The Roots of Direct Democracy in the United States: South Dakota’s 1898 Referendum Creating the First Statewide Initiative Process

Abstract

We investigate voter preferences for changes in voting rules, focusing specifically on the creation of citizen-initiative processes that were originally adopted in South Dakota in 1898 and eventually enacted by half of the states. Various claims have been advanced about why the process was adopted and who supported or opposed it, but without presenting evidence from referenda where voters approved the creation of the process. We test these claims by examining county-level election returns from South Dakota’s 1898 referendum that created the first statewide initiative process in the United States. We find that support for the initiative process was generally higher among groups that are disadvantaged in various ways by existing representative institutions and perceive advantages in creating direct democratic institutions capable of bypassing representative processes. These findings stand in contrast to the notion that the adoption of constitutional rules will be relatively free from calculations rooted in self-interest and perceived advantage from the rules changes.

Introduction

Direct democratic institutions play a prominent role in governance in many of the 26 states that allow for the initiative and/or referendum. The initiative process is used on a regular basis to enact policies regarding marijuana legalization, minimum-wage increases, taxes, redistricting commissions, and animal welfare that are blocked by governing officials but supported by the public. Voters have also forced referenda on policies that are passed by state legislators but turn out not to enjoy public support and are overturned by voters, as has occurred with referenda overturning state laws limiting the power of labor unions, expanding school choice, and restricting abortions. States that allow voter initiatives are more likely to adopt policies consistent with majority preferences, leading Boehmke (2005) to conclude that direct democratic institutions empower broad-based citizen groups and counteract the influence of narrow special interest groups. Shadbegian (1998) and Sun (2014) find the presence of voter initiative and referendum process and the success rate of passing voter referenda are correlated with the adoption of tax and expenditure limits.

In view of the importance of the initiative and referendum for state policy-making, it is important to understand the origins of these institutions. South Dakota was the first state to adopt direct democracy, when the legislature crafted a state constitutional amendment establishing the initiative and referendum and voters approved the amendment in 1898. During the 1900s and 1910s, 20 more states adopted and retained the initiative and/or referendum. Generally, this took place through the same process as in South Dakota, with legislators submitting to voters the question of whether to adopt direct democracy and voters giving their approval. After a number of decades when no additional states joined the ranks of states with direct democracy, five more states added the initiative and/or referendum between the late 1950s and early 1990s.

In explaining the origins of direct democracy, some scholars have focused on the role of legislators and investigated why they were willing to reduce their own power by allowing establishment of the initiative and referendum. Of the 26 states that currently allow for direct democracy in some fashion, legislature-referred amendments were the vehicle for adopting these institutions in two-thirds of these states, with constitutional conventions responsible for creating them in the remaining states. It is understandable why conventions were willing to create institutions that empowered the public and constrained the legislature. It is more surprising – and this puzzle has generated several recent studies – that legislators were willing to do so on a number of occasions. Heightened interparty competition in state legislatures and the prominence of third parties appears to have played a role in explaining why legislators were in some states willing to submit to voters the question of creating direct democracy (Smith and Fridkin 2008).

Although direct democracy measures clearly reduce the power of the state legislature, the effect on gubernatorial power is less certain. While the initiative process allows an avenue for the governor to bypass the legislature when party control of the legislative and executive branches is divided, voters can also thwart a governor’s veto in the same way. Thus, the effect on the relative power of the governor would depend on whether or not there is a coincidence of wants between the governor and median voter (Matsusaka 2008, 118–119). Using campaign expenditures as a proxy for gubernatorial power, Randolph (2011) finds governors in states with the initiative process have reduced power compared to governors in other states.

In this article, we investigate a different aspect of the origins of direct democracy, focusing not on governing officials but rather on voters and analyzing their decision to approve measures establishing the initiative and referendum. Although the expectation is that voters would be willing to enhance their power by gaining the ability to initiate and overturn policies, voters on a handful of occasions have rejected measures that would have created the initiative and referendum. Even in states where voters approved measures establishing the initiative and/or referendum, a large number of votes were cast against creating these institutions. In this study, we analyze determinants of voter support for and opposition to direct democracy, by studying county-level election returns for South Dakota’s 1898 vote that created the country’s first state-wide initiative and referendum process and was followed in quick succession by adoption of direct democracy in a number of other states.

In doing so, we contribute to a longstanding debate about the role of interests in creation of constitutional rules. Buchanan (1976) argued that rules developed at the constitutional stage for establishing the manner in which post-constitutional questions are considered will be free from bias due to the inherent uncertainty over one’s future station in life. Yet, empirically, there is growing evidence that constitutions are often developed and amended with personal interests in mind (McGuire and Ohsfeldt 1984; Kenny and Rush 1990; Heckelman and Dougherty 2007).

Adopting the latter view, we expect voters’ level of support for direct democratic institutions will be driven by calculations about whether they are advantaged or disadvantaged by the design and operation of representative institutions. Voters will be most likely to support the initiative and referendum when their political party, policy preferences, or socio-demographic group are not adequately represented in the legislative process. This could be a result of one party enjoying electoral dominance and voters who affiliate with the minority party or a third party supporting direct democracy as a vehicle to force consideration of measures that would otherwise be kept off the legislative agenda. Voters could also be led to support the initiative and referendum because they strongly support particular policy issues that have failed to pass due to influence wielded by interest groups in the legislative process or for other reasons. Finally, voters could be led to support direct democracy because they are members of socio-demographic groups that are not adequately represented in the legislative process, whether due to malapportionment of legislative districts or because members of certain regional, religious, racial, or occupational groups are under-represented in legislative assemblies.

We test these expectations by considering whether variation in county-level support for South Dakota’s 1898 referendum creating direct democracy is correlated with partisan affiliation, issue preferences, and socio-demographic factors. Our overall expectation that support for direct democratic institutions would be higher among groups disadvantaged in representative institutions is generally supported. When considering the role of partisan affiliation, areas of the state exhibiting higher levels of support for the insurgent Fusionist Party registered higher levels of support for direct democracy, as expected. In terms of issue preferences, areas of the state that supported women’s suffrage registered higher levels of support for direct democracy, also as expected. Regarding socio-demographic groups, rural areas were on balance more likely to oppose direct democracy, in line with our expectation that voters in rural areas that were advantaged in the legislative apportionment process would oppose allowing measures to be put to a direct popular vote where malapportionment advantages would be reduced. Wealthier areas of the state, as measured by farm and manufacturing output, were, as expected, more likely to oppose direct democracy, given that interests representing the wealthy would typically be able to wield influence more readily in the legislature than through direct democratic institutions. Although we expected support for direct democracy to be higher in areas with high percentages of Catholics, who were less well represented than Protestants in public offices, support for direct democracy was actually lower in heavily Catholic areas, perhaps due to Catholic voters’ opposition to prohibition and their well-founded expectation that the initiative process would be a vehicle for adopting prohibition.

Analyzing popular support for South Dakota’s referendum on establishing the initiative and referendum contributes to a better understanding of the origins of direct democracy in the first state to adopt these institutions. Moreover, focusing on county-level election results and investigating variation in support for direct democracy across counties in a single state enables scholars to move beyond cross-state studies that take the state to be the unit of analysis and explain variation in state adoption of these institutions.

Scholarship on the origins of direct democracy

Historians and scholars of American political development have traced the philosophical roots of the initiative and referendum in the U.S. Some studies have focused on the role of the Socialist Labor Party and Populist Party in including these institutions in their late-19th-century platforms (Ellis 2023; Goebel 2002). Attention has also been paid to the influence of individuals such as W.J. Sullivan and William S. U’Ren, who championed the initiative and referendum and played critical roles in pressing for their adoption (Ellis 2022; Schuman 1994). Scholars have also focused on the pioneering role of Western states in adopting these direct democratic devices (Persily 1997) and their eventual spread to some states east of the Mississippi (Piott 2003), as well as the debates throughout the Progressive era about the advantages and disadvantages of allowing the people to rule directly (Dinan 2006).

In recent years, political scientists have undertaken several empirical analyses of the factors associated with adoption of direct democracy. In general, these studies investigate why direct democracy was adopted in some states but not others. The focus has been in part on explaining why legislators were willing to reduce their own power by framing state constitutional amendments creating the initiative and referendum (Smith and Fridkin 2008; Bridges and Kousser 2011). Other studies have focused more generally on patterns of state adoption of these institutions (Bowler and Donovan 2006; Lawrence et al. 2009). All of these studies focus on the state as the unit of analysis and seek to explain variation in which states adopted direct democracy and the speed with which they did so.

Few studies have investigated the origins of direct democracy by analyzing county-level election returns for referenda creating these institutions in particular states. Dinan and Heckelman (2020) analyzed county-level results of California’s 1911 vote to adopt the initiative and referenda, as part of a broader project assessing the coherence of Progressive-era reforms. Otherwise, little has been done to analyze county-level election results on referenda to establish direct democratic institutions, in the way that we undertake in this paper analyzing South Dakota’s 1898 referendum.

South Dakota’s 1898 referendum creating direct democracy

South Dakota’s 1898 vote on adopting the initiative and referendum holds particular interest. South Dakota was the pioneer in enacting these direct democratic institutions. South Dakota voters’ approval of the initiative and referendum was then quickly followed by similar votes and approvals in a number of other (mostly western) states during the next two decades. The initiative and referendum were eventually adopted in nearly half of the states and have played a prominent role in governance in the ensuing 125 years.

South Dakota’s 1898 referendum has been analyzed by several scholars, generally with a focus on the persons, groups, and parties that played roles in getting this measure on the ballot (Ellis 2002, 26–27; Grant 1973; Piott 1992). Several individuals are credited with playing key roles in recommending that South Dakota adopt direct democratic institutions and building support for them. Robert W. Haire, a Catholic priest who held leadership roles with the Knights of Labor, began in the late 1880s to advocate for the adoption of an initiative process (Tiffany 1924, 331–332). Henry L. Loucks, who was elected president of the Dakota Farmers’ Alliance in the mid-1880s, eventually became another leading proponent of direct democracy in South Dakota. In 1890, members of the Knights of Labor and the Farmers’ Alliance joined forces to create the Independent Party, which later took the name of the People’s (Populist) Party and became the biggest champion of the initiative and referendum in South Dakota throughout the 1890s (Grant 1973, 395–397).

In the 1896 election, the Populist Party won the governor’s office, when Populist/Fusionist candidate Andrew E. Lee headed a coalition of Populists, Democrats, and Silver Republicans and defeated Republican A. O. Ringsrud by 319 votes (Lindell 1992, 347). The 1896 election also led to the Fusionist coalition gaining a majority in the state legislature (Piott 1992, 189). In the 1897 session, the legislature approved placing on the ballot a state constitutional amendment creating the initiative and referendum, by a 49-32 vote in the house and a 26-17 vote in the senate, with most of the supporting votes coming from the Fusionist coalition and most of the opposing votes coming from Republicans. In the November 1898 election, South Dakota voters approved the amendment by a margin of 23,816 to 16,483.

The initiative and referendum measure was one of three legislature-referred amendments that appeared on South Dakota’s 1898 ballot, which also included elections for state and federal office-holders. The other two amendments dealt with enfranchising women and giving the state government full responsibility for the manufacture and sale of intoxicating liquor. The woman suffrage amendment was defeated, 22,983 to 19,698. The liquor regulation amendment was approved, 22,170 to 20,557. The number of votes cast on the initiative and referendum amendment amounted to over 90% of the total votes cast on the other two amendments, and more than half of the votes recorded in the gubernatorial election.

Scholars who have analyzed South Dakota’s adoption of the initiative and referendum have not undertaken sustained investigations of county-level election returns on the 1898 vote, in the way we carry out in this paper. In his study of the origins of direct democracy in South Dakota, Grant (1973, 405) simply noted that the amendment “carried in all parts of the state – wheat, corn, ranching, and mining areas” and failed only in counties that were “largely populated by conservative Russian-German farmers.” Meanwhile, in his study of South Dakota, Piott (1992, 190) after noting similarly that the amendment “carried in all parts of the state,” went on to mention that the “largest bloc in opposition came from four counties … clustered in the southeast corner of the state” and that “the ‘city’ vote in South Dakota also seemed to support the amendment.”

The distribution of county support is pictured in Figure 1, calculated from the vote totals listed in Official Votes of South Dakota by Counties (1912, 59). The five counties showing the lowest support were (in order) the following: Hutchinson (18.7%), Bon Homme (25.7%), Marshall (27.3%), Turner (36.9%), and Campbell (38.4%). Three of the counties share a common border with Yankton (42.1%) in the southeast but the other two are separated by Brown (57.0%) and McPherson (58.6%) at the very top of the state. Three additional counties failed to majority support the amendment (Aurora – 45.2%, Gregory – 48.0%, Faulk – 48.9%) but these counties do not border any of the others with only minority support. At the other extreme, four counties returned greater than 3/4th support: Buffalo (75.7%), Hand (76.1%), and Sully (81.4%) in the central part of the state, and Butte (78.1%) in the northwest corner. The first three share a common border with Hyde (62.2%), which is closer to the median county Beadle (62.8%) in terms of support. Beadle also flanks Hand to the east. In sum, there appears to be little overall spatial relationship among the counties despite Piott’s (1992) observation.

Figure 1. Support for initiative and referendum amendment.

Hypotheses and data

We expect support for creating direct democratic institutions to be greatest among groups disadvantaged by the design and operation of representative institutions and opposition to be highest among groups that are advantaged in some way by the status quo of representative institutions. Our expectation is that out-groups would support direct democracy as a way of securing a vehicle for placing issues on the policy agenda that are blocked by governing officials but potentially supported by the public. This disjunction between the preferences/priorities of governing officials and the public could be a product of legislative malapportionment that gives certain regions or parties an undue advantage in legislative seats. The disjunction could also be a result of certain groups and interests enjoying outsized representation and influence in the legislative process while other groups and interests are under-represented.

This general expectation leads to a series of specific hypotheses. First, we expect support for direct democratic institutions to be greater among persons who are affiliated with the minority party or third parties and who would view the initiative process as a way to place issues on the policy agenda that may be blocked by the dominant parties in the legislature. After being granted statehood in 1889, South Dakota Republicans dominated the electoral landscape until suffering narrow losses in 1896. Throughout the 1890s, the Populist Party heavily pushed the initiative and referendum, and after a Populist/Fusionist Party candidate won the gubernatorial election in 1896 and Populists and Democrats gained a majority in the legislature, they viewed these direct democratic institutions as a means to circumvent long-standing Republican party dominance (Lawrence et al. 2009, 1028–1029). Indeed, the governorship and state legislature returned to Republican control following the 1900 election, which was the next election after the initiative and referendum amendment was approved by voters in 1898. We therefore expect that areas registering high levels of support for the insurgent Fusionist Party would back direct democracy, whereas areas exhibiting high support for the long-dominant Republican Party would be less likely to back direct democracy. To test this, we utilize the percentage of the county two-party vote between the Republican and Fusionist candidates that favored the Fusionist candidate in the 1898 gubernatorial election.¹

Second, we expect support for direct democracy to be higher among persons who support specific policy issues on which governing officials’ support lagged behind public support, whether because of a disjunction between the preferences of the public and elected officials or because interest groups wielded undue influence in the legislative process and could block the passage of these policies in the legislature. In particular, we expect advocates of enfranchising women to support the initiative process as a way of increasing the probability of securing its passage. Liquor companies and interests were seen as wielding undue influence over legislators and the legislative process and were able to prevent the passage of restrictive liquor policies that enjoyed public backing. These same companies and interests were also opposed to enfranchising women, out of a fear that women supported liquor prohibition; these companies and interests were joined by other entrenched interests that opposed woman suffrage for various reasons. Supporters of woman suffrage could have been expected to support creation of the initiative process as a way of overcoming these interests (Sponholtz 1973, 57). Figure 2 demonstrates the positive bivariate correlation between supporting women’s suffrage and the initiative and referendum amendments ( $ \rho =0.68 $ ).

Figure 2. Amendment support scatter plot.

We also expect areas with higher levels of labor union activity to be more supportive of the initiative and referendum, which were viewed as a means of overcoming corporate-dominated legislatures and securing passage of various worker-protective measures supported by the public but blocked in the legislature, including an eight-hour work-day, workers’ compensation programs, and bans on child labor and blacklists (Sponholtz 1973, 46–47, 50–51). We focus in particular on the role of the Knights of Labor, a leading proponent of the initiative and referendum in South Dakota in the early 1890s (Sannes 1992, 412). We rely on a database compiled by Garlock (n.d.), who identified each county where one or more Local Assemblies of the Knights of Labor were active and the dates when they were active. We assign a 1 to each of the sixteen counties that had at least one Knights of Labor Local Assembly operating in the decades leading up to the 1898 referendum, and we assign a zero to the remaining counties.

Third, we expect support for direct democracy to be affected by whether particular socio-demographic groups are advantaged or disadvantaged by legislative apportionment or by their degree of representation and influence in the legislature. The apportionment process has long benefited rural areas at the expense of urban areas in most states (Sponholtz 1973, 57), and this was generally true in South Dakota as well (Clem 1961, 4); therefore, we expect rural areas to be more opposed to direct democracy and urban areas to be more supportive. We also expect areas with greater wealth to be more opposed to direct democracy, because advantages enjoyed by rich persons and interests in the legislative process would be diminished when measures were put to a direct popular vote, especially because backers of the initiative process sought to use the process as a vehicle for increasing taxes. We also expect areas with higher percentages of Catholic voters to be more supportive of direct democracy, because Catholics were under-represented in the state’s Protestant-dominated political leadership and officeholders (Lauck 2010, 71).

The number of rural inhabitants in each county is taken from the 1900 Census and divided by the total population of the county. As posted on their website, The Census defines “rural” as all territory, persons, and housing units not defined as urban. Our measure of the rural percentage of each county represents the proportion of county residents not in an area of at least 2,500 persons.² The wealth of a county is measured by the value of farm output per capita and manufacturing output per capita. Farm and manufacturing output values are taken from the 1900 Census.³ Religious designations were not enumerated in the decennial census until 1940. Instead, there were periodic Census publications based on religious organizations reporting their own membership numbers. The most recent report of religious institutions at the time had been published in 1890 and the next occurred in 1906. We use the average of the 1890 and 1906 values normalized by the 1900 total county population. Our proxy measure for Catholics is the percentage of the population holding membership in a Roman Catholic church.

The census data we used are distributed by the Inter-university Consortium for Political and Social Research (ICPSR). All census data reflecting 1900 values are taken from ICPSR 2896, which represents an updated, expanded, and corrected version of ICPSR 0003 (Haines 2010). Religious data for 1896 and 1906 are taken from ICPSR 0008. Descriptive statistics for all variables are reported in Table 1.⁴

Table 1. County-level Descriptive Statistics (n = 50)

Regression analysis on support for I&R amendment

Although our primary goal is identifying the type of voters who tended to cast a ballot in favor of, or in opposition to, direct democracy, we stress that reliance on aggregated data prevents us from doing this directly. Consider, for example, our measure of partisan impacts, as proxied by the percentage of county votes for the Fusionist candidate in the 1896 gubernatorial election. A positive correlation will occur if more votes for the Fusionist candidate align with more votes for the initiative and referendum process. This would be consistent with either Fusionist voters consistently supporting the initiative and referendum amendment more often than Republican voters, or Republican support for the initiative and referendum amendment depending on the concentration of Fusionist voters in their county. While we are careful in our interpretations to avoid causal language, we do presume a constant marginal effect from each independent variable throughout, so that the behavior of one group does not change due to the presence of another group in the same county. We would, therefore, present the former rather than the latter explanation if a positive coefficient is estimated for the Fusionist variable, even though the latter is possible. With these caveats in mind, we now proceed to present our results.

In column 1 of Table 2, we report estimated coefficients and robust standard errors from our primary specification. All variable coefficients have the expected sign, except for areas with higher percentages of Catholics. We expected support for direct democracy to be greater in areas with higher percentages of Catholics, who wielded less political influence than Protestants and who might have been expected to view the initiative and referendum as vehicles for gaining influence in the political process. Yet areas with higher percentages of Catholics were less supportive of direct democracy.

Table 2. County-level support for initiative and referendum amendment

Standardized coefficients in brackets and robust standard errors in parentheses.

* significant at the 10 percent level.

^** significant at the 5 percent level.

^*** significant at the 1 percent level.

This result is consistent with the findings of various scholars who have undertaken state-based analyses of the adoption of direct democracy, but there are various possible ways to account for this relationship between areas with higher percentages of Catholics and lower levels of support for the initiative and referendum. Lawrence et al. (2009, 1032) reasoned that the presence of high proportions of Catholics in some states may have led Protestant voters and officials in those states to oppose direct democratic institutions because they feared that these institutions “could alter the relative power of social groups” by benefiting Catholics. We offer an alternative explanation for opposition to direct democracy being higher in areas with higher percentages of Catholics. Catholics generally opposed alcohol prohibition (Dinan and Heckelman 2014), and, despite their minority status, were able to get their way in South Dakota on this issue from the mid-1890s onward, because their preferences happened to be aligned with influential special interests that were successfully blocking prohibition. Direct democracy was expected to serve as a vehicle for adopting prohibition, as occurred in a number of states during the next several decades (Sponholtz 1973, 51–52). In fact, the first initiated measure to qualify for the ballot in South Dakota, a decade after adoption of the amendment creating the initiative process, sought to enact a local-option alcohol policy, whereby localities could vote to prohibit alcohol in their jurisdictions (Ballotpedia n.d.). The bottom line is that higher percentages of Catholics in a county are associated with lower support for direct democracy; but the ecological fallacy prevents us from distinguishing between two competing explanations for this relationship.

Our other hypotheses are supported. Areas with higher percentages of rural residents are associated with less support for adopting the initiative and referendum, whereas greater support for the Fusionist party candidate and women’s suffrage are associated with more support for the initiative and referendum. More wealth in the county, whether in the form of farm or manufacturing output, reduces support for this direct democracy measure, although the former measure is not statistically significant. Support for the initiative and referendum is higher in counties that had (concurrently or prior) an established Knights of Labor association, but not to a statistically significant degree.

Based on our estimates, a 10 dollar per person increase in county wealth from manufacturing suggests a 1.77 percentage point decline in support for the initiative and referendum but only about a half percentage point decline when the same wealth increase comes from farm output. Referring back to Table 1, farm output value is roughly five times larger on average than manufacturing output value; the same 10 dollar increase represents a seven percent increase in farm output value but a whopping forty percent increase in output value generated from manufacturing. Likewise a 10 percentage point change in the Catholic population is associated with a five percentage point, or eight percent, change in initiative and referendum support, whereas a 10 percentage point change in the rural population yields only a two percentage point, or three percent change, in initiative and referendum support on average. To make the coefficient estimates more comparable, we present standardized coefficient estimates in brackets below the OLS coefficient estimates. Our estimates suggest that a one standard deviation change in the rural population of a county would alter support for direct democracy by one-fifth of a standard deviation, whereas a one standard deviation change in the percentage of the Catholic population would impact support by one-fourth of a standard deviation. In terms of voting returns, a one standard deviation change in county-level support for women’s suffrage suggests a little over one-half standard deviation change in support for the initiative and referendum, by far the largest effect.

We check robustness in a number of ways. We note Catholic church members represented 10.4% of the county population on average but 25.9% of current church members in a county on average. Redefining the religion variables as percentages where the base was limited to the number of people who were members of a church, rather than the entire county population, did not affect signs or significance of the Catholic variable.

When the dependent variable is a percentage, as with our model here, a logistic transformation⁵ is sometimes preferred instead of assuming linearity throughout. Constant marginal effects can result in nonsensical predicted values outside the logical range of 0 to 100. For example, a hypothetical county with the sample maximum values for the political variables and the sample minimum values for its demographic and wealth variables⁶ is predicted to have 103% of its votes cast in favor of direct democracy. One way to avoid such nonsensical predictions is by transforming the dependent variable into a logit where the estimated coefficients now represent the marginal effect on the log of the odds that someone votes ‘yes’ on the initiative and referendum amendment. Estimates using the logistic transformation are presented in the second column of Table 2. All signs and levels of statistical significance remain the same. (The same hypothetical county in the previous example would now have a predicted logit of 2.26, or 91% predicted yes votes.)

The median voter in the state is typically not the median of the county medians due to malapportionment and different distributions of voters and abstainers across the counties. Voters in areas disadvantaged through the apportionment process would have greater incentive than others to favor the initiative, and to turn out to vote on the initiative, because their votes would be more equally represented in referenda than in the legislature. Counties with more active voters might also expect the resulting state median voter to be closer to their median voter on those issues in which the local electorate shows a strong majority support or opposition. On the other hand, turnout differences on the amendment would have no impact on determining the median legislator. This might suggest counties with greater turnout to disproportionately show greater support for direct democracy. We test this incentive by including the turnout for the initiative and referendum amendment, defined as the ratio of total number of votes cast on this question relative to the number of males aged 21 and above, representing the age and gender-eligible electorate. Estimates are presented in the final column of Table 2. The coefficient on this additional variable is positive but not statistically significant, and its inclusion has little effect on the other estimated parameters.

Conclusion

The benefit of analyzing county-level election returns for South Dakota’s 1898 referendum creating an initiative process is to gain a better understanding of which groups of voters supported direct democratic institutions in the Progressive Era. Because South Dakota was the first state to institute a state-wide initiative process, this referendum holds particular interest. Whereas other scholars have investigated variation in support for the initiative process by explaining why certain states were more likely to adopt direct democracy, in this study we focus on county-level election results within a particular state, with the goal of generating insights about who supported and opposed direct democracy.

Our expectation is that support for direct democracy would be highest among groups that are for various reasons disadvantaged in the legislative process. These disadvantages can be a product of legislative malapportionment plans that reduce the influence of certain areas of the state and increase the influence of other areas. These disadvantages can also be a result of over-representation of members of certain demographic groups in state offices and the under-representation of members of other groups. These disadvantages can also stem from the ability of certain interest groups to wield influence in the electoral or legislative process and to the detriment of other groups and interests. In each of these situations, groups that are disadvantaged in the legislative process would be expected to support creation of the initiative process as a means of achieving outcomes unattainable through the legislature and groups advantaged in the legislative process would be expected to oppose the initiative process.

In considering the role of partisanship, policy motivations, and demographic characteristics in accounting for variation in counties’ support for direct democracy, we generally find support for our expectations that groups that were disadvantaged in representative institutions would be more supportive of, and groups that enjoyed advantages in representative institutions would be more opposed to, the initiative process. Areas of the state registering greater support for the insurgent Fusionist Party were more likely to support direct democracy. Areas of the state demonstrating greater support for a woman suffrage amendment were also more likely to support direct democracy. Rural areas were less likely to support direct democracy, as were areas with greater wealth, particularly based on manufacturing.

Although our study focuses on South Dakota, which pioneered direct democracy, we expect our analysis and findings to apply to other states that considered adopting the initiative and referendum, in the sense that the particular groups disadvantaged in the political process in each state would be more supportive and groups advantaged in the political process in each state would be more opposed to these institutions. Analyzing election results at the county level, as we have done in this study, provides a way of moving beyond cross-state studies and delving into the dynamics of particular states, whether other states that adopted direct democracy in the Progressive Era or the handful of states that adopted direct democracy in the latter part of the 20th century.