Looking beyond the district: The representation of geographical sub-constituencies across Europe

Abstract

A key question in representation is how institutional settings bring about particular representational roles among legislators. In this regard, the strategic dilemma that representatives face of whether to represent all people in the district equally or, alternatively, to prioritize some area within the district, has been vastly understudied. Using innovative survey data collected in 12 European democracies, we demonstrate that a striking number of legislators favour representing the interests of their home town over the district as a whole and that the number of representatives elected by the district critically impacts their choice as to whom to represent. As district magnitude increases, an increasing number of legislators will not cater to district opinion but will prioritize the interests of a geographical sub-constituency. These findings have important implications for the study of political representation, challenging the conventional wisdom that – compared to single-seat districts – proportional representation tends not to provide geographical representation.

Keywords

Political representation geographical sub-constituencies district magnitude cross-national legislators’ survey focus of representation

Introduction

What do legislators see, Fenno (1978) famously asked, when looking at their district? Will they seek to represent all people in the district or will they concentrate instead on some geographical subpart thereof? The answer lies in the way the lines on the map define constituencies and represent people by where they live (Rehfeld, 2005). The simple fact that constituencies are defined in geographical terms – especially in the context of single-member plurality – creates incentives for legislators seeking re-election to think of political representation in terms of district opinion. However, conventional wisdom dictates that proportional representation (PR) does not provide geographical representation to the same extent (Latner and McGann, 2005). As Cox (1997: 228) argues, the number of representatives elected by a district – that is, district magnitude – has the greatest impact on a legislator’s strategic choice as to whom to represent. The district focus of representation is of minor importance, as one leading comparative study noted (Weßels, 1999), beyond single-member districts. If the legislator is the district’s sole representative, he or she is sure to be held to account for not being responsive; whereas multi-member districts are assumed to be too heterogeneous to constitute a ‘community of interest’ capable of being represented and accountability is obfuscated, which often leads to shirking on the part of legislators.

However, the failure to look beyond the district, we argue, has caused scholars to underestimate the extent of geographical representation in PR systems. Rather than spreading their resources thinly, and, therefore, ineffectively, across large multi-member districts, legislators will not seek to represent all people in the district but will prioritize only part of the district (Crisp and Desposato, 2004). This notion of a sub-constituency builds on the insights of Fenno (1978) that congressmen tend to see different circles of support within the district – only the widest of which is the geographical district. Whereas some legislators think of the other circles of support in partisan, religious, or ethnic terms, many are aware they disproportionally draw support from particular areas in the district (see also Jewell, 1982). A large number of studies have confirmed that electoral support is higher in a legislator’s home town and that distance exacts a toll in areas further afield where people are unfamiliar with the legislator (for a recent account, see Gimpel et al., 2008). The import of geography in this regard has long been recognized by parties fielding candidates from all parts of the district in PR (see Gallagher and Marsh, 1988). But this notion of a geographical sub-constituency has yet to inform the comparative study of legislators’ representational roles.

This article contributes to the literature examining the degree to which institutional settings shape political representation by arguing that, while they do not cater to the district opinion, a sizeable number of legislators in multi-member districts prioritize the interests of a geographical sub-constituency – more specifically, the people residing in their home town – over the majority of the district. The implication is that research so far may have underestimated the extent of geographical representation in PR systems. As such, we provide the first – as far as we know – comprehensive, cross-national test of geographical sub-constituency politics. In particular, we present new data on legislators’ sub-constituency focus of representation in 12 statewide – and over 50 regional – legislatures across Europe. Moreover, these legislators are elected using the widest variety of electoral institutions, allowing us to test for their impact. Additionally, the paper makes an important methodological contribution: survey instruments will need to be revised in order to include the representation of areas smaller than the district lest we misunderstand the focus and extent of geographical representation. The article should be of interest to scholars of political representation, democratic institutions, legislative behaviour, political geography, and comparative politics.

Whom to represent: a geographical sub-constituency focus of representation?

Different legislators have different focuses of representation. Confronted with the many people living in their district and the many conflicting demands made on them, legislators have to decide where to put the emphasis. Facing multiple and often competing principals, they have to decide to whose views and interests they will be most responsive. Single-member districts, the conventional wisdom argues, should result in a district focus of representation. Multi-member districts, by contrast, do not provide geographical representation of this kind and legislators elected by multi-member districts can be expected to favour partisan or otherwise organized interests over unorganized constituents (see Bawn and Thies, 2003).

Representing the views and interests of all people in the district, the careful comparison of single-country studies suggests, is more important in countries using single-member districts than in countries using multi-member districts (see Bogdanor, 1985; Esaiasson, 2000; Jewell, 1982; Müller and Saalfeld, 1997). Mixed-member systems, moreover, observe similar differences in focus between legislators elected by the nominal and the proportional tier (Chiru and Enyedi, 2015; Lundberg, 2007). The 1996 European Representation Study further attests that ‘the smaller the district magnitude, the narrower the representational focus representatives will choose’. As district magnitude increases, by contrast, a legislator’s probability of having a district focus decreases and geographical representation is stunted (Weßels, 1999: 221).

However, these previous studies have overlooked the strategic dilemma that legislators face either to represent all people in the district or, alternatively, to privilege some part thereof. They face the conundrum of either appealing to all unorganized citizens in the district equally or, alternatively, to prioritize a subsection. This notion of sub-constituency politics highlights the trade-off involved between seeking narrow, intense support rather than wide – albeit shallow – support. Constituents feel strongly about very few issues. But because on many issues there is a minority that does care – legislators will frequently represent the minority at the majority’s expense (Bishin, 2009).

While this notion of sub-constituencies is about issue publics who share a social identity, concerns and preferences, it can be easily extended to pertain to subgroups that are geographically concentrated. Some scholars have indicated that districts are often too large to constitute ‘communities of interest’ (Rehfeld, 2005), throwing in densely populated urban cores together with small towns and sparsely populated rural areas. Urban and rural areas are frequently at odds over economic development, poverty, mobility, the environment, and a plethora of other issues (e.g. Bishop, 2008). Facing uncertainty about district opinion as a result, at least some legislators can be expected to prioritize some part over the district as a whole. Geographical sub-constituencies in particular facilitate communication among constituents and can be more easily mobilized, aiding legislators’ ability to get their message across (see Bishin, 2009).

Building on the sub-constituency politics argument, we can hypothesize that the larger the multi-member district, the more legislators will grow uncertain about district opinion (Weßels, 1999) and the greater will be their propensity to favour some part over the district as a whole and to focus on such a geographical sub-constituency. That is, the larger the district magnitude, the more legislators are expected to turn to ‘narrower’ focuses of representation, including – but not exclusively – geographical sub-constituencies below the district level, in order to reduce uncertainty about whom to represent.¹

Moreover, the district magnitude’s expected positive effect on geographical sub-constituency representation is also consistent with Myerson’s (1993) model of candidate strategies under different electoral systems. As district magnitude increases and electoral systems become more proportional, legislators have a strong incentive to promise more goods to favoured minorities. District magnitude decreases the thresholds for winning a seat, allowing legislators to win re-election with smaller proportions of the vote (see also André et al., 2014b; Grofman, 2005). The argument does not make the distinction between preferential and non-preferential systems, which is contrary to the position advocated by Carey and Shugart (1995). In both systems, the inter-party allocation of seats proceeds along the same lines and, as such, thresholds decrease as a function of district magnitude in preferential and non-preferential systems. Where the threshold for winning a seat is lower, the legislator who appeals to the entire multi-member district is expected to be vulnerable to challengers who concentrate their efforts on a narrow sub-constituency and are, therefore, able to outbid the incumbent in this area. The importance of a legislator’s local roots, for instance, has been related to the threshold or the votes needed to win re-election; large vote shares simply cannot be obtained when concentrating on the local community alone (André et al., 2014a).

Key hypothesis: As district magnitude increases, legislators are more likely to shift their focus of representation away from the district towards a geographical sub-constituency.

However, in line with Carey and Shugart’s (1995) seminal argument, one could expect a legislator’s propensity to shift towards a sub-constituency focus to be stronger in preferential than in non-preferential systems. In preferential systems, the order in which the candidates running under the same party label are elected is governed by the preference votes they poll. As such, any candidate may benefit from attracting additional votes among ‘friends and neighbours’, hoping in this manner to leapfrog past a co-partisan ranked higher on the party list and to ward off lower-ranked challengers. In these systems, typically, far fewer votes will gain the candidate a seat from, or lose his or her seat to, a co-partisan than are needed to win the party an extra seat (see André et al., 2014a; Katz, 1986). In non-preferential systems, by contrast, the order of election is pre-fixed and only the candidates in the marginal positions on the list (most notably, those in the last-winner and the first-loser positions) may benefit from putting a face to the party in a narrow sub-constituency. Thus, while on average the sub-constituency focus can be expected to be more common in preferential systems, the expectation is that it grows more important as district magnitude increases in preferential and in non-preferential systems alike.

While the area that a legislator privileges representing likely varies from legislator to legislator (see Fenno, 1978), there is one place they typically have in common: the legislator’s place of residence. The place of residence is, first, an important factor affecting his or her recruitment (Gimpel et al., 2011). Party lists in PR systems typically balance candidates from different localities in the district (Gallagher and Marsh, 1988; for a striking example, see Carty, 1983). Even in the unlikely context of the nationwide constituency in Israel and the Netherlands, political parties have been found to value geographically balanced lists (Latner and McGann, 2005). Legislators frequently started in the local politics of their home town (Shugart et al., 2005) and as many as 18 per cent in Spain and 24 per cent in Germany continue to hold local office in their home town once elected (Navarro, 2013). Second, knowing where a legislator is from constitutes a valuable information shortcut to voters, signalling shared interests and inspiring trust (Shugart et al., 2005). In presenting themselves to voters, candidates rarely fail to divulge where they live and the community they are rooted in. It constitutes an important element of voter evaluation (Campbell and Cowley, 2014; Gimpel et al., 2008; Górecki and Marsh, 2014).

Across wide-ranging institutional contexts, legislators’ support has, therefore, been found to spike in their place of residence, their bailiwicks boosted by friends-and-neighbours voting (Gimpel et al., 2008). In Ireland, in particular, electoral bailiwicks have a long pedigree – whether observed from the unofficial vote tallies recorded by party representatives present at the count (Parker, 1982; Sacks, 1970) or, more recently, using survey data and information on the geographic coordinates of candidates and voters (Górecki and Marsh, 2012). There are also clear indications that Irish political parties go so far as to divide up the district, delimiting where a candidate will canvass most heavily and where the local party will instruct voters to give him or her their first preference (Carty, 1983; Górecki and Marsh, 2012; Sinnott, 1995). But in other institutional contexts, this home-town advantage also constitutes a personal vote – though by no means the only one (Grofman, 2005; Nemoto and Shugart, 2013) – that legislators have been known to nurture by frequent trips home (Crisp and Desposato, 2004), diligent constituency service (André and Depauw, 2013), the establishment of local offices (Lundberg, 2007), and pork expenditure (Ames, 1995). As such, if there is some merit to our argument, the representation of a legislator’s place of residence constitutes a natural starting point to look for it.

Data

Data on legislators’ focus of representation have been collected as part of the PARTIREP cross-national study of representation. The project, hosted by the Vrije Universiteit Brussel, has surveyed the members of 12 national legislatures. Included in the study are Austria, Belgium, Germany, Hungary, Ireland, Italy, Norway, Poland, Portugal, Spain, Switzerland, and the UK.² In addition to the national legislators, regional legislators were surveyed in the seven multi-level democracies. In Austria, Belgium, Switzerland, and the UK, members of all regional legislatures have been surveyed; the selection of regional legislatures that were surveyed in Germany, Italy, and Spain is reported in Table 1. The selection balances areas from the east and the west in Germany, from the north and the south in Italy, and autonomous regions in Spain with stronger and weaker regionalist traditions.

Table 1.

The case selection of legislatures and variation in electoral context.

	Regional legislatures included in the selection	Response rate		District magnitude			Preferential system
	Regional legislatures included in the selection	Responses	Proportion	Range	Mean	SD	Preferential system
Austria	All nine regional legislatures	216	0.36	2–36	11.7	8.1	0–1
Belgium	Flanders and Wallonia	125	0.36	3–33	16.9	8.0	1
Germany	Brandenburg, Lower Saxony, Rhineland-Palatinate, and Thuringia	279	0.27	1–65	23.2	25.0	0
Hungary		83	0.26	1–28	4.8	6.5	0
Ireland		34	0.22	3–5	3.9	0.8	1
Italy	Calabria, Campania, Lazio, Lombardy, and Tuscany	115	0.13	1–43	23.3	14.4	0–1
Norway		46	0.27	4–17	10.1	4.2	0
Poland		55	0.12	7–19	11.6	2.4	1
Portugal	The Azores	107	0.26	2–47	20.4	16.1	0
Spain	Andalusia, Catalonia, Basque Country, and Valencia	272	0.35	3–85	25.9	23.3	0
Switzerland	23 cantonal legislatures	577	0.22	1–36	14.8	9.4	1
UK	Scotland and Wales	107	0.13	1–7	2.0	2.1	0

Note: the table displays the response rates calculated as the ratio between the number of respondents and the size of the population. It further reports the distribution of district magnitude and the coding of ballot structure by country.

SD: standard deviation.

The survey was conducted between spring 2009 and winter 2012, using a combination of web-based and print questionnaires with a closed-ended question format. To guarantee cross-national comparability the survey was kept rigorously constant in content when translated into 14 languages. Members of the legislatures in the selection received a personalized introduction letter and email presenting the project and inviting them to participate. They were further sent at least two online reminders (excluding ‘hard’ refusals) and the option was offered to them to fill out a print questionnaire. A final invitation was by telephone. In Hungary, face-to-face interviews were used. The fieldwork was done at different moments in the electoral cycle in the different countries, meticulously avoiding election campaigns periods. Further details on the survey are described in Deschouwer et al. (2014).

Response rates per country, depicted in Table 1, range from 13 to 36 per cent, averaging about one in four. In spite of the variation in the response rates, deviations from the population of legislators surveyed are within acceptable margins. The Duncan dissimilarity index amounts to 8 per cent with regard to the national/regional level of government, 3 per cent with regard to ruling and opposition parties, and 4 per cent regarding the sexes. Deviations to the party distribution in the legislature do not exceed 15 per cent, except in Italy and Spain – where deviations among the main parties nonetheless amount to no more than 10 per cent (see Deschouwer et al., 2014: 10–11). Certainly, there is no record of non-response or drop-out being associated with the focus of representation questions included in the questionnaire. Any cultural social desirability bias in responding should be constant within each country and would, therefore, be absorbed by the country-fixed effects in the analysis. Post-stratification weights are used, moreover, to compensate for minor deviations in the sample with respect to the party distribution in the legislature. In addition, the members of the Swiss regional legislatures constitute a particularly large group in the selection – due to the number of cantonal legislatures in the country – and are underweighted to ensure that no one country dominates the analysis.³

The legislatures that were surveyed allow for the widest variety of electoral institutions, the key explanatory variable. Included are majoritarian and proportional systems – as well as mixed-member systems that combine plurality and PR. The distinction is further made between preferential systems that enable voters to indicate a preference among co-partisans running under the same party label and non-preferential systems that do not. Moreover, electoral institutions are frequently not uniform within a single country. Central to our purpose, the selection includes legislators elected by districts of various magnitudes (i.e. the number of seats that are allocated in the district). District magnitude, Table 1 attests, ranges from the single-member districts in Germany, Hungary, and the UK to the 85 members that the Barcelona district elects to the Parliament of Catalonia, providing sufficient analytical leverage to address the question at hand. Not only is there ample variation in district magnitude within and across countries, there is no near separation in the data between the countries. Low or high district magnitudes, moreover, are not confined to either preferential or non-preferential systems. In addition, low district magnitudes are not restricted to regional legislatures, nor are high magnitudes restricted to national legislatures.

The dependent variable

A legislator’s focus of representation, the dependent variable, is – as in previous studies (see Weßels, 1999) – captured by the question ‘how important is it to you, personally, to promote the views and interests of all the people in your constituency?’ The importance the legislator attributes to representing the district is tapped into using a seven-point scale, ranging from ‘no’ to ‘great importance’. Contrary to other cross-national elite surveys, however, the PARTIREP survey also looks into the possibility of geographically defined sub-constituencies by inquiring ‘how important do you, personally, find it to promote the collective interests of your municipality?’, using an identical seven-point scale. The ensuing trade-off between prioritizing a legislator’s district or home town is then computed by subtracting the importance the legislator attributes to representing the district from the importance of representing his or her municipality. The trade-off ranges from −6 to 6. Positive values reflect the prioritization of a sub-constituency focus, whereas negative values indicate a district focus. Social desirability is unlikely to bias responses, as full anonymity was guaranteed to all respondents. Nor is the issue very politically sensitive or partisan (see Searing, 1994).

The key trade-off is illustrated by Figure 1. It highlights the critical variation in legislators’ thoughts about representation, the dependent variable being normally distributed and picking up on every possible nuance in the legislator’s representational role. Facing the strategic conundrum of representing all people in their district or privileging their home town, at least as many legislators favour the interests of their home town as there are legislators who adopt a district focus. The middle of the scale is, however, especially well-populated. Facing the strategic conundrum, it seems, quite a few legislators are ambivalent about the two, thinking it very important to represent the interests of their district and home town. There is also substantial cross-national variation in legislators’ thoughts about representation. The distributions per country can be found in Appendices 1 and 2 of the online supplemental material.

Figure 1.

The distribution of the district–sub-constituency trade-off.

Trade-offs of this kind present an additional puzzle. To some legislators the strategic conundrum is irrelevant, because they find representing both the district and their home town unimportant. Even though they do not think either adequately describes what they do, they can be found from minus four to plus four on the scale, adding noise to the measurement of legislators’ focus of representation. Recoding the trade-off into four categories allows us to shed more light on the issue. As Table 2 indicates, only about 17 per cent of the legislators find the district–home town trade-off to be largely irrelevant: they attribute both values under 5 on the seven-point scale (see Weßels, 1999). Since the strategic choice is inconsequential to them, it could be argued that it would be better to exclude them from the analysis. By this more conservative standard, about 22 per cent of the legislators prioritize representing the district, attributing greater values to representing the district than to promoting the views of their home town (that is, scoring from −6 to −1 on the metric trade-off and having valued at least one of the options over 5 on the seven-point scale). By contrast, almost 26 per cent give precedence to their home town (reflected by scores of 1 to 6 on the metric trade-off and having valued at least one of the options over 5 on the seven-point scale). About 35 per cent of the legislators are ambivalent about the two, attributing both options equal values over 5 on the seven-point scale. As such, we find strong support for the contention that geographical sub-constituency representation is important and that a sizeable number of – national and regional – legislators across Europe engage in the representation of a geographically defined sub-constituency unobserved by previous studies.

Table 2.

Variation in the trade-off between district and sub-constituency focus by district magnitude.

	Range of district magnitude	District focus	Both equally important	Sub-constituency focus	None very important
	Range of district magnitude	[–6→–1]	[0]	[1→6]	[–4→4]
1st quartile	1–5	28.9	38.4	20.1	12.6
2nd quartile	6–12	21.9	40.0	23.6	14.5
3rd quartile	13–25	16.8	31.3	29.0	23.0
4th quartile	26–85	17.1	30.4	32.6	19.9
Total		21.7	35.4	25.8	17.0

Note: the table shows the distribution of the dependent variable, the trade-off between a district focus and a sub-constituency focus. Entries are the percentages of legislators by country prioritizing the district (negative values); prioritizing their home town (positive values); considering both foci equally important (a value of zero); or valuing neither the district, nor their home town (values ranging from −4 to 4).

Table 2 further presents a first glance of how the district–sub-constituency trade-off varies across district magnitude. The percentage of legislators prioritizing a sub-constituency focus increases, Table 2 indicates, from the first to the second quartile of district magnitude, from the second to the third, and from the third to the fourth. Correspondingly, the percentage of legislators favouring the district focus decreases from the first to the second quartile, from the second to the third, and remains at a low level in the fourth quartile. While, in general, the evidence is supportive of our key hypothesis, caution is advised interpreting raw percentages, as they do not account for alternative system-level or individual-level explanations. To isolate the effect of district magnitude more rigorous testing is required.

Method

The trade-off between favouring the district or, alternatively, the sub-constituency, is treated as metric and best estimated using ordinary least squares (OLS) regression. There are two electoral system variables of main interest: one is district magnitude.⁴ In line with previous research, the decimal logarithm of district magnitude is taken to account for the intuition that changes in magnitude in the low range will be more influential than changes in the high range. The log transformation also reduces variation caused by extreme values. The other electoral system variable is ballot structure, coded ‘1’ for preferential systems that enable voters to indicate a preference among co-partisans running under the same party label and ‘0’ for non-preferential systems that do not offer voters the option. Moreover, alternative system-level determinants are controlled for by clustering standard errors at the parliament level. As such, we correct for legislators, working within the same legislative arena, sharing similar conceptions on representation.

To isolate the effect of electoral institutions, the analysis further controls for a number of individual-level factors that have commonly been associated with variation in legislators’ focus of representation. First, we take into account gender differences in how legislators conceive their representational role. Reflecting differences in life experiences, female legislators are often found to be more district-oriented than their male colleagues (Richardson and Freeman, 1995). Second, we control for a legislator’s seniority in terms of the number of years he or she has held a seat in Parliament. Over the course of his or her career, a legislator moves from the expansionist to the protectionist stage and comes to focus on a more narrow base of hard-core supporters (Fenno, 1978). Third, the Speaker of Parliament, Parliamentary Party Group leaders, and committee chairs shoulder more leadership responsibilities in Parliament compared to rank-and-file members, possibly drawing their attention away from district work towards their tasks in the legislative arena (Heitshusen et al., 2005). Fourth, differences in the focus of representation might be a function of the policy preferences of the legislator in relation to his or her party. Legislators who place themselves more distant from the party on an 11-point left–right scale may be able to retain the district’s trust by a stalwart district orientation (Fenno, 1978). Besides the policy distance, the party’s left–right position enters the regression equation separately to account for possible floor and ceiling effects. In addition, the natural constituencies of leftist parties likely place higher service demands upon their representatives (Cain et al., 1987).

We further have to account for two contextual factors that may shape legislators’ focus of representation: the geographical location of the district measured in kilometres and the level of government.⁵ First, peripheral districts that are far removed from the seat of Parliament are likely to harbour distinct identities that bolster district orientations among the legislators elected by them. People living in remote districts resent being ruled from the centre and can be expected to put more pressure on their representatives to defend the particular interests of their area (Thomassen and Esaiasson, 2006). Second, Patzelt (2007) suggests that, because of the jurisdictions dealt with at the regional level, state legislators in Germany interact more often with local authorities than federal legislators, and are contacted more frequently by constituents and interest groups on local politics. Should this pattern translate to other countries, this might mean that regional legislators, in turn, may be more focused on their municipality than national legislators. Full details on the operationalizations and summary statistics of these alternative explanations can be found in Table 3.

Table 3.

Operationalization and summary statistics of the control variables.

	Operationalization	N	Mean	SD	Min.	Max.
Female	Dichotomous indicator coded ‘1’ for women and ‘0’ for men	1679	32.97%		Dummy
Seniority	Number of years a member has served in parliament	1679	5.06	6.53	0	54
Leadership position	Dichotomous indicator coded ‘1’ for the Speaker of Parliament, Parliamentary Party Group leaders, and committee chairs and ‘0’ for backbenchers holding no leadership position	1679	14.51%		Dummy
Policy distance from the party	The distance between legislators’ self-placement on an 11-point left–right scale and the score they assigned their party in absolute terms	1679	0.85	0.97	0	8
Party’s left–right position	The mean position of the party legislators’ self-placement on an 11-point left–right scale	1679	4.9	1.88	0	10
Geographical distance	The decimal logarithm of the distance in kilometre as the crow flies between the largest town or city in the district and the seat of Parliament using Google Maps. In mixed-member systems, the distance was measured to the first tier the legislator stood in. Only for those standing exclusively on the party list was the distance set to zero	1679	1.67	0.87	0	3.24
Regional parliament	Dichotomous indicator coded ‘1’ for members of regional parliaments and ‘0’ for members of national parliaments	1679	49.00%		Dummy

SD: standard deviation; Min.: minimum; Max.: maximum.

The regression coefficients, robust standard errors, and statistics of model fit are reported in Table 4. Model 1 reports the bivariate effect of district magnitude on a legislator’s dilemma between district and sub-constituency representation. Model 2 retests the effect of district magnitude after controlling for common alternative explanations of a legislator’s representational role. In Model 3, country-fixed effects are added, further isolating the effect of the electoral system from other cross-national differences in representation. To test whether the effect of district magnitude is dependent on the ballot structure, we add the multiplicative interaction term (Interaction model).

Table 4.

District magnitude and the district–sub-constituency trade-off.

	Model 1		Model 2		Model 3		Interaction model
	B	SE	B	SE	B	SE	B	SE
District magnitude	0.347	(0.130)**	0.316	(0.133)*	0.471	(0.089)***	0.460	(0.097)***
Preferential system			0.389	(0.108)***	0.675	(0.208)**	0.575	(0.286)*
District magnitude / Preferential system							0.099	(0.273)
Regional parliament			0.045	(0.120)	0.211	(0.070)**	0.212	(0.071)**
Geographical distance			0.005	(0.070)	0.027	(0.061)	0.029	(0.061)
Policy distance from the party			−0.077	(0.045)^#	−0.088	(0.045)^#	−0.088	(0.045)^#
Party’s left–right position			−0.048	(0.021)*	−0.051	(0.021)*	−0.051	(0.021)*
Female			0.009	(0.088)	0.042	(0.085)	0.041	(0.085)
Seniority in years			0.008	(0.005)	0.008	(0.006)	0.008	(0.006)
Leadership position			−0.256	(0.104)*	−0.232	(0.101)*	−0.232	(0.101)*
Country dummies	Omitted		Omitted		Included		Included
Constant	−0.281	(0.151)^#	−0.124	(0.317)	−0.790	(0.234)**	−0.767	(0.233)**
N	1679		1679		1679		1679
Log-likelihood	−2961.8		−2943.5		−2912.3		−2912.2
LR (df)	30.0	(1)***	66.5	(9)***	128.8	(20)***	129.0	(21)***

Note: entries are the parameter estimates and robust standard errors (in parentheses) of ordinary least squares (OLS) regression models as well as the model’s likelihood ratio test. Standard errors are clustered by Parliament. The dependent variable is the trade-off between adopting a district focus (low values) or a sub-constituency focus (high values). The regression estimates are not affected by multicollinearity: the mean Variance Inflation Factor (VIF) of Model 3 is as low as 1.98.

p ⩽ 0.1.

p ⩽ 0.05.

p ⩽ 0.01.

***

p ⩽ 0.001 using two-tailed t-values.

B: regression coefficients; SE: standard error; LR: likelihood ratio.

To make sure our findings do not result from some legislators’ inability, or refusal, to make a distinct choice, we re-examine the evidence in the online supplementary material, excluding in a first subsample the 347 legislators who find neither representing the district, nor the sub-constituency, important. In a further restricted subsample, we also exclude the 545 legislators who find representing both equally important (Appendix 3). To further ascertain the robustness of our findings, we explore different model specifications. In order to demonstrate that our findings are not driven by outliers who attach very high values of importance to one end of the spectrum and very low values to the other, we use multinomial logistic regression in Appendix 4, in which the dependent variable is recoded in three categories. Appendix 5 uses multilevel estimation techniques to account for the hierarchy in the data in a more stringent way. In addition, we present three additional subsample analyses. First, we test whether our inferences are driven by the peculiarity of the single-member districts by excluding them from the analysis (Appendix 6). Second, we run split-sample analyses on the national and regional legislatures (Appendix 7). Third, a series of models are run to ensure no one country dominates the analysis, excluding each of the 12 countries in turn from the analysis (Appendix 8).

Results: district magnitude and the trade-off between district and sub-constituency

Our findings demonstrate that electoral institutions shape legislators’ decision as to whom to represent. Table 4 firmly supports our key theoretical expectation, establishing that in low-magnitude districts legislators favour representing all people in the district but that in high-magnitude districts they are more likely to concentrate on representing their home town. Both the bivariate (Model 1) and multivariate (Model 2) model specifications report that district magnitude has a positive and statistically significant effect on the trade-off between district and sub-constituency. Recall that positive values on the trade-off reflect an emphasis on the sub-constituency compared to the district, while negative values are associated with a district rather than a sub-constituency focus of representation. As district magnitude increases, legislators become less likely to find it important to represent all people in the district equally; whereas a growing number of them will concentrate on representing the interests of their home town instead. Moreover, including individual-level controls hardly detracts from the explanatory power of district magnitude (see Model 2). As Model 3 attests, adding country-fixed effects even increases the magnitude of the effect, providing further evidence of the robustness of the finding.

Figure 2 depicts the effect size of district magnitude (on the basis of Model 3) to ensure correct interpretation: the average legislator in a single-member district has an expected value of −0.40, that is, towards the district orientation end of the spectrum. Increasing the district magnitude to its observed maximum moves the average legislator more than one point towards the sub-constituency end of the spectrum to a value of 0.51. Thus, from a district magnitude of eight, Figure 2 demonstrates that legislators predominantly find it more important to represent their home town than to represent all people in the district. Even though the effect of district magnitude is the strongest found, taken together the control variables and country dummies account for a larger proportion of the explained variance. These results suggest that the number of seats to be allocated in the district does have an appreciable impact on legislators’ focus of representation, but that the size of the effect should not be overstated.

Figure 2.

The simulated effect of district magnitude on legislators’ focus of representation.

In line with Carey and Shugart’s (1995) seminal argument, we find that the geographical sub-constituency orientation is more common in systems where personal vote-seeking on the part of legislators may affect their re-election. The coefficient of ‘preferential systems’ is positive and statistically significant in both Models 2 and 3 in Table 4. As expected, the effect of district magnitude is not conditional upon the presence or absence of preferential voting. Adding a multiplicative interaction term to the regression equation, the slope of district magnitude is invariably positive in both open- and closed-list systems. The interaction term cannot meaningfully be distinguished from zero. Figure 3 depicts the conditional effect for ease of interpretation. While a legislator’s sub-constituency focus varies with district magnitude in a similar fashion in preferential and non-preferential systems, the regression line of non-preferential systems systematically lies below that of preferential systems. The confidence intervals do not overlap, moreover, with the exception of the highest-magnitude districts (where only 2 per cent of the observations can be found).⁶ We can, therefore, be confident that the effects of district magnitude and ballot structure on the district–sub-constituency trade-off are simply additive.

Figure 3.

The conditional effect of ballot structure on legislators’ focus of representation.

Further robustness checks reported in the online supplemental material all provide additional credence to our main conclusion. The observed effect of district magnitude is not driven by the 347 legislators to whom the trade-off is irrelevant, nor by the 545 legislators who attach equal importance to their district and home town. Multinomial regressions and two-level random intercept models both emphasize that legislators in high-magnitude districts tend to engage in geographical representation but focus on a sub-constituency centred around their home town rather than on the district as a whole. Additionally, the results are not sensitive to the inclusion in the data of single-seat districts, regional legislatures, or particular countries.

In sum, the study presents new insights into the impact of the electoral context on legislators’ decision as to whom to represent, even if the presented models cannot fully explain variation in the district–sub-constituency trade-off. After taking into account differences in district magnitude, ballot structure, and all control variables, four country dummies continue to have a substantial impact. Relative to the country closest to the overall mean – that is, Ireland – the sub-constituency focus is of above-average importance in Hungary, Norway, and Poland and below-average importance in Switzerland – but recall that none of these countries are driving the results (see Appendix 8). Whereas, ultimately, we cannot rule out the possibility that our findings pertain only to the European countries where they were observed, studies in Brazil (Ames, 1995) and Colombia (Crisp and Desposato, 2004) certainly suggest that they may well travel to other parts of the world.

Conclusion

Districts are ‘legally bounded spaces’ (Fenno, 1978: 1) whose boundaries identify the people who are eligible to vote for a legislator. In turn, the legislator has an electoral incentive to be responsive to district opinion, relating interest to place. In this sense, how districts are defined can be expected to have an influence on legislators’ opinions as to whom to represent. More specifically, legislators face the strategic dilemma whether to represent all people in the district equally or, alternatively, to privilege only part of the district – typically, the area centring on their home town. This study provides evidence that a striking number of legislators favour representing the interests of their home town over the district as a whole and that the number of representatives elected by the district critically impacts their focus of representation, thereby contributing to the broader research agenda as to whether institutional settings shape democratic representation.

Using innovative survey data collected in 12 advanced industrial democracies, we found that, as district magnitude increases, legislators are less likely to represent all people in the district equally and to grow increasingly more likely to privilege their home town – providing the strongest support for our key hypothesis. The larger the district magnitude, it can be surmised, the more uncertain legislators are about district opinion and the more viable electoral strategies focusing on a narrow geographical sub-constituency become in both preferential and non-preferential systems. Different operationalizations and alternative modelling strategies – even when restricting the analysis to various subsamples – only confirm the robustness of our conclusions.

These findings have important implications for the study of political representation. In spite of the long pedigree of research regarding the focus and style of representation, our findings point to the importance of geographical sub-constituencies below the district that studies so far have not identified. At the least, our findings call for a revision of the standard survey question wording, including different options for geographical sub-constituency representation. Otherwise, this study strongly suggests, we will continue to underestimate the extent of geographical representation – especially in high-magnitude PR systems.

Even the growing personal vote literature has rarely examined whether legislators spread their efforts to cultivate personal votes around the district or specialize locally. These studies have focused on a wide range of activities and strategies measured at the district level, including trips to and time spent in the district, central government funds obtained for the district, and casework assisting constituents. While some have suggested that the particular activities a legislator decides on in order to nurture a personal reputation depends on the conditions under which he or she competes for re-election (André and Depauw, 2013), the important question as to where within the district’s boundaries they engage in personal vote-seeking has thus far received little attention. Yet, our finding that the likelihood that a legislator focuses on a geographical subsection of the district is dependent on district magnitude has important implications for this burgeoning literature, calling on scholars to inquire into the geography of personal vote-seeking and its roots in the electoral system (see Ames, 1995; Crisp and Desposato, 2004).

Of course, we have only begun to understand how institutional settings shape sub-constituency politics, as the amount of variance left unaccounted for suggests. Factors omitted from the analysis may cause some coefficients to be under- or overestimated. While there is a growing interest in the impact of a representative’s local roots on voter evaluations, the behavioural impact of the representative’s place of residence continues to be understudied. This study has explored – for the first time – the relation from the subjective perspective of the representatives, indicating a sizeable number of them think it is of striking importance. But future research will have to delve deeper into what sub-constituencies legislators see when looking at their district by using in-depth qualitative research techniques. In particular, future studies will need to inquire into the determinants of legislators’ strategic choice to cater to geographically concentrated or, alternatively, to dispersed issue sub-constituencies. More insight is needed into how legislators divide the multi-member district between them (much like ‘spokespersons’ do in terms of issues) and the ‘turf wars’ that might ensue. Only in this manner may a genuine accumulation of knowledge develop over time regarding how institutional settings – how districts are defined, in particular – shape legislators’ answers to the critical question as to ‘whom to represent’.

Footnotes

Acknowledgements

The authors would like to thank Scott Schraufnagel for his helpful comments on earlier drafts of this paper. Audrey André further acknowledges F.R.S.-FNRS for financial support.

Funding

The data used in this publication were collected by the PARTIREP MP Survey research team. The PARTIREP project was funded by the Belgian Federal Science Policy Office (grant number P6/37). Neither the contributors to the data collection nor the sponsors of the project bear any responsibility for the analyses conducted or the interpretation of the results published here. The dataset can be obtained from .

Notes

Author biographies

Audrey André is an F.R.S.-FNRS postdoctoral researcher at the Centre d’Étude de la Vie Politique (CEVIPOL) of the Université libre de Bruxelles. Her research focuses on the impact of electoral institutions on parties’, legislators’, and voters’ behaviour. Key findings have been published in Electoral Studies, Comparative Political Studies, Party Politics, Acta Politica, and Political Research Quarterly.

Sam Depauw is Assistant Professor and postdoctoral researcher at the Vrije Universiteit Brussel. His research concentrates on legislative and electoral studies. He is co-editor of Representing the People (Oxford University Press, 2014) and has published extensively on political representation and party discipline in European Journal of Political Research, West European Politics, Parliamentary Affairs, Journal of Legislative Studies, and Party Politics.

References

Ames

Barry

(1995) Electoral Strategy under Open-List Proportional Representation. American Journal of Political Science 39(2): 406–433.

André

Audrey

Depauw

Sam

(2013) District Magnitude and Home Styles of Representation in European Democracies. West European Politics 36(5): 986–1006.

André

Audrey

Depauw

Sam

Deschouwer

Kris

(2014a) Legislators’ Local Roots: Disentangling the effect of district magnitude. Party Politics 20(6): 904–917.

André

Audrey

Depauw

Sam

Shugart

Matthew S

(2014b) The Effect of Electoral Institutions on Legislative Behavior. In Shane

Martin

Saalfeld

Thomas

Strøm

Kaare

(eds) Oxford Handbook of Legislative Studies. Oxford: Oxford University Press, 231–249.

Bawn

Kathleen

Thies

Michael F

(2003) A Comparative Theory of Electoral Incentives Representing the Unorganized Under PR, Plurality and Mixed-Member Electoral Systems. Journal of Theoretical Politics 15(1): 5–32.

Bishin

Benjamin G

(2009) Tyranny of the Minority: The Subconstituency Politics Theory of Representation. Philadelphia, PA: Temple University Press.

Bishop

Bill

(2008) The Big Sort: Why the Clustering of like-Minded America Is Tearing Us Apart. Boston, MA: Houghton Mifflin.

Bogdanor

Vernon

(1985) Representatives of the People? Parliamentarians and Constituents in Western Democracies. Aldershot: Gower.

Cain

Bruce E

Ferejohn

John A

Fiorina

Morris P

(1987) The Personal Vote: Constituency Service and Electoral Independence. Cambridge: Harvard University Press.

10.

Campbell

Rosie

Cowley

Philip

(2014) What Voters Want: Reactions to candidate characteristics in a survey experiment. Political Studies 62(4): 745–765.

11.

Carey

John M

Shugart

Matthew S

(1995) Incentives to Cultivate a Personal Vote: A rank ordering of electoral formulas. Electoral Studies 14(4): 417–439.

12.

Carty

R Kenneth

(1983) Electoral Politics in Ireland: Party and Parish Pump. Dingle: Brandon.

13.

Chiru

Mihail

Enyedi

Zsolt

(2015) Choosing Your Own Boss: Variations of representation foci in mixed electoral systems. Journal of Legislative Studies 21(4): 495–514.

14.

Cox

Gary W

(1997) Making Votes Count: Strategic Coordination in the World’s Electoral Systems. Cambridge: Cambridge University Press.

15.

Crisp

Brian F

Desposato

Scott W

(2004) Constituency Building in Multimember Districts: Collusion or conflict? Journal of Politics 66(1): 136–156.

16.

Deschouwer

Kris

Depauw

Sam

André

Audrey

(2014) Representing the People in Parliaments. In Deschouwer

Kris

Depauw

Sam

(eds) Representing the People: A Survey among Members of Statewide and Sub-State Parliaments. Oxford: Oxford University Press, 1–18.

17.

Esaiasson

Peter

(2000) How MPs Define Their Task. In Esaiasson

Peter

Heidar

Knut

(eds) Beyond Westminster and Congress: The Nordic Experience. Columbus, OH: Ohio State University Press, 51–82.

18.

Fenno

Richard F

(1978) Home Style: House Members in Their Districts. Boston, MA: Longman.

19.

Gallagher

Michael

Marsh

Michael

(1988) Candidate Selection in Comparative Perspective: The Secret Garden of Politics. London: SAGE Publications.

20.

Gimpel

James G

Karnes

Kimberly A

McTague

John

Pearson-Merkowitz

Shanna

(2008) Distance-Decay in the Political Geography of Friends-and-Neighbors Voting. Political Geography 27(2): 231–252.

21.

Gimpel

James G

Lee

Frances E

Thorpe

Rebecca U

(2011) The Wellsprings of Candidate Emergence: Geographic origins of statewide candidacies in the United States. Political Geography 30(1): 25–37.

22.

Górecki

Maciej A

Marsh

Michael

(2012) Not Just ‘Friends and Neighbours’: Canvassing, geographic proximity and voter choice. European Journal of Political Research 51(5): 563–582.

23.

Górecki

Maciej A

Marsh

Michael

(2014) A Decline of ‘Friends and Neighbours Voting’ in Ireland? Local candidate effects in the 2011 Irish ‘Earthquake Election’. Political Geography 41: 11–20.

24.

Grofman

Bernard

(2005) Comparisons among Electoral Systems: Distinguishing between localism and candidate-centered politics. Electoral Studies 24(4): 735–740.

25.

Heitshusen

Valerie

Young

Garry

Wood

David M

(2005) Electoral Context and MP Constituency Focus in Australia, Canada, Ireland, New Zealand, and the United Kingdom. American Journal of Political Science 49(1): 32–45.

26.

Jewell

Malcolm E

(1982) Representation in State Legislatures. Lexington, KY: University Press of Kentucky.

27.

Katz

Richard S

(1986) Intraparty Preference Voting. In Grofman

Bernard

Lijphart

Arend

(eds) Electoral Laws and Their Political Consequences. New York: Agathon Press, 85–103.

28.

Latner

Michael

McGann

Anthony

(2005) Geographical Representation under Proportional Representation: The cases of Israel and the Netherlands. Electoral Studies 24(4): 709–734.

29.

Lundberg

Thomas C

(2007) Proportional Representation and the Constituency Role in Britain. Basingstoke: Palgrave Macmillan.

30.

Müller

Wolfgang C

Saalfeld

Thomas

(1997) Members of Parliament in Western Europe: Roles and Behaviour. London: Frank Cass.

31.

Myerson

Roger B

(1993) Incentives to Cultivate Favored Minorities Under Alternative Electoral Systems. American Political Science Review 87(4): 856–869.

32.

Navarro

Julien

(2013) Le Cumul des Mandats: Une comparaison Européenne. In François

Abel

Navarro

Julien

(eds) Le Cumul des Mandats en France: Causes et Conséquences. Bruxelles: Editions de l’Université de Bruxelles, 117–132.

33.

Nemoto

Kuniaki

Shugart

Matthew S

(2013) Localism and Coordination under Three Different Electoral Systems: The national district of the Japanese House of Councillors. Electoral Studies 32(1): 1–12.

34.

Parker

Anthony J

(1982) The ‘Friends and Neighbours’ Voting Effect in the Galway West Constituency. Political Geography Quarterly 1(3): 243–262.

35.

Patzelt

Werner J

(2007) The Constituency Roles of MPs at the Federal and Länder Levels in Germany. Regional & Federal Studies 17(1): 47–70.

36.

Rehfeld

Andrew

(2005) The Concept of Constituency: Political Representation, Democratic Legitimacy, and Institutional Design. Cambridge: Cambridge University Press.

37.

Richardson

Lilliard E

Freeman

Patricia K

(1995) Gender Differences in Constituency Service Among State Legislators. Political Research Quarterly 48(1): 169–179.

38.

Sacks

Paul M

(1970) Bailiwicks, Locality, and Religion: Three elements in an Irish dail constituency election. Economic and Social Review 1(4): 531–554.

39.

Searing

Donald

(1994) Westminster’s World: Understanding Political Roles. Cambridge: Harvard University Press.

40.

Shugart

Matthew S

Valdini

Melody Ellis

Suominen

Kati

(2005) Looking for Locals: Voter information demands and personal vote-earning attributes of legislators under proportional representation. American Journal of Political Science 49(2): 437–449.

41.

Sinnott

Richard

(1995) Irish Voters Decide: Voting Behaviour in Elections and Referendums since 1918. Manchester: Manchester University Press.

42.

Thomassen

Jacques

Esaiasson

Peter

(2006) Role Orientations of Members of Parliament. Acta Politica 41(3): 217–231.

43.

Weßels

Bernhard

(1999) Whom to Represent? Role orientations of legislators in Europe. In: Schmitt

Hermann

Thomassen

Jacques JA

(eds) Political Representation and Legitimacy in the European Union. Oxford: Oxford University Press, 209–234.

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

0.44 MB

0.25 MB