} (u_{E} (E) + M) + (1 - f_{E}) u_{E} (R) - p_{R} υ_{E} n_{E R}^{b},

V_{R} = (1 - f_{E}) (u_{R} (R) + M) + f_{E} u_{R} (E) - p_{E} υ_{R} n_{R E}^{b},

where $f_{E}$ is the probability that party E wins the election and M is the utility value of political office. Under secret ballot, the parties do not, by assumption, buy votes.

B. Electoral turnout

The enfranchised voters decide whether they want to vote. Under open ballot, each voter compares the sum of his expressive benefit of voting ( $θ$ ) and the expected utility value of the bribe ( $v_{j}^{e}$ ) with the utility cost of voting, $c \in (0, 1)$ . He votes if $θ + v_{j}^{e} \geq c$ for $j \in {E, R}$ where the expected utility value of the bribe is

v_{j}^{e} = \frac{n_{~ j j}^{b}}{N_{j}} p_{j} λ_{j}

and $λ_{j}$ is the marginal utility of income. Letting the expressive benefit $θ$ be uniformly distributed on $[0, 1]$ for all voters, the number of voters of type $j$ turning out to vote is

n_{j}^{O B} (n_{~ j j}^{b}, N_{j}) = N_{j} (1 - c) + n_{~ j j}^{b} p_{j} λ_{j} .

Turnout among voters of type $j$ increases in the number of enfranchised voters of this type. More importantly, by offering bribes, a party gives opposition voters an extra reason to turn out to vote. We notice the parties are not paying voters directly to turn out to vote. Rather, this positive turnout effect is a side effect of vote buying. Under secret ballot, no bribes are offered, and turnout is lower:

n_{j}^{S B} (0, N_{j}) = N_{j} (1 - c) .

It is possible that parties, as it becomes impossibly to buy votes outright, start engaging in negative turnout buying, that is, pay opposition voters to abstain. We do not model this possibility explicitly, but note that turnout would fall even further after the ballot becomes secret if this were the case. This, in turn, reinforces our main results.

C. Vote buying

The voters who decided to vote show up at the polling station. Under open ballot, a vote market can operate. A voter of type $j$ is willing to shift his allegiance to party $~ j$ if offered at least his reservation price ( $p_{j}$ ). The reservation price for a voter of type R is the monetary sum needed to compensate him for the economic loss of having policy E instead of policy R, and similar for a voter of type E.⁸ Formally, $p_{j} = Δ λ_{j}^{- 1}$ for $j \in {E, R}$ , where we use the marginal utility of income ( $λ_{j}$ ) to convert the utility differential (∆ $Δ$ ) into a monetary amount. The radicals are poorer than the elite. Because the marginal utility of income falls with income ( $λ_{E} < λ_{R}$ ), the reservation price is higher for voters of type E than for voters of type R. Under secret ballot, the vote market does not operate.

D. Polling

The election outcome depends on the relative electoral support of the two parties and on random events that might induce some voters to shift their allegiances (after bribes, if any, are paid). This induces a preference shock ( $η$ ). The shock can, in principle, be positive (a shift to the radicals) or negative (a shift to the elite) and is introduced to capture the unpredictability of elections. Under open ballot, party E wins a majority among those who turn out to vote if

\frac{n_{E}^{O B} - η + g (v)}{n_{E}^{O B} + n_{R}^{O B}} \geq \frac{1}{2},

where $v = α_{E} n_{E R}^{b} - α_{R} n_{R E}^{b}$ is the (productivity-adjusted) difference between the number of votes bought by party E and party R. The function $g$ , which is increasing, strictly concave and satisfies $g (0) = 0$ , describes the exogenous vote buying technology. The two parameters ( $α_{E}$ and $α_{R}$ ) represent measures of vote buying productivity. As Baland and Robinson (2008) point out, the productivity of vote buying depends, among other factors, on the economic and social sanctions that the buyer can impose on the seller if the vote contract is broken. One of the ways in which modernization can reduce electoral corruption is, therefore, by undermining the scope for economic and social sanctions, that is, by lowering $α_{j}$ .

Under the assumption that $η$ is distributed uniformly on the interval $[- η / 2, η / 2]$ , the win probability of party E under open ballot is

f_{E} (n_{E}^{b}, n_{R}^{b}, n_{E}^{O B}, n_{R}^{O B}) = {\begin{array}{l} 0 \\ \frac{1}{\hat{η}} (\frac{\hat{η}}{2} + \frac{n_{E}^{O B} - n_{R}^{O B}}{2} + g (v)) \\ 1 \end{array}} f o r {\begin{array}{l} (\frac{n_{E}^{O B} - n_{R}^{O B}}{2} + g (v)) < - \frac{\hat{η}}{2} \\ (\frac{n_{E}^{O B} - n_{R}^{O B}}{2} + g (v)) \in (- \frac{\hat{η}}{2}, \frac{\hat{η}}{2}) \\ (\frac{n_{E}^{O B} - n_{R}^{O B}}{2} + g (v)) > \frac{\hat{η}}{2} \end{array}} .

The win probability of party R is $1 - f_{E}$ . Under secret ballot, the vote market shuts down, $g (0) = 0$ , and the win probability of party E is $f_{E}^{S B} \equiv f_{E} (0, 0, n_{E}^{S B}, n_{R}^{S B})$ . Because party R got more core supporters than party E, party E is likely to lose under secret ballot $(f_{E}^{S B} < 1 / 2)$ . After the election, the winner implements its most-preferred policy and earns the office rent ( $M$ ).

E. ballot reform

Under open ballot, the winning party may propose that the secret ballot is adopted for future elections. Although the majority party can implement its policy platform without the consent of the opposition, the opposition can, at a cost, $ρ > 0$ , veto such a proposal. If a reform proposal is made and not vetoed, then the new ballot regime applies for all future elections. If the ballot is already secret, it stays secret. The idea we want to capture with the veto is that reforming the ballot system is fundamentally different from day-to-day policy making and needs to pass a stricter test than the simple majority rule. A natural way to think of the veto is that it represents the ease with which an upper chamber can block proposals from the lower chamber and at what cost.⁹ For example, even in the 19th century where the British House of Lords had real power, most day-to-day policy making was a matter for the House of Commons, but institutional reforms would be subject to a veto constraint from the Lords, as the politics of the Great Reform Act of 1832, for example, illustrates. The veto cost captures this reality in a simple and workable way.

We analyze the model in two steps. First, we characterize the equilibrium in the vote market. Second, we study the reform process.

The Vote Market

If the secret ballot was introduced in the past, party E wins with probability $f_{E}^{S B} < 1 / 2$ . If not, the vote market can flourish and the two parties must, in stage A decide how many opposition voters to target with bribes in the upcoming election, anticipating events as they unfold in the subsequent stages. Using Equations 3 and 4 and recalling from Equation 6 that $\partial n_{j}^{O B} / \partial n_{~ j j}^{b} = Δ$ and $p_{j} = Δ λ_{j}^{- 1}$ for $j \in {E, R}$ , the two first-order conditions governing these choices can be stated as follows:

n_{E R}^{b} : \frac{1}{\hat{η}} \frac{\partial g}{\partial v} (v) \leq \frac{υ_{E} Δ}{λ_{R} α_{E} (Δ + M)} + \frac{Δ}{2 \hat{η} α_{E}} with = if n_{E R}^{b} > 0,

n_{R E}^{b} : \frac{1}{\hat{η}} \frac{\partial g}{\partial v} (v) \leq \frac{υ_{R} Δ}{λ_{E} α_{R} (Δ + M)} + \frac{Δ}{2 \hat{η} α_{R}} with = if n_{R E}^{b} > 0 .

The two terms on the right-hand side shows the cost of buying an extra vote. The first term is the direct (utility) cost per vote. The second term represents the reduction in the win probability of the party that offers the extra bribe because bribery increases turnout among opposition voters. The left-hand side represents the increase in the win probability of allocating an extra dollar to buying opposition votes before adjusting for productivity differences (captured by $α_{j}$ ) and is the same for the two parties. An implication, then, is that only the party with the lowest productivity-adjusted cost buys votes.¹⁰ Under Assumption 1, this party is the minority party representing the elite.

Assumption 1 : \frac{υ_{E}}{λ_{R} α_{E} (Δ + M)} + \frac{1}{2 \hat{η} α_{E}} < \frac{υ_{R}}{λ_{E} α_{R} (Δ + M)} + \frac{1}{2 \hat{η} α_{R}} .

Assumption 1 is likely to be satisfied because voters of type $R$ are cheaper to buy than voters of type $E$ , as they are poorer ( $λ_{R} > λ_{E})$ , because the elite can raise funds at lower cost than the radicals ( $υ_{E} < υ_{R}$ ), and because the elite is more effective at buying votes than the radicals ( $α_{E} > α_{R}$ ), as they can use social sanctions more effectively. The equilibrium win probability of party E is

f_{E}^{O B} \equiv f_{E} (n_{E R}^{b^{*}}, 0, n_{E}^{O B} (0), n_{R}^{O B} (n_{E R}^{b^{*}})) > f_{E}^{S B},

where $n_{E R}^{b^{*}}$ represents how many radical voters party E offers bribes to at equilibrium.

The equilibrium in the vote market ( $n_{E R}^{b^{*}}$ ) and the win probability of the minority party, $f_{E}^{O B}$ , are affected by three key parameters, α_Ε, λ_R, and $N_{R}$ . First, an increase in the transaction cost (a fall in $α_{E}$ ) reduces the marginal benefit of each bribe and, as a consequence, party E targets fewer opposition voters, and the vote market shrinks. This gives radical voters one less reason to vote, and their turnout falls. Yet, the net effect is that party R is more likely to win a majority. Second, a fall in the private marginal value of income ( $λ_{R}$ ) increases the price of a vote. Intuitively, radical voters value money less and so require a higher monetary compensation to shift their allegiance. This increases the cost of bribery, and party E buys fewer votes. The vote market shrinks, and although turnout falls among voters of type R, the consequence is, again, that the win probability of party E goes down. An exogenous franchise extension—an increase in $N_{R}$ —does not affect the absolute number of opposition voters party E wants to bribe, but the relative share drops and, as a consequence, the win probability of party E falls. Suffrage reform makes vote buying uneconomical.

We contend that “modernization”—income growth, urbanization, and education—are systematically related to the transaction cost of vote buying and to the marginal value of income. The latter relationship is immediate: Income growth combined with the standard assumption of diminishing marginal utility of consumption implies that the private marginal value of income ( $λ_{R}$ ) decreases with income growth. This increases the price of a vote and gradually reduces the volume of transactions in the market for votes. The first relationship between “modernization” and the transaction cost of vote buying ( $α_{E}$ ) is more complex and requires elaboration. First, a vote market operates most effectively in environments with a high degree of economic dependency, social control, and deference to authority. The examples of Britain before the Ballot Act of 1872 and Imperial Germany, which nominally had secret ballot from 1871, demonstrate how the social structures of small towns and villages enabled local authorities and elites to control elections through outright bribery or through social deference (Anderson, 1993; Lee, 1994; Mares, 2015). Industrialization and urbanization disturb this equilibrium by opening up new economic possibilities for working and middle class voters and by making them economically mobile. This destroys traditional patron–client networks, which, once they are gone, are difficult to re-create (Hicken, 2007). Anderson (1993) paraphrases the argument about urbanization in her analysis of Imperial Germany as follows:

Supported by the anonymity of the new megalopolis, . . . voters escape both the meshes of deference and the terrors of officials and employers. Town air makes free. Such cities were simply too large for traditional authorities to control. (p. 1458)

The transition from a static agrarian economy to a dynamic industrial economy with deeper markets and economic specialization, therefore, makes it harder for the old elite to enforce and monitor vote contracts, and the transaction cost of vote buying shoots up. Moreover, economic development

weakens the power of the landlord class and strengthens subordinate classes. The working and the middle classes . . . gain an unprecedented capacity for self-organization due to such developments as urbanization, factory production, and new forms of communication and transportation. (Rueschemeyer, Stephens, & Stephens, 1993, p. 75)

By allowing new political parties and special interest groups (e.g., unions) to emerge, this also serves to undermine old structures of deference. Although it is clear, then, that urbanization breaks down structures that previously enabled vote markets, large urban electorates are not a guarantee that electoral corruption is eliminated. The experience of many American cities, for example, New York, in the Gilded Age clearly demonstrates that politicians and parties determined to control the electorate can find new ways of doing so (Fredman, 1968). Urbanization may, therefore, have encouraged electoral corruption in some contexts (e.g., in U.S. migration cities), while having discouraged it in others (e.g., in Britain and Germany). Second, modernization is associated with a rise in literacy and because “education presumably broadens men’s outlooks, enables them to understand the need for norms of tolerance, restrains them from adhering to extremist and monistic doctrines, and increases their capacity to make rational electoral choices” (Lipset, 1959, p. 79), this also makes it harder to enforce vote contracts.

Taken together, these arguments suggest that modernization, either directly through the effect on the marginal value of income or indirectly through urbanization and the rise of education, reduces the incentive of parties, which benefit from electoral corruption, to use vote buying as an electoral strategy. Consequently, as modernization progresses under open ballot, vote markets continue to exist, but they become less and less important and fewer and fewer votes are bought and sold in them. This, in turn, has, as we shall now show formally, implications for the likelihood of ballot reform.

Ballot Reform

At stage E the winning party can, if the ballot regime at time t is open ballot, propose to adopt the secret ballot for future elections. Clearly, party E has no reason to do so. A reform proposal must come from party R which stands to improve its electoral prospect $(f_{E}^{O B} > f_{E}^{S B})$ . The implication, then, is that a reform proposal is tabled at time t only if party R wins the election under the open ballot. This happens with probability $1 - f_{E}^{O B}$ . As discussed above, both modernization and franchise extension increase the cost of buying votes in the market. Party E, therefore, rationally buys fewer votes, and its electoral prospect deteriorates. This increases the likelihood that ballot reform is proposed. Whether it will be reformed, however, depends on the elite’s willingness to veto.

Suppose that party R wins the election at time t and proposes a reform. Party E can veto this proposal at cost $ρ > 0$ . By doing so, the ballot remains open in period $t + 1$ , but another veto may have to be called if party R happens to win again and so on. To evaluate the elite’s willingness to veto, we compare the present discounted value of a veto with the present discounted value of accepting the ballot reform:

W_{E} (no veto) = β \frac{(1 - f_{E}^{S B}) u_{E} (R) + f_{E}^{S B} (u_{E} (E) + M)}{1 - β} .

W_{E} (veto) = β \frac{(1 - f_{E}^{O B}) u_{E} (R) + f_{E}^{O B} (u_{E} (E) + M)}{1 - β} - \frac{β v_{E} Δ λ_{R}^{- 1} n_{E R}^{b *} (.)}{1 - β} - \frac{β (1 - f_{E}^{O B}) ρ}{1 - β} - ρ .

Under secret ballot, power simply alternates between the two parties according to the win probability $f_{E}^{S B}$ , which is determined by the relative turnouts of the two groups and random events. This is what the numerator of Equation 13 represents. Because not calling a veto at time t means a permanent switch to secret ballot starting at time $t + 1$ , this expected value is discounted by the factor $β / 1 - β$ . A veto preserves the open ballot for another period. Under open ballot, power alternates between the two parties according to the win probability $f_{E}^{O B}$ , which takes into account the vote market. This is represented by the first term of Equation 14. On top of this comes the cost of buying votes each period (the second term) and the veto cost, which is paid after each electoral defeat in the future (the third term). The fourth term is the veto cost at time t.

The elite’s willingness to veto is the difference between $W_{E}$ (veto) and $W_{E}$ (no veto). We can state the condition under which the elite will veto as

\frac{β}{1 - β} {(f_{E}^{O B} - f_{E}^{S B}) (Δ + M) - v_{E} Δ λ_{R}^{- 1} n_{E R}^{b^{*}} (.)} \geq ρ + \frac{β (1 - f_{E}^{O B}) ρ}{1 - β} .

The left-hand side represents the net benefit of the veto. This depends positively on the electoral advantage that the vote market gives party E $(f_{E}^{O B} - f_{E}^{S B})$ and negatively on the cost of bribery. The right-hand side represents the veto cost.

The costs and benefits of a veto depend on the two modernization parameters ( $α_{E}$ and $λ_{R}$ ) and how many of party R’s core voters have been enfranchised $(N_{R})$ Modernization—represented either by an increase in the transaction cost or by income growth—makes the elite less likely to veto. The reason is that modernization reduces the electoral advantage gained from the vote market (the gap between $f_{E}^{O B}$ and $f_{E}^{S B}$ narrows) while the frequency with which a veto is required goes up because party R is more likely to win office. Franchise extension also makes the elite less likely to veto a subsequent ballot reform. Because the total number of opposition voters targeted is unaffected, the gap between $f_{E}^{O B}$ and $f_{E}^{S B}$ is unaffected by an increase in $N_{R}$ . However, party R becomes more likely to win elections, and, for that reason, the expected veto cost increases, and this is what makes party E less willing to call a veto.

We summarize these insights in three hypotheses which we test empirically.

Hypothesis 1 (The modernization hypothesis): Modernization increases the likelihood that the secret ballot is adopted.

Hypothesis 2 (The franchise hypothesis): An extension of the franchise increases the likelihood that the secret ballot is adopted subsequently.

Hypothesis 3 (The turnout interaction hypothesis): The secret ballot is associated with a fall in electoral turnout, and this fall is smaller where modernization has progressed the most.

The third hypothesis requires elaboration. Under open ballot, modernization gradually reduces turnout. This is because as the price of a vote goes up, party E becomes less willing to buy votes. As a consequence, from the point of view of a voter of type R, the value of voting falls and more and more such voters decide to stay at home. Our theory does not necessarily suggest that the vote market has disappeared by the time the secret ballot is introduced, only that fewer votes are being bought and sold than in previous times. Accordingly, secret ballot also leads to a fall in turnout because at that point, what remains of the vote market shuts down altogether. To see how the two effects interact to generate Hypothesis 3, imagine two societies which happen to introduce the secret ballot at the same time. The only difference between the two is that one is, say, richer than the other. Consequently, the vote market is less vibrant, and fewer voters bother to vote in the rich than in the poor country in the years leading up to the reform. When the reform happens, the fall in electoral turnout is smaller in the rich than in the poor country.

Our theory is developed in the context of direct vote buying. In reality, indirect vote buying through employment relationships or other favors played an equally important role in many countries, most notably perhaps Imperial Germany (Leemann & Mares, 2011; Mares, 2015; Ziblatt, 2009) or Chile (Baland & Robinson, 2008). So, do our predictions regarding “modernization” and the secret ballot also apply to a market dominated by indirect vote buying? The answer is that they do. To see this, notice that we can reinterpret the price, $p_{j}$ , as an effort cost incurred by the parties. Effort can coerce voters to “accept” the utility cost ∆ $Δ$ , which we might reinterpret as a psychic cost of voting against once preferred party, because the sanction associated with not doing so is greater. “Modernization” reduces the scope for social control by increasing the effort cost required to buy votes and in that way, undermines the value of indirect vote buying.¹¹

Event History Studies of the Secret Ballot

The aim of our event history study is to explain the timing of the secret ballot. We model the (conditional) probability that a country or a state that has not yet adopted the secret ballot adopts it in a given year as a function of quantitative measures of modernization, the franchise, and other potential determinants of ballot reform.

We explore two different historical samples—Western Europe plus off-shoots (1820-1913) and the U.S. states (1840-1950)—that cover the relevant period during the 19th and 20th centuries when the secret ballot replaced the open ballot. We are aware that this is a small subset of the countries or states that have undergone the transition from open to secret ballot. Given the constraints imposed by the availability of quantitative historical data on modernization, in particular, GDP per capita, it is not possible to expand the Western Europe plus off-shoots sample further, say, with countries from Eastern Europe.¹²

We are interested in the year in which the secret ballot became effective in a country or a state. By effective, we mean that the ballot rules were such that voters could cast their vote in private (i.e., without the vote choice being observed and unduly influenced by others) and that it was impossible ex post to reconstruct how a particular voter voted. We focus on when the secret ballot became effective because this is what our theory suggests is relevant. In the theory, the secret ballot is introduced when the benefits to the elites are less than the veto cost. Introducing secret ballot de jure but with ways to circumvent secrecy works similarly to the veto in our theoretical model. From a practical point of view, de jure secrecy may not be de facto secrecy. This may be so because of non-standard ballot paper of different colors or shapes, because of non-standard urns (for example, a small urn which preserves the order of voting can reveal the choices of individual voters ex post), or because of the absence of a place where the ballot can be filled in and placed in the urn without outsiders being able to observe the act (Mares, 2015).

A number of different arrangements can in principle and in practice make the ballot secret in this sense. One such arrangement is the Australian ballot. The Australian ballot requires, at least, two things: First, all parties/candidates are listed/printed on the same ballot paper, and second, these ballot papers are printed by the state at the public’s expense and distributed only at the polling stations and placed in a standard-sized urn in privacy. Another arrangement is the so-called “ballot and envelope” system, practiced in, for example, France, Sweden, and Spain. It requires the state to print the ballot papers, but one ballot paper is printed per party or candidate. The voter then chooses the paper for the party/candidate he wants to cast his vote for, puts it in an envelope, and deliver it to the ballot box. This type of arrangement (and others like it) is more open to abuse than the Australian ballot, but de jure makes the ballot secret. In coding the year in which secret ballot was introduced in the countries covered by the Western European plus off-spring sample, our rule is that a new ballot arrangement qualifies as “secret ballot” if our sources indicate that the relevant change in balloting procedures effectively made the act of voting private. In most cases, it is straightforward to establish the year of adoption thus defined, but in some cases, most notably Germany, there is disagreement about when the ballot became effectively secret. We use 1913, when standard urns were introduced in the main specification, but note that our results are unchanged if we instead use 1903 when ballot envelopes and isolating places were introduced (Mares, 2015). In the supplementary material (Appendix S2), we provide information on how we coded the year of adoption for each country and why. For the sample of U.S. states, we code the year in which the ballot in a state became Australian in statewide elections based on the information provided by Lott and Kenny (1999).

The dependent variable $r e f o r m_{i t}$ is coded 1 if country (or state) $i$ introduced the secret ballot in year t and as 0 in the years before and after that. A country (or state) drops out of the sample in the year after its adoption. We assume that all polities were at “risk” of adopting the secret ballot from the beginning of the (relevant) sample period or when they became independent. This assumption is justified by historical facts. By the late 18th century, the secret ballot was widely debated in liberal circles and supported by philosophers, such as Jeremy Bentham. In Britain, the secret ballot was proposed by the Westminster Committee as early as 1780 (Schofield, 2004). Thus, the principle of secret ballot was known and on the agenda in many countries by 1820. It is, therefore, reasonable to assume that the countries (states) under consideration could have adopted the secret ballot from 1820 (1840) onward.

As in Beck, Katz, and Tucker (1998), we estimate a discrete logistic model¹³:

P ({reform}_{i t} = 1 | x_{i t}, m_{i t - 1} = 0) = \frac{1}{1 + e^{- (x_{i t} γ + H (.))}} .

The variable $m_{i t - 1}$ is an indicator variable equal to zero in each year before introduction of the secret ballot and equal to one thereafter. Through the function $H (.)$ , we allow the hazard rate to be time-varying and a function of the number of years a country has been at risk of adoption.¹⁴ This is important because the likelihood of adoption may increase as time passes. By taking duration dependence into account, we avoid confounding any modernization or franchise effect with a spurious correlation between two upward trending variables. The vector $x_{i t}$ represents three main groups of explanatory variables. The first group contains indicators of modernization, such as the log of real GDP per capita, the urbanization rate, and measures of education attainment standards. The second group contains variables related to the extension of the franchise. The third group contains variables that capture alternative causes of ballot reform. Many scholars emphasize the importance of landholding and income inequality in relation to democratization in general. We expect that landholding inequality makes the secret ballot less likely because inequality along this dimension often reflects a social order that favors vote buying. Other scholars, for example, Wejnert (2005), Gleditsch and Ward (2006), and Leeson and Dean (2009), emphasize international diffusion of democracy. Governments in one polity may learn from political reforms—in our case, ballot reforms—in other polities, and more so from those which are either linguistically or physically nearer. Finally, we control for population size.

Western Europe Plus Off-Shoots

The Western Europe plus off-shoots sample covers, for the period from 1820 to 1913, 11 Western European countries¹⁵ plus the United States, Canada, and New Zealand. The first country in the sample to introduce the secret ballot was the Netherlands in 1849; the last ones were France and Germany in 1913. Before the secret ballot, electoral corruption was widespread. In many Western European countries, vote buying was, during the 19th century, concentrated in the countryside where social control and employment relations made it relatively easy for the landed elites to run vote markets (e.g., Mackie, 2000; Seymour, 1915; Ziblatt, 2009). In the United States, vote markets were also widespread in the countryside, but they were often particularly vibrant in the big migration cities where party machines exploited that colored voting papers, indicating party choice, could be handed out at the polling stations (Cox & Kousser, 1981; Ostrogorski, 1964).

In Table 1, we report the results of the event history study. We see that the coefficients on the two modernization variables, urbanization rate and real GDP per capita, are positive and statistically significant (columns 1 and 2).¹⁶ When they enter together (column 3), they are jointly significant, and real $G D P$ per capita remains significant on its own. The two variables are positively correlated, with a correlation coefficient of .75, and are likely to capture the same underlying concept of modernization. It, therefore, makes sense to extract the principal components. In column 4, we replace the two modernization variables with the first principal component which correlates positively with real GDP per capita and with urbanization rate.¹⁷ This component has a positive and statistically significant coefficient. Ballot reform is a rare event, and this may bias the estimates. King and Zeng (2001) propose a logit estimator that deals with this. In columns 5 to 8, we report results using this estimator. The two modernization variables remain individually significant and jointly significant when entered together or as the first principal component. To quantitatively evaluate the modernization effect, we may begin by noticing that based on the specification in column 3, the probability of a ballot reform in a given year conditional on the average values of the covariates is 0.3%. Suppose now that real GDP per capita increases by one standard deviation, keeping all other variables at their average values. As a consequence of this, the predicted probability of reform in a given year increases to 2.6%. While ballot reform remains an unlikely event in any given year, we observe that the likelihood increases many fold as GDP per capita increases.¹⁸

Table 1.

The Effect of Modernization of the Probability of Secret Ballot in Western Europe Plus Off-Shoots 1820-1913.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
log (real GDP per capita)	6.36**		6.62*		5.43**		5.32		0.15*
	[2.44]		[1.77]		[2.11]		[1.44]		[1.99]
Urbanization rate		0.01***	−0.001			0.09***	−0.001
		[3.65]	[−0.21]			[3.38]	[−0.01]
First principal component				1.46***				1.29***
				[3.27]				[2.94]
Electorate/adult population	0.04	0.06**	0.03	0.05*	0.03	0.05**	0.03	0.03	0.002
	[1.26]	[2.30]	[0.94]	[1.87]	[1.19]	[2.17]	[0.94]	[0.94]	[0.92]
Gini coefficient	28.49	12.33	28.35	16.80	25.36	12.03	26.21	26.21	−1.041
	[1.44]	[0.66]	[1.39]	[0.93]	[1.30]	[0.66]	[1.30]	[1.30]	[−1.12]
log (population)	−0.50**	−0.76***	−0.46	−0.71**	−0.41*	−0.66**	−0.39	−0.39	0.13
	[−2.04]	[−2.80]	[−1.35]	[−2.54]	[−1.72]	[−2.47]	[−1.18]	[−1.18]	[1.65]
Learning	−1.18	1.2	−1.26	0.22	−1.03	0.93	−1.03	−1.03	0.001
	[−0.94]	[1.56]	[−0.95]	[0.27]	[0.84]	[1.44]	[−0.78]	[−0.78]	[0.03]
Joint significance of modernization variables			6.24**				4.91*
Control for duration dependence	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	779	732	732	732	779	732	732	732	779
Estimation method	Logit	Logit	Logit	Logit	Rare events	Rare events	Rare events	Rare events	Fixed-effects OLS

Dependent variable: Reform. Robust z statistics correcting for clustering by country in brackets (i.e., we allow the errors to be correlated over time within countries, but not between countries). Constants not reported. All estimations based on sample of 14 countries. The small number of countries raises the concern that the standard errors corrected for clustering are misleading. We have done two checks to judge the relevance of this issue. First, we use more conservative critical values for the t distribution based on G–K degrees of freedom with G being the number of clusters and K the number of variables (Angrist & Pischke, 2009). Second, we apply a wild cluster bootstrap in the linear probability model (Cameron, Gelbach, & Miller, 2008). None of these techniques indicate that the inference based on the clustered standard errors reported in the table is misleading. OLS = ordinary least squares.

p < .1. **p < .05. ***p < .01.

We use information on the number of eligible voters as a proportion of the adult population, electorate/adult population, to test the franchise hypothesis. The results reported in Table 1 are not favorable. Although the point estimate is positive, it is only statistically significant in specifications where real GDP per capita is excluded. Overall, it was not the pre-secret ballot expansion of the suffrage that triggered the secret ballot.

Income inequality, as measured by the gini coefficient, is (statistically) unrelated to the timing of the secret ballot. This runs counter to other recent evidence on the effect of inequality on democratization (Boix, 2003, Chap. 2; Ziblatt, 2008; or Ansell & Samuels, 2010). The control variable population always has a negative coefficient, but is usually not significant, suggesting that scale effects were unimportant. The variable learning captures diffusion effects. It is a “distance”-weighted index of reforms in neighboring countries, where we use the information on linguistic similarities provided by Fearon (2003) to measure distance. Despite the fact that the adoptions of the secret ballot cluster in the 1870s, we find little evidence that social learning was important. A similar result is reported in Aidt and Jensen (2014) in relation to international diffusion of suffrage reforms. These results contradict the strong evidence on international diffusion of democracy coming from studies of large panels of countries, either in the second half of the 20th century (e.g., Gleditsch & Ward, 2006) or the entire period from 1820 or 1850 to 2000 (Persson & Tabellini, 2009 or Leeson & Dean, 2009). These studies focus on composite democracy measures (mostly the Polity IV index). Our results pertain to the first wave of democratization and to specific aspects of the package of democratic institutions. It appears that international diffusion of specific institutions was not all that important for the early democratizations in Western Europe.

The results reported in Table 1 are estimated from a combination of between- and within-country variation in modernization, the size of the electorate, inequality, and so on. It is, therefore, possible that the correlations are driven by the same unobserved factors and that they are coincident rather than causal. To show that this is most likely not the case, we have undertaken two additional tests, both of which force us to work with a linear probability model.¹⁹ First, we have estimated a linear fixed-effects model that takes into account country-specific unobserved heterogeneity. We report the results in column 9 of Table 1. We observe that the coefficient on the main modernization variable, real GDP per capita, now estimated only from variation within each of the 14 countries over time, is significant with a robust t statistic of 1.99. Second, we have estimated the linear probability model with an instrumental variables estimator. We use a weighted index of the level of (log) real GDP in other countries as an instrument for (log) GDP per capita in a particular country. This is similar to Acemoglu, Johnson, Robinson, and Yared (2008), except that we use inverse distance rather than trade shares as weights. The main threat to the validity of the instrument is the possibility that an increase in real GDP per capita abroad induces ballot reform, and this experience diffuses internationally. We deal with this by including the variable learning, which controls for the diffusion process, in the estimations. We use a measure of revolutionary shocks in neighboring countries as an instrument for the extension of the franchise in particular countries (see Aidt & Jensen, 2014). The instrumental variables results are reported and discussed in Supplementary Appendix S4 (Table S4) in the supplementary material. They show that the baseline results are robust and may be given a causal interpretation.

U.S. States

The first U.S. state to adopt the secret ballot, which for this sample we define as Australian ballot, statewide was Massachusetts in 1888 and the last was South Carolina in 1950. Before the Australian ballot, electoral corruption was widespread. Bensel (2004) notes that the introduction in many states of the ticket system prior to the Australian ballot not only provided some privacy for voters but also stresses that the degree of secrecy should not be overstated. The issue was that parties were able to ascertain voting behavior because they supplied voters with ballot papers (Lehoucq, 2007). Fredman (1968) describes how it worked:

The simplest form of bribery occurred when ballot peddlers or district captains paid a voter as he emerged from the polling place. To check that he actually used the ballot it was colored or otherwise recognizable and the compliant voter was followed up to the booth. (p. 22)

McCook (1892) estimates that 16% of voters in Connecticut were up for sale at prices ranging from US$2 to US$20. Stokes et al. (2013) emphasize that the Australian ballot reduced the effectiveness of vote buying in the United States, and that it did so by diminishing “the observability of voter’s choices” (p. 183). The most corrupt 19th-century state elections are said to have occurred in New York and San Francisco. The reason was the high concentration of poor voters and recent immigrants unused to the franchise (Fredman, 1968). Congleton (2011) notes that the introduction of the Australian ballot “allowed votes to be cast without fear of rebuke by landlords or employers” (p. 560). Taken together, this suggests that the ballot became effectively secret in the U.S. states with the arrival of the Australian ballot.

We use three quantitative indicators of modernization—real income per worker, average years of schooling, and urbanization rate. From Table 2, we see that each of the modernization variables is positively and significantly correlated with the timing of the secret ballot (columns 1-3). When included together in column 4, they are jointly significant, and real income per worker and average years of schooling remain significant on their own. We replace the three variables with the first principal component in column 5 and note that it is highly significant.²⁰ These results are robust to alternative estimation techniques, as shown in columns 5 and 6.

Table 2.

The Effect of Modernization of the Probability of Secret Ballot in U.S. States, 1840-1950.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
log (real income per worker)	1.49***			1.24**		1.19**	0.60
	[3.62]			[2.37]		[2.30]	[1.02]
Average years of schooling		0.53***		0.32*		0.31*	0.78***
		[3.73]		[1.91]		[1.84]	[3.05]
Urbanization rate			0.02**	−0.01		−0.01	0.02
			[2.05]	[−0.38]		[−0.29]	[0.97]
First principal component					0.58***
					[2.99]
Women’s suffrage	0.21	−0.12	0.11	0.16	0.20	0.19	−0.17
	[0.46]	[−0.26]	[0.22]	[0.33]	[0.42]	[0.40]	[−0.28]
Literacy test	−0.84	−0.88	−0.86	−0.78	−0.85	−0.72	−0.50
	[−1.21]	[−1.19]	[−1.33]	[−1.09]	[−1.28]	[−1.01]	[−0.93]
Poll taxes	−0.30	−0.45	−0.57	−0.25	−0.34	−0.20	−0.32
	[−0.82]	[−0.91]	[−1.36]	[−0.61]	[−0.88]	[−0.49]	[−0.63]
Land inequality							−31.10***
							[−3.67]
log (population)	−0.07	−0.28***	−0.20	−0.09	−0.12	−0.10	−0.11
	[−0.65]	[−2.62]	[−1.48]	[−1.03]	[−1.25]	[−1.23]	[−0.92]
Learning	−0.05	−4.29	−12.19	−5.35	−9.69	−5.58	9.76
	[−0.01]	[−0.36]	[−0.94]	[−0.39]	[−0.77]	[−0.40]	[0.81]
Joint significance of the modernization variables				20.70***		20.10***	18.10***
Joint significance of the suffrage variables	2.33	2.02	3.40	1.72	2.70	1.57	1.49
Observations	2,239	2,214	2,251	2,186	2,186	2,186	738
Number of states	45^a	48	47^b	44	44	44	44
Estimation method	Logit	Logit	Logit	Logit	Logit	Rare events	Logit

Dependent variable: Reform. Robust z statistics correcting for clustering by state in brackets. All estimations control for duration dependence.

Income data are missing for South and North Dakota and Oklahoma. b. Urbanization is missing for Idaho.

p < .1. **p < .05. ***p < .01.

To evaluate the size of the modernization effect, suppose that real income per worker increases by one standard deviation. Using the specification in column 4 and keeping all other variables at their average values, this increases the predicted probability of a ballot reform from 0.21% to 0.37%. This effect is smaller than in the Western Europe plus off-shoots sample, but still substantial.

Unlike in Western Europe, 60% to 90% of adult (White) males were enfranchised already in the 1840s. Nevertheless, the states applied various strategies to de facto restrict the suffrage. These included requiring payment in full of poll taxes and subjecting potential voters to literacy tests. These steps served to keep poor and illiterate males off the election roll, often aimed at disenfranchising African Americans (Keyssar, 2009, Chaps. 4 and 5). Women’s suffrage rights also varied, and some of the frontier states granted women the right to vote long before it became mandatory in 1920 (Braun & Kvasnicka, 2013; Lott & Kenny, 1999). We use these restrictions to capture over time and across-state variation in the size of the electorate. From Table 2, it is clear, however, that these restrictions had very little impact on the secret ballot. We must, again, conclude that there is little evidence supporting the franchise hypothesis.

We use the share of land held by the 20% largest farms (land inequality) to capture landholding inequality (Galor, Moav, & Vollrath, 2009). Information on this share is only available from 1880, so we lose more than half the observations when we include it in the specification shown in column 7 of Table 2. Yet, landholding inequality is highly significant and exerts a negative impact on the secret ballot. This is in line with the findings of Ziblatt (2008) and Ansell and Samuels (2010, 2015) in other contexts. We find little evidence that scale effects or social learning mattered among the U.S. states.

In Appendix S4 in the supplementary material, we discuss a number of additional robustness checks. This includes instrumental variable estimation based on the same strategy as for the Western Europe plus off-shoots sample. The results are consistent with a causal interpretation of the relationship between modernization and secret ballot. Kuo and Teorell (2017) report independent evidence which further supports such an interpretation and suggests that the Australian ballot reduced vote buying and intimidation in U.S. state elections between 1860 and 1930. Moreover, the case study of Stokes et al. (2013, Chap. 8) also corroborates the hypothesis that modernization helped end clientelism in the United States.

Turnout and Vote Buying

The evidence presented so far supports the view that modernization systematically affects the timing of the secret ballot. The underlying mechanism, however, remains unclear. In our theory, the causal link runs through the vote market. The available quantitative evidence on the operation of vote markets comes from a handful of insightful studies of particular markets.²¹ We provide new evidence on the interaction between modernization and the secret ballot by testing the turnout interaction hypothesis. The logic behind this hypothesis is that modernization gradually reduces the volume of trade in the vote markets operating under open ballot, and that the secret ballot, in one go, eliminates (most of) what remains of these markets. As argued, for example, by Converse (1974), Rusk (1974), and Heckelman (1995), this reduces electoral turnout. The elimination of the vote market is, however, not the only possible explanation for a fall in turnout after secret ballot. For example, it is possible that there is a shift from vote buying toward negative turnout buying (i.e., to pay expected opposition voters to stay home) after a reform. The seminal study by Cox and Kousser (1981) of newspaper reports in New York State about instances of electoral corruption before and after the introduction of the secret ballot in 1890 provides examples of this. It is also possible that turnout will increase as a consequence of the secret ballot. This would be the case if the expressive benefit of voting shoots up by making the act of voting private or if the secret ballot “protects” voters and gives them the freedom to vote how they like. A possible example of this is Imperial Germany where the adoption of the secret ballot was correlated with an increase in turnout. Turnout could also increase if parties start using “brokers” to deliver blocks of votes (Stokes et al., 2013). The idea is that party leaders can contract with those brokers, who themselves use personal connections and knowledge of individuals to foster favorable turnout. This leads to positive turnout buying, and if this is taking place at a sufficiently large scale, then turnout would not fall as a consequence of the secret ballot. We notice, however, that all these additional effects (which are not captured by our model) operate ex post, that is, after the ballot has become secret. Our theory and the turnout interaction hypothesis are about what happens to the vote market and turnout ex ante, that is, while the ballot is still open. What we argue is that the fall in turnout, documented by Heckelman (1995), is smaller in polities that have undergone significant modernization in the years before ballot reform than in polities that have not. This would also be true if there is a switch to negative turnout buying after the reform. If, however, turnout is encouraged by the secret ballot because of changes in expressive benefits, positive turnout buying, or the development of broker-mediated forms of clientelism, then the prediction from the model would be that the this increase would be bigger in places where modernization is more progressed. In all cases, the point we want to stress is that modernization reduces turnout under open voting by reducing vote buying. We test this on a sample of U.S. states from 1870 to 1950 and on a sample of parliamentary constituencies in Great Britain before and after the Ballot Act of 1872.²²

The U.S. States

For the sample of U.S. states, we extend Heckelman’s (1995) baseline model for the voter turnout rate in gubernatorial elections in the 48 contiguous U.S. states and estimate the following panel model:

t u r n o u t r a t e_{i t} = δ_{i} + η_{t} + β_{0} A B_{i t} + β_{1} A B_{i t} \times M_{i t} + β_{2} M_{i t} + Z_{i t} β_{3} + e_{i t},

where $i$ represents a state, $t$ represents elections (from 1870 to 1950), $A B_{i t}$ is a dummy variable equal to zero before the Australian ballot and equal to one after, $M_{i t}$ is a “modernization” variable of interest (real income per capita, average years of schooling, or urbanization rate), $Z_{i t}$ is a vector of additional control variables, and $δ_{i}$ and $η_{t}$ are state- and time-fixed effects. The turnout interaction hypothesis predicts that $β_{0} < 0$ and $β_{1} > 0$ . The hypothesis that modernization under open ballot reduces vote buying implies that $β_{2} < 0$ .

Table 3 reports the results. In column 1, we reproduce Heckelman’s (1995) finding that the Australian ballot has a significant, negative effect on turnout, consistent with the hypothesis that before the Australian ballot, turnout is kept high because of vote buying. Columns 2 to 4 report on specifications with the interaction between modernization and secret ballot. To facilitate interpretation, we introduce one interaction term at the time. The coefficient on the interaction term is positive and significant in each specification. In column 5, we include all three interactions simultaneously. They are less precisely estimated, but two of them remain significant at the 10% level, and they are jointly significant at the 1% level. The turning points for the modernization variables are all within sample and are indicated at the bottom of the table. We can, for example, imagine a state which starts with a low education level but expands education over the years. If this state happened to introduce the Australian ballot early, the consequence would be a fairly large drop in turnout, but if it adopts later, the drop will be smaller and eventually turnout might even increase. The last effect is not explained by our model because the expressive benefit of voting is constant. In practice, this benefit may increase when voting becomes secret. This produces a countervailing effect that could dominate in cases where the importance of the pre-Australian ballot vote market has already been curtailed by modernization. We observe that the point estimate on two of the modernization variables, real income per worker and urbanization rate, is negative. This supports the view that modernization undermined the vote market prior to the Australian ballot. We note, however, that the positive point estimate on average years of schooling is inconsistent with this. We conjecture that this variable is picking up a positive correlation between the expressive value of voting and education that dominates any effect running through the vote market. Overall, these results corroborate the hypothesis that modernization undermines the vote market.

Table 3.

The Effect of the Interaction Between Modernization and Secret Ballot on the Turnout Rate in 48 U.S. States 1870-1950.

	(1)	(2)	(3)	(4)	(5)
Australian ballot	−5.59*	−164.30***	−27.37***	−7.90*	−133.30**
Australian ballot	[−1.96]	[−4.01]	[−3.15]	[−1.85]	[2.29]
Australian Ballot × Real Income per Worker		17.50***			12.53*
Australian Ballot × Real Income per Worker		[3.99]			[1.76]
Australian Ballot × Average Years of Schooling			5.14***		3.07
Australian Ballot × Average Years of Schooling			[3.23]		[1.59]
Australian Ballot × Urbanization Rate				0.15*	0.01
Australian Ballot × Urbanization Rate				[1.68]	[0.04]
Average years of schooling		4.31***	1.84	3.81**	3.25**
Average years of schooling		[2.84]	[1.13]	[2.09]	[2.06]
Real income per worker		−20.48***	−9.83**	−13.53***	−15.76**
Real income per worker		[−4.48]	[−2.55]	[−3.50]	[−2.27]
Urbanization rate		−0.49**	−0.54***	−0.50***	−0.56**
Urbanization rate		[−2.50]	[−2.77]	[−2.78]	[−2.58]
Women’s suffrage	−13.33***	−11.56***	−9.53**	−11.33***	−10.28***
Women’s suffrage	[−4.23]	[−3.47]	[−2.43]	[−3.00]	[−3.05]
Poll tax	−19.04***	−14.61***	−15.53***	−15.85***	−13.98***
	[−6.03]	[−5.54]	[−5.53]	[−5.19]	[−5.10]
Literacy test	−8.29***	−4.62*	−5.63**	−6.44**	−4.59*
	[−2.87]	[−1.68]	[−2.05]	[−2.16]	[−1.72]
Joint significance of the interaction terms					5.34***
R ²	.79	.82	.82	.81	.83
Turning point	NA	9.38	5.33	52.64	Not reported

Dependent variable: Turnout Rate. Robust z statistics correcting for clustering by state in brackets (i.e., we allow the errors to be correlated over time within states, but not between states). State- and year-fixed effects included. Number of states is 48; total number of observations is 1,408. Specification (1) corresponds to the baseline model in Heckelman (1995, Table 4).

p < .1. **p < .05. ***p < .01.

The Ballot Act of 1872

The Second Reform Act of 1867 was followed 5 years later, in 1872, by the Ballot Act which introduced the Australian ballot in elections to the British House of Commons. We use the Ballot Act as a quasi-natural experiment.²³ We study electoral turnout patterns at the constituency level in the general elections in 1868 and 1874, both of which were conducted according to the new franchise and district rules laid down in 1867, but under different ballot conditions. To test the turnout interaction hypothesis, we proxy modernization with the number of inhabitants per house in each constituency ( $d e n s i t y_{i t}$ ) that is closely related to the degree of urbanization.²⁴ The outcome variable ( $t u r n o u t_{i t}$ ) records the number of voters who voted in constituency $i$ in election $t$ for $t \in {1868, 1874}$ . We consider the following model for $t u r n o u t_{i t}$ :

\begin{array}{l} t u r n o u t_{i t} = α_{i} + ν_{t} + α_{1} R_{t} + α_{2} d e n s i t y_{i t} R_{t} + α_{3} d e n s i t y_{i t} + \\ α_{4} e l e c t o r s_{i t} + ε_{i t}, \end{array}

where $R_{1868} = 0$ and $R_{1874} = 1$ indicate that the reform of the ballot took place between 1868 and 1874, $α_{i}$ is a constituency-specific fixed effect, $ν_{t}$ is an aggregate time effect, and $e l e c t o r s_{i t}$ is the number of registered voters. We allow $d e n s i t y_{i t}$ to have a direct impact on turnout, in addition to its interaction with the ballot reform. Taking the first difference of Equation 18, we get

\begin{array}{l} Δ t u r n o u t_{i t} = (v_{1874} - v_{1868} + α_{1}) + α_{2} d e n s i t y_{i 1874} + α_{3} Δ d e n s i t y_{i t} + \\ α_{4} Δ {(e l e c t o r s)}_{i t} + (ε_{i 1874} - ε_{i 1868}) . \end{array}

Clearly, the direct effect of the reform ( $α_{1}$ ), which we expect to be negative, cannot be identified independently of the common time effect. However, we can recover the interaction effect between the reform dummy variable and the indicator of modernization and test whether $α_{2} > 0$ We can also test whether turnout was lower in places that experienced an increase in population density between the two elections, that is, if $α_{3} < 0$ . Finally, we expect that the number of electors affects turnout positively, that is, $α_{4} > 0$ .

Unfortunately, we only know the votes polled for each candidate and in constituencies with more than one seat, voters could cast as many votes as there were seats (Craig, 1977). Consequently, the number of votes is not, in general, equal to the number of voters. However, for the constituencies with only one seat, the number of votes is equal to the number of voters who turned out to vote, and our test is based on a sub-sample of 74 one-seat constituencies. The estimation results are

Δ t u r n o u t_{i t} = \underset{(- 1.57)}{- 263} + \underset{(1.92)}{61.5} d e n s i t y_{i 1874} - \underset{(- 2.09)}{614} Δ d e n s i t y_{i t} + \underset{(5.31)}{0.47} Δ e l e c t o r s_{i t},

where the figures in parentheses are (robust and clustered) z statistics. The combined estimate of the reform and the aggregate time effect is negative, consistent with a post-reform drop in turnout, but not statistically significant. More importantly, the coefficient on density is positive with a p value of .06. Consistent with the turnout interaction hypothesis, this suggests that the vote market, before the reform, was more vibrant in places that were less densely populated. This conclusion is supported by the negative coefficient on change in density, which indicates that urbanization undermined the voting market and caused turnout to fall. Overall, the results are consistent with the causal mechanism suggested by our theory, and they are corroborated by the insightful case study by Stokes et al. (2013, Chap. 8), which concludes that the reason why political parties in Britain adopted legislation to limit electoral corruption and vote buying was that industrialization had made vote buying less economical.²⁵

Alternative Mechanisms and Theories

All our empirical tests are consistent with the hypothesis that modernization paves the way for secret ballot. Our theory emphasizes the interaction between vote markets and modernization as the underlying causal mechanism. This is, however, not the only possibility, and in this section, we evaluate the plausibility of alternative channels through which modernization might have triggered ballot reform.

First, although we stress the effect of modernization on demand and supply in a competitive market for votes, Stokes et al. (2013) emphasize the importance of brokers, such as party agents, and the way modernization may reduce the value of clientelism. In particular, they explain that

poverty may increase the returns to clientelism by making voters (who on average will have larger marginal utilities of income in poorer societies) more responsive to transfers, whereas growth of average income weakens these returns and thus increases the incentives of brokers to extract rents. (pp. 179-180)

Given the data marshaled for this project, it is, however, not possible to discriminate effectively between the vote buying mechanism and broker-mediation mechanism. Both mechanisms are plausible, and both predict a positive relationship between secret ballot and modernization, and, as such, our evidence is consistent with both, and we view them as complementary rather than as substitutes. Modernization may also affect the old regime elites in other ways than proposed by our theory where they are “reluctant” reformers. In Prussia, for example, conservative elements of the elite supported secret ballot in the 19th century presumably because “intimidation of voters in the rural districts of the East was more than matched by the SPD (Sozialdemokratische Partei Deutschlands) intimidation of shopkeepers, artisans and non-Socialist voters in the cities of the West” (Retallack, 1988, p. 164). In fact, between 1869 and 1912, 18 bills were proposed in the Reichstag to introduce elements of secrecy in voting. Mares (2015, Chap. 6) shows that bills were sponsored by the politicians who were elected in districts with a high degree of occupational fragmentation. This suggests that economic development through its effect on economic diversity created demand from within the German elite for ballot reform. Along similar lines, Elklit (1988) notes that voters of all parties, including the conservative and social democratic party, experienced electoral coercion in Denmark prior to the secret ballot. Depending on shifts in the relative ability of parties to coerce voters, subsets of the elite might, therefore, see ballot reform as an expedient way to gain electoral advantage. This line of reasoning suggests an alternative way that modernization may drive ballot reform not present in our model: Modernization empowers the working class and creates economic diversification that makes the old elite favor ballot reform simply because intimidation becomes unworkable. One way to evaluate the power of this mechanism in our context is to include the share of left-wing party seats in our cross-country model for secret ballot adoption and test whether it has a positive effect. We find that it does, but the effect is statistically insignificant. Because the coefficient on GDP per capita remains positive and significant, this demonstrates that modernization must work through other channels than expedient elite support for reform.²⁶

Second, another consequence of modernization, which is stressed by, for example, Kitschelt and Wilkinson (2007) and Keefer and Vlaicu (2008), is the emergence of “modern” political parties. That is, parties with either credible policy platforms that individual politicians respect or with organizational structures that enable monitoring of new forms of exchange (other than outright vote buying) between the party and the voters. Once a modern party is established, the value of open voting is reduced, either because the party can contest elections with programmatic policy platforms or because it can support forms of clientelism that do not require open voting. It is, therefore, possible that it is the emergence of modern political parties that triggers secret ballot, and that modernization operates through this channel. To investigate this, we determine the approximate year in which political parties with nation-wide organizations started to contest elections emerged in our Western Europe plus off-shoots sample. The recorded timing pattern relative to the timing of secret ballot reform provides mixed support for this hypothesis (see Supplementary Material Appendix Table S1). In some countries, modern parties did indeed emerge prior to the introduction of the secret ballot (e.g., in Germany and Denmark), but in others, they did not (e.g., in the United Kingdom and the Netherlands). We add a dummy variable capturing the year in which modern political parties emerged to the baseline specification. We find that this variable denoted by “effective parties” itself has a positive, but insignificant, effect on the probability of secret ballot reform. Importantly, the positive effect of the modernization variables remains statistically significant.²⁷ This suggests that the modernization effect did not primarily operate through the emergence of modern political parties.

Third, the historical narrative from the United States suggests that the old regime elites played a proactive role in ballot reform. In fact, it has been suggested that they used the Australian ballot to disenfranchise illiterates among foreign-born citizens and Blacks to obtain a partisan advantage (Heckelman, 2000). If this mechanism is important, we would expect higher support for the Australian ballot in states with lower levels of education that would imply higher levels of illiteracy. This does not seem to be borne out by the data as years of schooling is positively correlated with the probability of ballot reform in the 48 U.S. states.

Forth, modernization may encourage the formation of “public opinion” and induce a fall in the cost of mass communication. Stokes et al. (2013) see the later as a consequence of rising literacy and increasing years of schooling. In our analysis of the 48 U.S. states, we find that GDP per capita and average years of schooling are both significant. This, on one hand, is consistent with the possibility that modernization works through a fall in the cost of mass communication. On the other hand, this cannot be the whole story as income growth has an independent effect. Moreover, Stokes et al. (2013) suggest that campaigning in newspapers did not become common till the first decades of the 20th century. We have, therefore, checked whether the adoptions of secret ballot after 1900 drive the correlation between ballot reform and income in our samples and found that they do not.

Finally, our theory puts a strong emphasis on vote buying and less emphasis on intimidation and coercion. Yet, there are plenty of historical examples of how intimidation and coercion corrupted elections. Our theory offers some scope for discriminating between the two forms of electoral corruption. Although the effect on turnout of secret ballot is unambiguously negative, if vote markets are important, the effect is less clear if electoral corruption takes the form of intimidation and coercion. In that case, voter turnout may actually increase after the Australian ballot was introduced. In contrast to Heckelman (1995) who finds a clear fall in turnout among U.S. states after the vote became secret, the cross-country analysis by Przeworski (2008) shows that the effect can be positive. This suggests that context matters and that both vote buying and intimidation played a role. Importantly, however, whether the direct effect on turnout is negative or positive does not matter for the hypothesis that modernization causes secret ballot adoption. The reason, as argued by Stokes et al. (2013), is that coercion and intimidation are less effective means of controlling the electorate with higher levels of economic development. Accordingly, whether modernization works through the vote market or through a reduction in the returns to coercion and clientelism, the outcome is that modernization makes secret ballot more likely. Our evidence strongly suggests that this was the case.

Conclusion

We “unbundle” the concept of democracy to evaluate in a more nuanced way the interplay between modernization and democratization. Our event history studies of the introduction of the secret ballot demonstrate a remarkably robust relationship between modernization and its adoption. This finding is important because it grants the forces of urbanization, rising education standards, and income growth a role in explaining political development. The role is more limited than envisaged by Lipset (1959), but we contend that modernization, although probably not causally linked to the timing of the major suffrage reforms and thus to the evolution of the overall package of democratic institutions, was a causal factor in getting the secret ballot introduced.

Uncovering the precise mechanism behind this aggregate result requires more refined work than what we have presented here. We have proposed one mechanism based on the logic of a competitive vote market. The reported evidence on the interplay between modernization, ballot reform, and turnout is consistent with this mechanism, but it does not rule out that other forms of clientelism could also have played a major role. More work is clearly needed to gain a fuller understanding of what is behind the relationship between modernization and the introduction of secret ballot.

Footnotes

Acknowledgements

The authors thank the editor of journal and four anonymous reviewers for useful comments and suggestions. The authors also thank Julia Shvets, Miriam Golden, Jo Andrews, James Robinson, Thomas Barnebeck Andersen, Christian Bjørnskov, Carl-Johan Dalgaard, Casper Worm Hansen, Ingrid Henriksen, Lars Lønstrup, Martin Paldam, Paul Sharp, Linda Veiga, Jacob Gerner Hariri, Philipp Schröder, Daniel Ziblatt, Jørgen Elklit, Jan Teorell, Chris Kam, Fabrice Lehoucq, Isabela Mares, Mogens Justesen, John Gerring, and participants in the Harvard workshop “The Causes and Consequences of Electoral Fraud: Lessons From the History of Established Democracies” for helpful comments and useful suggestions on preliminary drafts of this article. The authors also thank Samuel Berlinski, Torun Dewan, and James Fearon, for sharing data with us. This article was partly written while Toke Aidt was visiting the Institute for Quantitative Social Science at Harvard University and, subsequently, the CESifo in Munich. The hospitality of these institutes is much appreciated.

Authors’ Note

The article was presented at the Second World Congress of the Public Choice Society and European Public Choice Society in Miami in March 2012. The usual qualifier applies.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biographies

Toke S. Aidt is a senior university lecturer of economics at the University of Cambridge. Most of his current research focuses on the causes and (economic) consequences of political institutions.

Peter S. Jensen is a professor of economics at the University of Southern Denmark. His current research focuses on political economics, economic growth, and economic history.

References

Acemoglu

Johnson

Robinson

J. A.

Yared

(2008). Income and democracy. American Economic Review, 98, 808-842.

Acemoglu

Robinson

J. A.

(2006). Economic origins of dictatorship and democracy. Cambridge, UK: Cambridge University Press.

Aidt

T. S.

Franck

(2015). Democratization under the threat of revolution: Evidence from the Great Reform Act of 1832. Econometrica, 83, 505-547.

Aidt

T. S.

Giovannoni

(2011). Constitutional rules. Social Choice and Welfare, 37, 219-268.

Aidt

T. S.

Jensen

P. S.

(2009). The taxman tools up: An event history study of the introduction of the personal income tax, 1815-1938. Journal of Public Economics, 93, 160-175.

Aidt

T. S.

Jensen

P. S.

(2014). Workers of the world, unite! Franchise extensions and the threat of revolution in Europe, 1820-1938. European Economic Review, 72, 52-75.

Aidt

T. S.

Leon

(2016). The democratic window of opportunity: Evidence from riots in sub-Saharan Africa. Journal of Conflict Resolution, 50, 694-717.

Alvarez

M. E.

Cheibub

J. A.

Limongi

Przeworski

(1996). Classifying political regimes. Studies in Comparative International Development, 31(2), 3-36.

Anderson

G. M.

Tollison

R. D.

(1990). Democracy in the marketplace. In Crain

W. M.

Tollison

R. D.

(Eds.), Predicting politics: Essays in empirical public choice (pp. 285-303). Ann Arbor: University of Michigan Press.

10.

Anderson

M. L.

(1993). Voter, junker, landrat, priest: The old authorities and the new franchise in imperial Germany. The American Historical Review, 98, 1448-1474.

11.

Andrews

J. T.

Jackman

R. W.

(2005). Strategic fools: Electoral rule choice under extreme uncertainty. Electoral Studies, 24, 65-84.

12.

Angrist

Pischke

J.-S.

(2009). Mostly harmless econometrics. Princeton, NJ: Princeton University Press.

13.

Ansell

Samuels

(2010). Inequality and democratization: A contractarian approach. Comparative Political Studies, 43, 1543-1574.

14.

Ansell

Samuels

(2015). Inequality and democratization: An elite-competition approach. Cambridge, UK: Cambridge University Press.

15.

Baland

J.-M.

Robinson

J. A.

(2007). How does vote buying shape the economy. In Schaffer

F. C.

(Ed.), Elections for sale: The causes and consequences of vote buying (pp. 123-141). Boulder, CO: Lynne Rienner.

16.

Baland

J.-M.

Robinson

J. A.

(2008). Land and power: Theory and evidence from Chile. American Economic Review, 98, 1737-1765.

17.

Barro

R. J.

(1999). Determinants of democracy. Journal of Political Economy, 107, 6-29.

18.

Beck

Katz

J. N.

Tucker

(1998). Taking time seriously: Time-series-cross-section analysis with a binary dependent variable. American Journal of Political Science, 42, 1260-1288.

19.

Bensel

(2004). American ballot box in the mid-nineteenth century: Law, identity and the polling place. Cambridge, UK: Cambridge University Press.

20.

Berlinski

Dewan

(2011). The political consequences of franchise extension: Evidence from the Second Reform Act. Quarterly Journal of Political Science, 6, 329-376.

21.

Blais

Dobrzynska

Indridason

I. H.

(2004). To adopt or not to adopt proportional representation: The politics of institutional choice. British Journal of Political Science, 35, 182-190.

22.

Boix

(1999). Setting the rules of the game: The choice of electoral systems in advanced democracies. American Political Science Review, 93, 609-624.

23.

Boix

(2003). Democracy and redistribution. Cambridge, UK: Cambridge University Press.

24.

Boix

(2011). Democracy, development, and the international system. American Political Science Review, 105, 809-828.

25.

Braun

Kvasnicka

(2013). Men, women, and the ballot: Gender imbalances and suffrage extensions in United States. Explorations in Economic History, 50, 405-426.

26.

Cameron

A. C.

Gelbach

J. B.

Miller

D. L.

(2008). Bootstrap-based improvements for inference with clustered errors. Review of Economics and Statistics, 90, 414-427.

27.

Collier

Vicente

P. C.

(2012). Violence, bribery, and fraud: The political economy of elections in sub-Saharan Africa. Public Choice, 153, 117-147.

28.

Congleton

R. D.

(2011). Perfecting parliament. Cambridge, UK: Cambridge University Press.

29.

Converse

P. E.

(1974). Comment on Burnham’s “theory and voting research.” American Political Science Review, 68, 1024-1027.

30.

Cox

G. W.

Kousser

J. M.

(1981). Turnout and rural corruption: New York as a test. American Journal of Political Science, 25, 646-663.

31.

Craig

F. W. S.

(1977). British parliamentary election results. Great Britain, UK: Macmillan.

32.

Dahl

(1971). Polyarchy: Participation and opposition. New Haven, CT: Yale University Press.

33.

Dal Bo

. (2007). Bribing voters. American Journal of Political Science, 51, 789-803.

34.

Dekel

Jackson

M. O.

Wolinsky

(2008). Vote buying: General elections. Journal of Political Economy, 116, 351-380.

35.

Elklit

(1988). Fra åben til hemmelig afsteming [From open to secret voting]. Aarhus, Denmark: Politica.

36.

Elklit

(2000). Free and fair elections. In Rose

(Ed.), International encyclopedia of elections (pp. 130-134). Washington, DC: CQ Press.

37.

Fearon

(2003). Ethnic and cultural diversity by country. Journal of Economic Growth, 8, 195-222.

38.

Fredman

L. E.

(1968). The Australian ballot: The story of an American reform. Ann Arbor: Michigan University Press.

39.

Galor

Moav

Vollrath

(2009). Inequality in landownership, the emergence of human-capital promoting institutions, and the great divergence. Review of Economic Studies, 76, 143-179.

40.

Gleditsch

K. S.

Ward

M. D.

(2006). Diffusion and the international context of democratization. International Organization, 60, 911-933.

41.

Gundlach

Paldam

(2009). A farewell to critical junctures. European Journal of Political Economy, 25, 340-354.

42.

Heckelman

J. C.

(1995). The effect of the secret ballot on voter turnout rates. Public Choice, 82, 107-124.

43.

Heckelman

J. C.

(2000). Revisiting the relationship between secret ballots and turnout: A new test of two legal-institutional theories. American Politics, 28, 194-215.

44.

Heckelman

J. C.

Yates

A. J.

(2002). Incumbency preservation through electoral legislation: The case of the secret ballot. Economics of Governance, 3, 47-57.

45.

Hicken

A. D.

(2007). How do rules and institutions encourage vote buying? In Schaffer

F. C.

(Ed.), Elections for sale: The causes and consequences of vote buying (pp. 33-60). Boulder, CO: Lynne Rienner.

46.

Kam

(2017). The secret ballot and the market for votes at 19th-century British elections. Comparative Political Studies, 50, 594-635.

47.

Keefer

Vlaicu

(2008). Democracy, credibility, and clientelism. Journal of Law, Economics, & Organization, 24, 371-406.

48.

Keyssar

(2009). The right to vote: The contested history of democracy in the United States. New York, NY: Basic Books.

49.

King

Zeng

(2001). Logistic regression in rare events data. Political Analysis, 9, 137-163.

50.

Kitschelt

Wilkinson

S. I.

(2007). Patrons, clients and policies: Patterns of democratic accountability and political competition. Cambridge, UK: Cambridge University Press.

51.

Kuo

Teorell

(2017). Illicit tactics as substitutes: Election fraud, ballot reform, and contested congressional elections in the United States, 1860–1930. Comparative Political Studies, 50,665-696.

52.

Lee

S. J.

(1994). Aspects of British political history 1815-1914. London, England: Routledge.

53.

Leemann

Mares

(2011). From “open secrets” to the secret ballot: The economic and political determinants of secret ballot reform (Unpublished working paper). New York, NY: University of Columbia.

54.

Leeson

Dean

(2009). The democratic domino theory: An empirical investigation. American Journal of Political Science, 53, 533-551.

55.

Lehoucq

(2007). When does a market for votes emerge. In Schaffer

F. C.

(Ed.), Elections for sale: The causes and consequences of vote buying (pp. 33-60). Boulder, CO: Lynne Rienner.

56.

Lipset

S. M.

(1959). Some social requisites of democracy: Economic development and political legitimacy. American Political Science Review, 53, 69-105.

57.

Lott

J. R.

Kenny

L. W.

(1999). Did women’s suffrage change the size and scope of government? Journal of Political Economy, 107, 1163-1198.

58.

Mackie

T. T.

(2000). The secret ballot. In Rose

(Ed.), International encyclopedia of elections (pp. 289-290). Washington, DC: CQ Press.

59.

Mares

(2015). From open secrets to secret voting: Democratic electoral reforms and voter autonomy. New York, NY: Cambridge University Press.

60.

McCook

J. J.

(1892). The alarming proportion of venal voters. Forum, XIV, 1-13.

61.

Miller

M. K.

(2012). Economic development, violent leader removal, and democratization. American Journal of Political Science, 56, 1002-1020.

62.

Moore

(1966). The social origins of democracy and dictatorship. Boston, MA: Beacon.

63.

Ostrogorski

(1964). Democracy and the organization of political parties. Chicago, IL: Quadrangle Books.

64.

Persson

Tabellini

(2009). Democratic capital: The nexus of political capital and economic Change. American Economic Journal, 1, 88-126.

65.

Przeworski

(2008). Constraints and choices: Electoral participation in historical perspective. Comparative Political Studies, 42, 4-30.

66.

Przeworski

(2009). Conquered or granted? A history of suffrage extensions. British Journal of Political Science, 39, 291-321.

67.

Przeworski

(2015). Suffrage and voting secrecy in general elections. In Elster

(Ed.), Secrecy and publicity in votes and debates. New York, NY: Cambridge University Press.

68.

Przeworski

Limongi

(1997). Modernization: Theories and facts. World Politics, 49, 155-183.

69.

Retallack

(1988). Notables of the right: The conservative party and political mobilization in Germany, 1876-1918. Boston, MA: Unwin Hyman.

70.

Rokkan

(1961). Mass suffrage, secret voting and political participation. European Journal of Sociology, 2, 132-152.

71.

Rueschemeyer

Stephens

E. H.

Stephens

J. D.

(1993). The impact of economic development on democracy. The Journal of Economic Perspectives, 7(3), 71-86.

72.

Rusk

J. G.

(1974). Comment: The American electoral universe: Speculation and evidence. American Political Science Review, 68, 1028-1049.

73.

Schofield

(2004). Jeremy Bentham, the French revolution and political radicalism. History of European Ideas, 30, 381-401.

74.

Schorske

(1955). German social democracy, 1905-1917. Cambridge, MA: Harvard University Press.

75.

Seymour

(1915). Electoral reform in England and Wales. New Haven, CT: Yale University Press.

76.

Snyder

J. M.

(1991). On buying legislatures. Economics & Politics, 3, 93-109.

77.

Stokes

(2005). Perverse accountability: A formal model of machine politics with evidence from Argentina. American Political Science Review, 99, 315-325.

78.

Stokes

Dunning

Nazareno

Brusco

(2013). Brokers, voters, and clientelism: Puzzle of distributive politics. Cambridge, UK: Cambridge University Press.

79.

Treisman

(2015). Income, democracy, and leader turnover. American Journal of Political Science, 59, 927-942.

80.

Wejnert

(2005). Diffusion, development, and democracy, 1800-1999. American Sociological Review, 70, 53-81.

81.

Ziblatt

(2008). Does landholding inequality block democratization? A test of the “bread and democracy” thesis and the case of Prussia. World Politics, 60, 610-641.

82.

Ziblatt

(2009). Shaping democratic practice and the causes of electoral fraud: The case of nineteenth-century Germany. American Political Science Review, 103, 1-21.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.54 MB