Who Counts? Non-Citizen Residents, Spatial Sorting, and Malapportionment

Abstract

Existing research argues that malapportionment primarily favours rural areas, resulting in conservative biases of electoral systems. In this paper, we provide a new perspective on the study of apportionment processes by identifying the institutional design under which malapportionment may favour other regions. Because of the geographical sorting of non-citizen residents, we argue that regions with high shares of non-citizen residents benefit from population-based apportionment, whereas the spatial sorting of non-citizens does not affect malapportionment in the case of citizen-based apportionment. Empirically, we use sub-national data from ten advanced democracies to forward evidence that differences in apportionment mechanisms and district-level shares of non-citizen residents systematically influence malapportionment. Our findings suggest that the impact of malapportionment on political representation and public policies may be more heterogeneous than previously thought.

A switch to the use of citizen voting age population as the redistricting population base for redistricting would be advantageous to Republicans and Non-Hispanic Whites.

Thomas B. Hofeller

Introduction

There is a long-standing debate in the USA about whether undocumented immigrants should be excluded from the numbers that Congress uses in apportionment, that is, the process by which seats in the House of Representatives are distributed among the fifty states. If excluded, states with larger populations of undocumented immigrants would lose seats in the House of Representatives. Republican lawmakers, in particular, continue to demand that non-citizen residents should not be considered in the apportionment process (Frost 2021). The USA is not the only case where the role of non-citizen residents in apportionment is controversial (cf. Städler 2016). This paper therefore asks: Does the choice of the relevant population for apportionment have a systematic effect on political representation?

Malapportionment denotes systematic biases in democratic systems because some geographical regions benefit from numerically better political representation than others (Albertus and Menaldo 2018; Ardanaz and Scartascini 2013; Kamahara, Wada and Kasuya 2021). Hence, in malapportioned systems, ‘the votes of some citizens weigh more than the votes of other citizens’ (Samuels and Snyder 2001, 654). Such malapportionment has been identified as a key reason for poor economic outcomes, low progressive taxation, high inequality, and severe fiscal imbalances (Horiuchi and Saito 2003; Ardanaz and Scartascini 2013; Bhavnani 2018, 2021; Hiroi 2019; Ansolabehere, Gerber, and Snyder 2002; Rodden 2002; Beramendi, Rogers, and Díaz-Cayeros 2022; Horiuchi 2004).

The empirical literature argues that malapportionment typically favours rural regions and conservative parties (Albertus and Menaldo 2018; Bruhn, Gallego and Onorato 2010; Boone and Wahman 2015; Bhavnani 2018; Ziblatt 2009; Cox and Katz 2002; Ansolabehere and Snyder 2008; Moriwaka 2008; Johnson and Miller 2023). There are two reasons for this pattern. First, constitutional rules sometimes guarantee regions a minimum number of parliamentary seats, independent of their size. Second, malapportionment may result from the non-adaptation of districts despite unequal population growth. In both cases, malapportionment benefits districts that suffer from depopulation – often rural regions – and the parties that represent them, which are often conservative (Rodden 2019; Ford and Jennings 2020).

Another factor influencing the apportionment process, which has been curiously overlooked in the literature, is whether the allocation of parliamentary seats to geographical regions is based on the regions’ total population or the number of citizens or voters. This difference would not matter if non-citizen residents, primarily immigrants, were equally distributed across the territory. However, this is rarely the case. Non-citizen residents tend to concentrate in urban areas (OECD 2016), and urban areas tend to support different political parties than rural areas (Ford and Jennings 2020). As a result, the representational consequences of the apportionment process might be more diverse than generally assumed. Depending on the specific rules for electoral district design, apportionment might even benefit urban regions and left parties.

This paper contributes to the literature on how electoral rules impact political representation and public policies (Becher 2016; Döring and Manow 2017; Iversen and Soskice 2006). We show that the allocation of seats to geographical regions varies between countries with approximately half of the countries apportioning parliamentary seats based on the total population instead of other reference groups such as citizens or voters. Using a novel, sub-national data set covering ten advanced democracies, we demonstrate that the choice of reference groups has a systematic and substantial effect on malapportionment. Finally, we show that the share of non-citizen residents at the district level systematically influences political representation when countries rely on population-based apportionment, but not when countries rely on citizen-based apportionment. Depending on the parties’ electoral geography, the choice of apportionment mechanism thus influences the political representation of parties. In this way, the paper adds an important comparative perspective to a debate that has so far mainly focused on the representation of minorities and been limited to the USA (Baumle and Poston 2004; Chen and Stephanopoulos 2021; Haas, Miller and Kimbrough 2022).

Who Counts for Apportionment?

In an unbiased electoral system with districts, every vote has the same impact on the composition of the legislature, meaning that the share of voters should be equal to the share of representatives elected in a district. To achieve this goal, district boundaries or magnitudes are adjusted, and new districts are created to account for population shifts. Given that this process must consider multiple, sometimes conflicting, goals such as maintaining traditional communities while creating equally sized districts, deviations from such perfectly apportioned systems are common, perhaps unavoidable. However, sometimes, these deviations are large and lead to an unfair distribution of representatives in the legislative body.

Malapportionment can be the result of institutional design, as some groups are deliberately overrepresented to gain partisan advantage (Ardanaz and Scartascini 2013; Ostwald 2013; Giugăl et al. 2017), or because some groups are excluded because they are considered unfit for political participation, such as felons in the USA (Baumle and Poston 2004). For most advanced democracies, the literature typically highlights two main sources of malapportionment.

First, constitutions or electoral laws can contain rules that guarantee geographical regions a minimum number of seats in the legislature. For example, Australia, Canada, Norway, and Switzerland have adopted such quotas for lower chamber elections, but similar rules also exist around the world in countries such as Turkey, Uruguay, and Ecuador (Nohlen, Grotz and Hartmann 2004; Nohlen 2005; Nohlen and Stöver 2010). While these quotas are typically designed to guarantee minority representation, they primarily benefit sparsely populated regions. Due to global trends towards urbanization (United Nations 2019; Baeumler et al. 2021), their distributional consequences are often skewed in favour of rural regions and the parties representing these regions’ interests (Johnson and Miller 2023).

Second, differences in representation may result from not reapportioning electoral districts despite significant population shifts. A widely known example is Imperial Germany in which no redistricting took place between 1871 and 1912 (Boix 2010, 406). Among advanced democracies, the USA, before the Supreme Court’s reapportionment decisions in the 1960s, is another prominent example of non-reapportionment despite large population movements (Ansolabehere and Snyder 2008; Cox and Katz 2002), as is the decade-long freezing of the seat distribution in India since 1976 (Bhavnani 2018). In the case of non-reapportionment, regions with low population growth gain most from freezing the existing district boundaries. Once again, rural interests are the likely beneficiaries of blocking the reapportionment process (see also Sauger and Grofman (2016) on France).

This paper argues that there is a third, often overlooked factor influencing apportionment, which has the potential to benefit urban regions. In a perfectly apportioned system, ‘no citizen’s vote weighs more than anothers’ (Samuels and Snyder 2001, 654, emphasis added). However, citizens are not always the only ones that count. In their overview of reference groups used for apportionment, Handley and Grofman (2008, appendix C) show that 54 per cent of all countries use the total population to apportion seats, whereas the remaining 46 per cent use the number of citizens or voters. The difference between citizens and voters is primarily concerned with age, as young citizens might not be entitled to vote yet. However, in advanced democracies, citizens and voters are almost perfectly correlated because the age structure rarely varies significantly by region (Kashnitsky, De Beer and VanWissen 2021). By contrast, the main difference between citizens and the total population is non-citizen residents.

The difference between total population and citizens in apportionment would not matter much empirically if non-citizen residents were distributed equally across the territory. In this case, the question of who counts in the process of seat apportionment would not have any significant consequences. However, this is not the case. Research shows that in most advanced economies, non-citizen residents are more likely to live in urban areas than rural ones (OECD 2016; Alessandrini, Natale and Tintori 2018). To be clear, there is no necessity for non-citizen residents to concentrate in urban areas, just like there is no necessity for countries to suffer from rural depopulation. However, in most countries, they do (see Figure A1 in the appendix).

The non-random distribution of non-citizens has implications for political representation if seats are apportioned to electoral districts as a function of the total population rather than just citizens or voters. Consider the case of Switzerland, a country that uses the total population to apportion seats to electoral districts and hosts a large number of non-citizen residents (25.7 per cent of the population in 2021). The share of non-citizens varies between 11.5 per cent in the comparatively rural district ‘Appenzell Innerrhoden’ and 40.5 per cent in the comparatively urban district ‘Genève’ (Federal Statistical Office 2022). Citizens residing in Genève benefit from the fact that the large number of non-citizens increases district magnitude, which in turn increases the weight of their vote relative to other Swiss voters. In Genève, there is one legislator in the lower chamber per 25,251 citizens, whereas, for the average Swiss citizen, the corresponding number is 32,469 citizens.

This difference in reference groups matters. Against the benchmark of equal weight of citizens’ votes, population-based apportionment, compared to citizen-based apportionment, increases malapportionment. This matters substantively because differences in shares of non-citizen residents between districts can be large. However, this also matters systematically because non-citizen residents tend to be concentrated in certain, often urban, regions. Therefore, voters residing in areas with large non-citizen populations are the main beneficiaries of population-based apportionment. Moreover, given that there are regional differences in electoral support, population-based apportionment has also important implications for political representation (Rodden 2019; Ford and Jennings 2020; Taylor et al. 2023).

Why did the literature overlook this third factor influencing apportionment? Following the ‘one person, one vote’ logic, scholars often define malapportionment with reference to voters or citizens (Bhavnani 2018; Boone and Wahman 2015; Kamahara, Wada and Kasuya 2021). Yet, in their measurement, most studies resort to the total population, presumably for reasons of data availability (Ardanaz and Scartascini 2013; Beramendi, Rogers and Díaz-Cayeros 2022; Bruhn, Gallego and Onorato 2010; Hiroi 2019; Horiuchi 2004; Horiuchi and Saito 2003). The agenda-setting paper by Samuels and Snyder (2001) is a typical example. They define malapportionment with reference to citizens’ votes but measure malapportionment using the total population. Samuels and Snyder (2001, 655) justify this measurement strategy by writing that they ‘use population per district whenever available [because] most countries apportion seats on the basis of population rather than registered voters’. However, as the overview of Handley and Grofman (2008) shows, this is problematic because roughly half of the countries use the number of citizens or voters to apportion seats.

This paper aims to demonstrate that differences in reference groups matter. In this way, we contribute to the literature that views the design of electoral institutions as the result of political conflict (Bhavnani 2018; Cox and Katz 2002; Walter and Emmenegger 2023). In addition, scholars have started to investigate the distributional consequences of excluding certain groups of the population from the apportionment formula (Baumle and Poston 2004; Chen and Stephanopoulos 2021; Haas, Miller and Kimbrough 2022). However, existing work mainly focuses on the representation of minorities and is limited to the USA. To broaden our understanding of these processes, we present evidence on apportionment from a cross-national perspective in the following.

Data and Methods

We employ cross-country data on the development of apportionment under different institutional designs to show that the share of non-citizen residents at the district level systematically influences political representation when a country relies on population-based apportionment rather than citizen-based apportionment. Previous research has struggled with considerable data limitations, relying mainly on cross-sectional country data without being able to exploit temporal variation and geographically locating which regions enjoy representational advantages (Samuels and Snyder 2001; Ardanaz and Scartascini 2013). Other studies have exploited sub-national variation in single countries. However, these analyses omit important institutional determinants influencing apportionment, in particular the definition of the reference group for apportionment (Ansolabehere, Gerber and Snyder 2002; Bhavnani 2018; Hiroi 2019; Horiuchi and Saito 2003).

We overcome these limitations by employing a large dataset of ten Western countries, which covers the elections from the 1970s until the late 2010s. Table A2 provides more detail on the structure of the data and our sources. Importantly, we focus on administrative structures (henceforth districts) to which seats are distributed as units of analysis. In countries with proportional representation (PR) electoral systems, these units are the same as electoral districts. In countries with majoritarian (MR) electoral systems, however, we focus on levels such as the states in the USA even though electoral districts are drawn on a lower level. In doing so, we avoid conflating representational biases caused by interstate apportionment with biases caused by intrastate districting since we are analytically interested only in the former (Kamahara, Wada and Kasuya 2021). In the USA, for instance, the apportionment rules are enshrined in the constitution, while the process of redrawing electoral districts is in the hands of the states. In our analysis, we will demonstrate that our results are driven by differences in the seat allocation mechanism and not by countries with PR or MR systems (Samuels and Snyder 2001).

As our dependent variable, we use the log-transformed ‘relative representation index’ (RRI) as a district-level measure of representation biases (Ansolabehere, Gerber and Snyder 2002). The RRI is the ratio of voters and district magnitude between the district and national levels and is computed as follows:

$$RRI = \log \left({{{{{\rm{District\ magnitude}}} \over {{\rm{Number\ of\ voters\ in\ district}}}}} \over {{{{\rm{Assembly\ size}}} \over {{\rm{Number\ of\ voters\ in\ country}}}}}}\right)$$

Values above zero indicate that districts have higher seat shares in parliament than shares in the electorate, whereas values below zero indicate that districts have lower seat shares in parliament than shares in the electorate.

To capture the influence of non-citizen residents on the distribution of seats, we use the share of non-citizen residents in a district and the apportionment mechanism as independent variables. We estimate the share of non-citizen residents by subtracting the number of eligible voters from the population and dividing the result by the population. ¹ For the apportionment mechanism, we employ a dummy variable that measures whether seats are distributed according to the number of citizens or voters (‘Reference Citizens’ = 1) or, alternatively, the total population (‘Reference Citizens’ = 0). The data comes from Handley and Grofman (2008, appendix C) and was double-checked using country-specific sources such as constitutions and electoral laws. ² We interact the share of non-citizen residents with our dummy variable for the reference group because our theoretical expectation is that the effect of non-citizen residents operates only in countries with population-based apportionment. Furthermore, districts are entitled to a minimum number of seats in some countries. To account for these minimum representation thresholds, we have created a measure that captures which districts fall below the threshold. Lastly, we expect smaller electoral districts, regardless of minimum representation thresholds, to be politically over-represented to compensate for their small size. Therefore, we add a control variable for the size of the voting population to our specifications.

Given that we have a panel data set with a longer temporal dimension, that is, elections covering multiple decades, we employ a fixed effects model to account for unobserved heterogeneity by year, country, and district, respectively. In addition, the relative representation index (RRI) is strongly autocorrelated over time. We add a one election lag of RRI to all our specifications and cluster the standard errors by districts because the observations within districts are interdependent. Lastly, the number of districts and elections in our data differ considerably between countries. For instance, we have ninety observations for Canada but 1,000 observations for the USA. To avoid that our results are driven by countries with many observations, we add inverse probability weights to our model specifications so that every country has the same impact on the estimates. ³

Results

How do differences in reference groups impact apportionment? Before we turn to our cross-national analysis of the link between non-citizen residents and malapportionment, we illustrate the relevance of reference groups for apportionment in two cases. More specifically, we have computed the distribution of seats to electoral districts for the number of voters and the size of the population in Finland and Switzerland. Both countries rely on PR systems with multi-member districts and allocate seats to districts with the largest remainder method using the population (Switzerland) or citizens (Finland). In contrast to many other cases, both countries do not have adjustment seats on higher levels than the districts or interdependencies in the seat allocation between majoritarian and proportional components as in mixed systems. This makes the calculation of seat distributions straightforward.

Figure 1 displays the share of seats as the fraction of all legislative seats that would be differently allocated under different reference groups. As the plot shows, the distribution of seats would affect between 1.5 per cent and 4 per cent of all legislative seats over time. For instance, 3 per cent of all seats would have been allocated to different districts in 2015 if Switzerland would have switched from population-based to voter-based apportionment. Given that the issue of reference groups has so far attracted little attention, these differences are substantial. To put it into perspective, Samuels and Snyder’s (2001, 660–661) estimates of malapportionment for Finland and Switzerland, based on population data, are 1 per cent and 2 per cent, respectively. Hence, the choice of reference groups can exceed malapportionment resulting from other sources. The reference group thus has a considerable impact on the magnitude of malapportionment.

Figure 1. Impact of Switching from Population-Based to Voter-Based Apportionment on the Regional Distribution of Seats (Per Cent Legislative Seats Affected).

Turning to our quantitative cross-country analysis, we expect that districts with higher shares of non-citizen residents display larger levels of apportionment bias but only in countries with population-based apportionment. Given that the effect of the share of non-citizen residents is conditional on the apportionment mechanism, we interact the share of non-citizens with our dummy variable for different reference groups in Table 1. All models include a lagged dependent variable as well as country and year fixed effects.

Table 1. The Effect of Non-Citizen Residents and the Apportionment Mechanism on the Relative Representation Index (RRI)

Cluster-Robust Standard Errors by District. $***p \lt 0.001$ ; $**p \lt 0.01$ ; $*p \lt 0.05$

Our preferred specification with inverse probability weights in model 1 shows that the interaction term is significant at the 1 per cent level. In addition, the coefficient for the share of non-citizen residents has a positive sign and is significant at the 0.1 per cent level. With the interaction term included in the specification, non-citizen residents display the effect of the baseline category for the binary variable ‘reference citizens’, namely population-based reapportionment. Put differently, the coefficient shows that non-citizen residents increase malapportionment in countries with population-based reapportionment. The negative sign of the coefficient of our interaction term then suggests that the effect of non-citizen residents on the relative representation index is weaker, neutralized, or even negative in countries with citizen-based apportionment. We return to the interpretation of the estimates for both apportionment mechanisms below.

Model 2 shows that our results are similar to model 1 if we do not include weights in our specification. Moreover, in model 3, we restrict our sample to countries with PR to exclude countries where apportionment and districting involve two different steps with different actors involved. The estimates show that the results remain largely unchanged. In Table A5, we used the same specifications as in models 1 to 3, except that we employ district fixed effects as well. As before, the estimates of our interaction term and the coefficient of non-citizen residents remain significant with the expected signs.

To provide a substantial interpretation of the interaction term, we present a marginal effects plot in Figure 2 using the estimates of model 1 in Table 1. The figure shows the effect of the share of non-citizen residents on the relative representation index under different seat allocation mechanisms. The figure demonstrates that the share of non-citizen residents does not have an effect on malapportionment in countries with citizen-based apportionment, which is distinguishable from zero. By contrast, non-citizen residents have a positive effect on malapportionment in countries with population-based apportionment. A 10 percentage point increase in the non-citizen population is associated with an increase in the relative representation index of about 1.7 per cent. In short, the reference group matters.

Figure 2. Marginal Effect of Non-Citizen Residents on the Relative Representation Index.

Conclusion

This paper has highlighted a previously overlooked factor influencing the apportionment process. According to Samuels and Snyder (2001, 654), in malapportioned systems, ‘the votes of some citizens weigh more than the votes of other citizens’. However, numerous countries do not consider the number of citizens or voters when apportioning seats to districts. Rather, they apportion seats based on the total population. This has implications for representation because non-citizen residents are not distributed equally across a territory. In the empirical analysis, we have demonstrated that population-based apportionment has a systematic and substantial effect on malapportionment.

Such malapportionment has important political consequences. Non-citizens tend to be concentrated in urban districts, which is why urban regions are likely to benefit from population-based apportionment. Similarly, the political left may benefit from population-based apportionment because it tends to be comparatively strong in urban areas. The correlations between the share of non-citizen residents and the district’s position on the left-right scale displayed in Figure A2 in the appendix suggest that, in several countries, the political left benefits from population-based apportionment. However, the picture is far from univocal, as, for example, the Nordic countries demonstrate. Clearly, the choice of apportionment mechanism has distributional implications but which parties benefit depends on the countries’ electoral geographies.

The choice of reference group in the apportionment process also raises important normative questions. Ultimately, all answers to the question of whether non-citizen residents should count in apportionment are controversial (Mansbridge 2003). For example, there is a strong current in political theory that argues in favour of the political representation of non-citizen residents (Angell and Huseby 2017; Arrighi and Bauböck 2017; Pedroza 2019). Importantly, however, there is an important difference between the enfranchisement of non-citizens and their consideration in the apportionment process because the political preferences of non-citizens and citizens residing in the same districts do not necessarily converge, which is why scholars often stress the importance of descriptive representation (Pitkin 1967; Mansbridge 2003; Campbell, Childs and Lovenduski 2010). It is thus not necessarily the case that population-based apportionment improves the political representation of non-citizen residents unless it is also paired with voting rights.