THE NEIGHBOURHOOD EFFECT IN A LOCAL ELECTION
J.S. WOLFE
Utii\w.rity of Giielph
and A.F. BURGHARDT
McMrrstrr University
ONEof the more popular ideas in electoral geography is that there is a positive
correlation between residential proximity and voting behaviour. Often referred to
as the neighbourhood effect, the idea is quite simply that individuals within a given
local area tend to vote similarly. The process involves, it is suggested, social
contacts between neighbours leading to political discussion and information flow
which exerts an influence on the way people vote. Closely related to, but distinguishable
from the neighbourhood effect is the friends-and-neighbours effect
whereby neighbours of a particular candidate will tend to know him better, discuss
him more, and support him more avidly than they will other candidates. Since the
concepts of the neighbourhood effect and the friends-and-neighbours effect have
been adopted into the literature of politicalgeography, and have become central as
explanatory models for certain spatial patterns of voting behaviour,' the rather
inconclusive and contradictory findingsof researchers usingthese concepts require
examination, and the concepts and process require an empirical test.
A recent body of literature has emphasized the salience of neighbours as a
secondary reference group for the voting decisions of individuals. Foldare2found
that individuals tend to vote in accordance with the majority of partisan voters in
their residential area. Putnam3 tested the hypothesis that community influence is
mediated through numerous personal contacts among members of the community
and that social interaction would support the political attitudes commonly held by
the community members. He found a high association between involvement in
community organizations and conformity with community norms.
Making use of research findings in innovation diffusion, Cox4 has further conceptualized
the neighbourhood effect as a product of political information flow. In
explanatory models for both the neighbourhood effect and the friends-andneighbours
effect he suggested that both effects can be attributed to the chardcteristics
of socialcontact networks, the physical and functional distance between an
individual and his social contacts, and the connectivity between them. Political
discussion and discussion of local candidates take place between neighbours and
friends, and in this way political support is transmitted outwards. Reynoldss
likewise attributed an observed pattern of distance decay in support for a local
candidate to a decrease in interpersonal contacts between his neighbours and
friends away from his place of residence. Johnston: in a seriesof papers, identified
three ways in which a spatial influence may be exerted upon political behaviour to
produce what may be regarded asa neighbourhood or local effect: the socialcontext
298
CANADIANGEOGRAPHER,XXII, 4, 1978
THE NEIGHBOURHOOD EFFECT IN A LOCAL ELECTION 299
of the neighbourhood; the residential location of the candidate; and the information
which would be generated and diffused.
These geographic models of the neighbout-hood effect and friends-andneighbours
effect rest upon the inadequately tested assumptions that social contacts
between neighbours and between friends are more effective than other social
contacts: that they do transfer information, specifically political information: and
that the contact and information transfer is influential in effecting a conversion and
yieldinga favourable vote.
There has been little detailed investigation of the processes at work producing
the identified effects. The effects have been adduced from aggregate data. While
there is considerable understanding of social contacts between neighbours' there
is less understanding of the transfer of specifically political informations and of
political discus~ion.~Although widely assumed, the process remains at base
poorly examined. It is the purpose of this paper to present the results of such an
examination.
Prior research has indicated that the neighbourhood effect and friends-andneighbours
effect are most likelytooperate in a local, multi-candidate, non-partisan
election. The study was, therefore, conducted in the small city of Guelph, Ontario
(population 66,000),prior to the civic election of 2 December 1974.In this at-large
election every eligiblevoter in the city could select up to eleven candidates from the
twenty-two who were running. As is customary in Canadian local elections, the
candidates did not bear party labels, although individual affiliationswere, in several
cases, known or presumed.
Three residential areas in the city were selected for detailed analysis. Because of
the difficulties inherent in defining macro-neighbourhoodsIO small areas were
chosen. Indeed, the study employs the concept of neighbours (people residentially
adjacent to one another who have contacts with each other, but also with friends
and business acquaintances) rather than that of neighbourhood. Selection of the
three residential areas was dependent upon criteria established for the selection of
candidates and the selection of types of residential areas.
Candidates were to be non-incumbents." To reduce confusion as to the meaning
of local candidate and neighbour, selected candidates had to be residentially isolated
from other candidates. The areas were to be residential, singlefamily, middle
to upper middle income areas, since research suggests that such areas are likely to
have high rates of political interest and political participation. Streets selected were
to be of a design conducive to interaction among neighbours.'*
These criteria narrowed the selection to the followingthree areas.
Arecr A . In area A the candidate was male, living in the corner house of a
T-junction (Figure 1).The houses in the area were built between 1967and 1969.The
candidate and over half of the residents had lived in the area six years; the others a
shorter time. Of the thirty-nine houses included two were vacant, six produced no
response, and one was the candidate's own. Thirty respondents were interviewed.
Area B. The candidate in area B was female and lived towards the middle of a
crescent-shaped street (Figure 1). The area was built up during the 1950s.Forty per
cent of the residents had lived in the area for more than ten years. The candidate had
300 LE GEOGRAPHE CANADIEN
J
____ NEIGHBOURS
ELECTWI
DISCUSSION OF UECTION AND
CANDIDATE WITH NEIGHBWR
-
REFUSE
VACANT
NORESPONSE
0 CANDIDATE
* Other houses sunwnd the area unless
otherwise ~ndicsled.
FIGUREI. Neighboui-s’discussion of election and candidates.
THE NEIGHBOURHOOD EFFECT IN A LOCAL ELECTION 301
been there for eight years. All fifteen of the households (excluding the candidate’s)
responded to the questionnaire.
Ai-eu C . Area c had a repeat male candidate (previously unsuccessful)who lived
near the entrance to a dead-end street leading to two culs-de-sac (Figure 1). The
houses were built between 1968and 1970.All residents had lived there less than six
years. The candidate was one of the first residents on the street. Of the thirty-five
households, thirty responded.
Questionnaires were administered at home during the six days prior to the
election.I3We felt it essential to interview a willing adult from each household, and
hence used no samplingtechnique. We asked a numberofquestions which we felt to
be basic to the neighbours and friends-and-neighbours models, and tabulated data
from the replies obtained.
Respondents were asked about the frequency of their informal and formal social
contacts during the two weeks preceding the election; about these channels and
media channels as sourcesof information concerning candidates; and about discussion
of candidates during their social contacts. They were questioned on their
knowledge of the candidates and on their voting intentions. Neighbours and friends
could thus be compared as “contact and source” channels with other social and
media channels. Respondent discussion of, knowledge of, and intent to vote for the
local candidate or other candidates could also be compared. Respondents were
asked to identify on a map those neighbouring households they visited, those with
which they had discussed the election, and those with which they had discussed
candidates.
Carried to a logical extreme the neighbourhood model assumes not only that
neighbours are a major social contact, but that they serve as a dominant source
channel for political information and that friends play a similar, if lesser role.
The Guelph data indicate that neighbours are no more important in this regard
than friends. Whereas forty per cent of the respondents mentioned receiving from
neighbours candidate information (including of course information about the candidate
who is a neighbour), forty-two per cent mentioned friends. Twenty per cent
gave businesscontacts, seventeenper cent clubs, eight percent work, four percent
school contacts, and none mentioned a church. There was considerable variation
among the three areas: the neighbours’ percentage was forty-threein A, fifty-nine in
B (the older area), and only twenty-seven in c.
Are potential voters more likely to discuss an impending election with their
neighbours than with their otherinformal social contacts?Greer14postulated that of
the various social contacts (neighbours, friends, and fellow workers) neighbours
would have the highest frequency of political discussion. His results, however,
revealed high levels of political discussion for kin and fellow workers, and lower
levels for neighbours and friends. He suggested that politics was too sensitive a
topic for social contacts with the latter. Table I , from the Guelph data, shows very
high levels of discussion within the household. Neighbours’ and friends’ rates of
discussion were virtually identical, and much lower; but the discussion rate with kin
outside the household was lower still.
In terms of discussionof candidates,discussionbetween neighbours(26.4%) was
.
302 THE CANADIAN GEOGRAPHER
TABLE I
DISCUSSION OF THE ELECTIONAMONG NElGHBOURS AND THROUGH OTHER INFORMAL
SOCIAL CHANNELS
(percentage of the respondents mentioning the channel, by area)
With With With non-local Within the
Area neighbours friends relatives household
A, N = 30 16.6 20.0 3.3 36.6
B, N = 15 13.3 20.0 13.3 66.6
C, N = 30 23.3 13.3 6.6 66.6
a poor second to that between members of a household (42.6%), but well ahead of
that between friends (13.3%), with relatives outside the household (5.3%),in clubs
(4.0%), and in professional organizations (1.3%). By area, discussion between
neighbours followed the above-mentioned pattern: area A had the highest figure
(40%),and area c the lowest (17%).
The specifichouse-to-house contacts indicated in Figure 1 make it clear that there
is a great variability in discussion among neighbours. In area A , where the candidate
lived at the T-junction, almost all of the discussion of candidates occurred on one of
the streets, and that mostly between five households adjacent to the candidate. In
area B all discussion took place in the half of the crescent which included the
candidate. In area c,however, there was virtually no discussion by households near
the candidate, but a great deal in the inner, circular end of the two culs-de-sac,
reinforcing the idea that physical configuration plays a very significant role at the
micro-level.
A key assumption of the friends-and-neighbours model is that neighbours will
have a higher level of knowledge of their local candidate than they will of other
candidates. Only in the compact crescent of area B was this demonstrably true;
there, seventy-three per cent of the respondents knew their local candidate personally,
whereas no other candidate was known personally by more than seven per
cent. In area A the local candidate was known personally by only a third; the next
highest candidate was known by only thirteen per cent. In area c only eleven and a
half per cent knew the local candidate personally, whereas another candidate was
known by fifteen per cent. Nor should too much be assumed fom these levels of
knowledge, which can be very superficial. Only 11.7 per cent of the respondents
knew the occupation of their local candidate. and only thirteen per cent knew his or
her political affiliation.
Table 11 indicates that those electors who did discuss candidates mentioned the
local contender more than other candidates. Area c was a noteworthy exception.
Within other social circles non-local candidates were discussed almost as frequently
as the local. The exception was discussion of the local candidate with
friends, by respondents from area A .
Because of their proximity to the local candidate it was expected that the
respondents would notice their neighbour more than his rivals when contenders
were featured by the media. The responses negated this expectation. In all the
pertinent forms of the media - posters, newspaper articles, newspaper advertise-
THE NEIGHBOURHOOD EFFECT IN A LOCAL ELECTION 303
TABLE I1
NEIGHBOURS’ DISCUSSION OF THE LOCAL CANDIDATECOMPARED WITH DISCUSSIONOF OTHER CANDIDATES
(percentage of respondents mentioning the channel, by area)
With non-local Within
With neighbours With friends relatives household
Highest
Highest Highest Highest
Area Local other Local other Local other Local other
A, N = 30 47.36 10.52 50.00 16.66 0 0 15.09 13.20
B, N = 15 85.71 14.28 16.66 16.66 33.33 33.33 19.23 15.58
C, N = 30 14.28 21.42 10.71 10.71 0 33.33 5.40 13.51
ments, and radio- the highest outside candidate gained a level of notice well above
that of the local candidate. If nothing else, this fact suggests that the attention
gained from proximity cannot be expected to screen out the media messages of the
competition.
Research utilizingaggregate data had led to the conclusion that political influence
is exerted by neighbours (or neighbourhoods) and that support for local candidates,
which declines with distance from the candidate, is attributable to neighbours and
friends, in those social interaction networks that are most active, supportive, and
influentialclose to the candidate’s place of residence.
Two clusters of ideas have been investigated in this paper. The first is that people
living in close proximity have a high probability of social contact with each other,
and that communication of a political nature is included in this contact. The general
belief was that this contact and communication is more frequent and more effective
than that with persons livingoutside the immediate vicinity. The second set of ideas
follows upon the first. Persons living close to a candidate will tend to discuss him
more, notice his name more, and vote more readily for him than for any other
candidate. Proximity has political consequences. There is an implicit belief that
physical closeness breeds support. These two setsof ideas make up the essentials of
the neighbourhood-based models as applied in electoral geography.
The analysis presented above throws some doubt on the general validity of this
theory and on the two processes which it assumes. Those areas and candidates
selected were believed likely to display high levels of political interest and communication.
Contrary to the first part of the theory, however, contact does not
necessarily include political communication. Political communication is highest
within a household and relatively low between neighbours and friends. Neighbours
do not necessarily have the same level of interest in politics, nor do they necessarily
communicate political information. The second part, concerning proximity to a
local candidate, is partially substantiated by the higher knowledge level of the local
candidate than of the others. However the lack of awareness of the candidate’s
affiliation and occupation shows how superficial such knowledge was. The local
candidate was indeed discussed more than were others, but this interest was not
often converted intofavourable votes. Only in the smallest and most stable area did
the local candidate get more local support than the others (Table 111).
Perhaps the most interesting findings of this study were the differences existing
304 LE GEOGRAPHE CANADIEN
TABLE 111
(asa percentageof total mentions by respondents, by area)
Area Local candidate Highest other candidate
INTENT TO VOTE FOR THE LOCAL CANDIDATE
A 8.42
B 11.36
C 5.64
10.52
10.22
11.29
among the three areas. We are left with strong support for the notion that the age of
the area and the age and length of residency of the inhabitants have a considerable
impact on the degree to which the neighbourhood theories apply at the local level.
Only area B , half of whose voters have lived there ten years or more, fits the model.
The newest area, c,deviates furthest from the expectations of the model.
At the micro-leveladditional extraneous factors intervene. Area B fitsthe theory,
but is also the smallest and most compact area, with only half as many respondents
as the other two. Area B also had the only woman candidate, which may or may not
have been a factor stimulating discussion and knowledge of the candidate. The
shapeof the area (twoculs-de-sac located slightlyaway from the candidate’s home)
may explain the deviating results from area c.
In conclusion, neighbours operate as one of several informal personal contact
channels, are generally of less significance than friends, and are not much more
significant than work or business channels. Neighbours are significant as an information
channel onlyfor a local candidate. The existence of the neighbourhood flow
of information is not, however, any guarantee of support for the local candidate.
Other channels, namely friends, fellow workers, business contacts, and the newspaper,
connect neighbours to other areas and other candidates. With these contacts
neighbours spread their voting support beyond the immediate neighbouring area.
Considered as explanations for the processes of political information flow and
political support, the neighbourhood effect and friends-and-neighbours effect
models need to be applied with caution and with due consideration for scale and
local circumstance.
REFERENCES
1. Cox, K., “The Neighbourhood Effect in Urban Voting Response Surfaces,” paper presented at a
conference on models of urban structure sponsored by the Ohio State University and Battell
Memorial Institute, Columbus, Ohio (1969); “The Spatial Structuring of Information Flow and
Partisan Attitudes,” in M. Doggan and S. Rokkan (eds.), Quuntitcrtive Ecological Analysis in the
Social Sciences (Cambridge, Mass.: M.I.T. Press, 1969);“The Voting Decision in a Spatial Context,”
in C. Board et al. (eds.). Progress in Geography, 1 (1969), 83-118; “The Genesis of
Acquaintance Field Spatial Structures: A Conceptual Model and Empirical Tests,” in K. Cox and
R. Golledge (eds.), Problems in Behavioral Geogruphy (Northwestern University, Studies in
Geography No. 17, 1969). 146-68. REYNOLDS, D., “A Friends-and-NeighboursVoting Model as a
Spatial Model for Electoral Geography,” in K. Cox and R. Golledge, Problems in Behuvioral
Geogruphy.81-100. JOHNSTON, R.J., “Spatial Patterns and Influenceson Votingin Multi-candidate
Elections: The Christchurch City Council Elections, 1968,” Urban Studies, 10(1973), 69-82.
2. FOLDARE, IS., “The Effect of Neighbourhood on Voting Behaviour,” Pol. Sti. Quort., 83 (1968),
516-29.
TH E NElCHBOURHOOD EFFECT IN A LOCAL ELECTION 305
3. PUTNAM.R., “Political Attitudes and the Local Community.” Ar?!er. Pol. Sc,i. RCI...60 (1966).
640-54.
4. Cox. ”The Neighbourhood Effect,’’ 1969.
5. REYNOLDS.“Friends-and-Neighbours Voting Model,” 1969.
6. JoHNs.roN. “Spatial Patterns.” 1973; “Spatial Elements in Voting Patterns at the 1968Christchurch
City Council Elections.” Pol. Sci.. 24 (1972). 49-61: “Local Effects in Votingat a Local Election,”
Atin. Assoc. Ar~7c.r.Geog., 64(1974). 418-29.
7. AXELROD,M., “Urban Structure and Social Participation.” An7rv. Sociol. Rrrz.. 21 (1956). 13-18:
Hrrriitrrr Factors ir7 Hoirsing (Stanford. Calif.: Stanford University Press, 1963); GRLE.R.S., “Urbanism
Reconsidered: A Comparative Study of Local Areas in B Metropolis.” Arnc~r.Sociol. Rei,..
FtSTINGER, L.. SCHACHTER.s..and BACK.K . , Socict/ fre.ssrrrrs if7 Irf$orrritr/ Gvolrp: A StrrcIy of
21 (1956). 19-25.
8. FITTON,M., ”Neighbourhood and Voting: A Sociometric Analysis,” Brit. J . Pol. Jci., 3 (19731,
445-72.
9. GREER,S.,Metropolirics: A Stucly ofPoliric,trlCiilrrw (NewYork: Wiley. 1963).
10. LEE, T.. “Urban Neighbourhoods a s a Socio-spatial Scheme,” Hurnrrri Rrlrtioris. 21 (1968),
241-67.
1I. JOHNSTON.“Spatial Patterns,” 1973, and REYNOLDS,“Friends-and-Neighbours Voting Model.“
1969.found both effects to be most pronounced for new candidates or non-incumbents.
12. FESTINGERet al., Socidl frrssrrves. 1963.
13. The complete questionnaire forms Appendix I of JACQUELINES. WOLFE.“Political Information
Channels in an At-Large Municipal Electoral System’’ (Doctoral Dissertation, McMaster- University.
1975). 292-321.
14. GREER,Merropolirics, 1963.