Political Studies (1994), XLII, 469-479
Measuring Local Campaign Effects:
Labour Party Constituency Campaigning
at the 1987 General Election
CHARLESPATTIE
University of Nottingham
PAULWHITELEY
College of William and Mary, Virginia
RONJOHNSTON
Unit7ersity of Essex
AND PATRICKSEYD*
Unicersity of Shefield
Thursday 31 March: Polling day. To headquarters at about 9 and from 9.30 am to
6 visiting all the polling stations and committee rooms. We had lunch at the
'Hole-in-the-Wall' and then went out again with our loudspeaker, doing final
knocking up. After the polls closed we cleared up the headquarters, went back to
the hotel and began watching the BBC Election programme. It was evident that a
big Labour victory was coming . . .'
During election campaigns, political parties engage in intense activity aimed at
winning over voter support. If they are to win, the parties must both mobilize
their existing support and reach out to those who have not yet decided. At the
1987 election, for instance, most voters had made their choice before the start
of the campaign (about 70% of respondents to the 1987 British Election Study
claimed to have done so over a year before, and 80% had decided before the
campaign began): thus a major goal of the campaign effort was to make sure
they actually voted. In addition a still substantial minority (around 21% in the
1987 BES) decided during the campaign itself (Miller and his co-workers, using
Part of the research for this paper was funded by ESRC grant no. ROOO 231522 and by the
University of Sheffield Research Fund. Data from the British Election Study were supplied by the
ESRC Data Archive at Essex University. The data for the Labour party membership study have
been lodged with the Data Archive. We gratefully acknowledge the support of both bodies.
' T. Benn, Our of /he Wilderness: Diuries 1963-1967 (London, Hutchinson, 1987) p. 399
t. PoliticalStudies Awoarion 1994. Publirhcdby Bkckwcll Publishers. 108Cowlcy Road. Oxford OX4 IJF. UK and 238 Main Slrar.
Cambndgc. M A 02142. USA.
470 Research Note
a different methodology, estimate that up to 38% of voters changed their minds
at least once during the ‘short campaign’ from March to June 19872);these
‘floating voters’ were a major target for the campaign.
The resources mobilized by parties during the campaign are very large
indeed. At a national level there are few restrictions on how much parties can
spend in Britain, and the major parties lay out large amounts of money for
advertizing and publicity in the press and on billboards3Television and radio
campaigns also pay an important role, though, unlike the print media, access
to the electronic media is carefully monitored to ensure fairness.* The effects
of this national campaign have been widely discussed.’ Less attention has been
given, however, to constituency campaigns. But there is controversy over
whether these local campaigns really affect the result of an election. Furthermore,
quite how they might affect results is a matter of debate. The paper picks
up these controversies and discusses whether, and how, local campaigns affect
local results, with particular reference at Labour’s 1987 campaign.
The Constituency Campaign
For most commentators, local campaigns make no difference to modern British
elections. In a mass media age, voters get their main political news from the
national television, radio and press. Few attend campaign meetings in their
constituencies (only 2.7% of respondents to the 1987 BES said they had done
so). Thus, it is argued, ‘the local campaign is indeed much of a ritual’, having
‘little success in changing political attitudes’.6
Local Campaign Actii~ism
A number of studies have produced findings which run against the accepted
view and suggest an important role for the local campaign. However, many of
these studies are one-off investigations of one or two constituencies.’ There has
been little comparative work on local campaign effects throughout the whole
country. Recent research on Labour party members provides an important
exception.’ Seyd and Whiteley questioned Labour party members throughout
W. L. Miller, H. D. Clarke, M. Harrop. L. Leduc and P. F. Whiteley. How Voters Change:
the 1987 British Election Campaign in Perspecriae (Oxford, Claredon, 1990).
’M. Pinto-Duschinsky, British Polirical Finance 1830-1980 (Washington. American Enterprise
Institute for Public Policy Research. 198I); M.Pinto-Duschinsky, ‘Trends in British political
funding, 1979-83’. Par1iamenrar.v Aflairs. 38 ( 1989, 32847; M. Pinto-Duschinsky. ‘Trends in
British political funding, 1983-87’ Parliamentary Aflairs, 42 (1989). 197-212.
‘H. F. Rawlings, Law and the Elerroral Process (London, Sweet and Maxwell, 1988).
’1. Crewe and M. Harrop (eds),Political Communications:the General Election Campaign of 1983
(Cambridge, Cambridge University Press. 1986); I. Crewe and M. Harrop (eds), Political
Communicarions:$heGeneral Eiection Campaign of 1987 (Cambridge, Cambridge University Press,
1989); Miller et (11.. How Voters Chuge.
’D. Kavanagh, Constituency Electioneering in Britain (Longman, London, 1970). pp. 47, 87.
’R. T. Holt and J. E. Turner, Politicul Parries in Action: rhe Battle of Barons Court (New York,
Free. 1968);J. M. Bochel and D. Denver ‘Canvassing, turnout and party support; an experiment‘
British /ournu/ oJPolirical Science. I (1971). 25749; J. M. Bochel and D. Denver ’The impact of
the campaign on the results of local government elections’ Brztish Journal of Political Science, 2
(1972). 239-60.
‘P.Seyd and P. Whiteley, Labour’s Grass Roots: the Politics of Parry Membership (Oxford,
Clarendon, 1992).
I: Polilical Studm A ~ r u o n .1994
Research Note 471
Britain on their levels of local activism during the 1987 election. Their
indicators of activism were: whether the party member displayed a poster
during the campaign; whether leaflets were delivered and whether voters were
canvassed. These were used to calculate activism scores for individual constituency
parties. Information from the 1987 British Election Study (BES) on
whether voters had been canvassed during the campaign was also used to
estimate how many voters had been canvassed in each of the 191 constituencies
sampled in the BES and in Seyd and Whiteley’s study.9
Seyd and Whiteley demonstrated the efficacy of local campaigning. They
found that the more active a constituency Labour party was then: more
voters reported being canvassed; Labour’s share of the vote was higher;
and the greater the swing to Labour between 1983 and 1987. Local
activism remained an important correlate of the Labour vote in a multivariate
model which controlled for the effects of constituency marginality,
social composition and previous levels of party support.” Seyd and
Whiteley’s results indicate the importance of local campaign effort in an era of
mass media politics, and they estimate that ‘if every local (Labour)
party had recruited an additional 100 members prior to the 1987 election, and
each of these increased the activism score by an average of 8.5 points’,
the party would have won nearly 36% of the vote, rather than the 31% it
actually got.’’
Local Campaign Spending
There is also a handful of studies which use the amount spent on campaigning
by a party in each constituency (to pay for, for instance, the printing of
handbills, posters, etc.) to estimate the effects of local campaign effort over a
large number of constituencies.” Their authors argue that spending has two
possible effects on the outcome. Firstly, and directly, the more a party is able
to spend locally, the more local votes it is able to ‘buy’ through advertizing.”
Secondly, local spending is assumed to be a surrogate for
campaign activity: the more a party spends locally, the more active it is
assumed to have been during the campaign there, so that if canvassing activity
wins votes, the level of spending is indicative of the volume of vote-winning
activity.
Seyd and Whiteley, Lobour’s Grass Roors. p. 182. In the constituencies sampled, there were on
average 15.3 BES respondents and 10.6 respondents to the Labour party survey.
lo Seyd and Whiteley, Lubour’s Grass Roots, Table 8.7.
” Seyd and Whiteley, Lobour’s Grass Roo& pp. 1976.
l 2 D. Denver and G. Hands, ‘Marginalityand turnout in Britishgeneral elections’. Brirish Journal
oJPoliricu1Science. 4 (1974), 17-35; D. Denver and G. Hands, ‘Marginality and turnout in British
general elections in the 1970s’ British Journal offolirical Science, I5 (1985), 381-8; R. J. Johnston.
‘Campaign spending and votes: a reconsideration’, Public Choice,33 (1979),83-92; R. J. Johnston,
‘Party strength, incumbency and campaign spending as influences on voting’, Tijdschryr umr
Economische en Sociale Geogrufle, 76 (1985). 82-7; R. J. Johnston, ‘A further look at British
political finance’, Political Studies, XXXIV (1986). 466-73; R. J. Johnston, Money and Votes:
Consriruency Campaign Spending and Election Results (London, Croom Helm, 1987); R. J.
Johnston. C. J. Pattie and L. C. Johnston, ‘The impact of constituency spending on the result of
the 1987 British General Election’. Electorial Studies, 8 (1989). 143-55.
” R . J. Johnston, ‘Political advertising and the geography of voting in England at the 1983
General Election’, lnrernationul Journal o/ Advertising, 4 (1985). 1-10.
e Pohiicpl Sludm Asscai!on. 1994
472 Research Note
Parties are limited in their spending on the constituency campaign in British
elections. Legislation imposes maximum limits.14and in 1987,the maxima were:
in a county constituency,f3,370 plus an extra 3.8 pence for every entry in
the electoral register used at the election; and
in a borough constituency, f3,370 plus an extra 2.9 pence for every entry
in the register.
The difference between county and borough constituencies reflects the larger
geographical extent of the former, and hence the assumed greater difficulties
and expense of campaigning there. Local parties did not spend to the upper
limit in many constituencies, however; in 1987, the average constituency
Labour party spent only 71% of the allowed amount, with a standard deviation
of 24%. (Spending ranged from 16% to 100% of the legal maximum.)
Constituency parties are not always rational in their spending, and the
relationship between constituency marginality and spending is not particularly
strong. But however irrationally it is used, local campaign spending is related
to the constituency vote, particularly where a party is the ~ha1lenger.l~
For the first time, then, there are three different estimates of the extent of
local campaign effort over a large number of constituencies for the Labours
party's performance in the 1987 general election: Seyd and Whiteley's direct
measurements of local activism in the Labour party; data from the 1987 BES
on canvassing; and Johnston et d ' s relatively indirect estimates using campaign
spending data. In the following analyses, all three types of data are used in
analyses of the 191 constituencies for which data are available both from Seyd
and Whiteley's study and from the 1987 BES.
Campaign Effort: Measures of the same Thing?
Work on local campaign spending has been criticized. Gordon and Whiteley,
in a critique of an earlier paper by Johnston, argue that expenditure data are
inaccurate and are thus poor indicators of local campaign effort, since 'in
Britain elections at the local level are run almost entirely on voluntary labour
and the bulk of election activity does not involve financial transactions at a11'.I6
Evaluating the worth of local campaign spending data has important
implications for the analysis of voting behaviour. If the local campaign is seen
as a potential influence on the vote, then some measure of campaign activity
is needed. But gathering data such as those collected by Seyd and Whiteley for
all constituencies and for all parties is expensive and time-consuming. Spending
data must be published by Parliament, are widely available and inexpensive.
I' Pinto-Duschinsky. British Politicul Finance; R. J. Johnston and C. J. Pattie 'Great Britain;
twentieth century parties operating under nineteenth century regulations' in A. B. Gunlicks (ed.),
Cornparalice Campaign and Partv Finance in North America and Western Europe (Boulder CO,
Westview, 1993).
"R. J. Johnston. 'Campaign spending and the efficacy of advertising at the 1974 General
Election in England, Poliriid Srudies. XXVIl (1979). 11&9; R.J. Johnston er a).. 'The impact
of constituency spending'; R. J. Johnston. Money and Votes.
l6 I. Gordon and P. Whiteley. 'Comment: Johnston on campaign expenditure and the efficacy
of advertising'. Poliricol Studies. XXVlIl (1980), 291; R. J. Johnston, 'Campaign expenditure and
the efficacy of advertising'. For a response, see R. J. Johnston, 'Campaign expenditure and the
efficacy of advertising: a response'. Political Studies, XXIX ( I98I ), I 1 3 4 .
I. P o l i t d S t d i n Asmciauon. 1994
Research Note 473
However, if there is doubt over their meaning, then their value is lost. Some
means of assessing the substantive relationship between ‘real’campaign activity
and spending is needed.
The data employed here fulfil that need. They make possible a test of
Johnston’s claim that campaign spending can be used as a surrogate for
campaign activism. Regression models were fitted to estimate the connections
between local party activism, local canvassing and party spending for the
Labour party in 1987.Local activism was measured using Seyd and Whiteley’s
constituency activism score (ACTIVE); the number of voters in each constituency
claiming to have been canvassed by Labour in 1987was derived from
the British Election Study (CANVASS); and the amount spent on the campaign
by constituency Labour parties was expressed as a percentage of the legal
maximum in that constituency (LSP87PC.)”
If campaign activism is related to voters’ experience then we should expect
that the more active the local party, the more voters who reported being
canvassed. A simple bivariate model, with CANVASS as the dependent,
confirms the hypothesis:
CANVASS = 11.43 +O.l6(ACTIVE) R ’=0.17
(6.20)
Labour’s local campaign activity was significantly and positively related to
the number of local people reporting being canvassed. However, the relationship
Is not a strong one.
If reported spending levels are reasonable surrogates of campaign effort,
the latter being measured both. by self-reported local activism and by the
‘output’ measure of voters reporting being canvassed, then the more active a
local party is, and the more voters who are canvassed by it, the closer to its
legal maximum its spending should be. A multivariate model was fitted, with
Labour spending as the dependent, and campaign activism and canvassing as
the independents:
LSP87PC = 43.83 +0.45(CANVASS) +0.I6(ACTIVE) R ’=0.33
(5.40) (4.94)
Once again, the regression model confirms expectations. Both CANVASS
and ACTIVE were significantly related to LSP87PC and in the expected
directions. However, the relationships are once again relatively weak: CANVASS
and ACTIVE together account for only a third of the variance in
LSP87PC.”
As a final test of the ‘surrogate measure’ argument, LSP87PC, ACTIVE and
CANVASS were entered into a principal components analysis. This produced
a one-component solution which accounted for almost two thirds of the
original variance (Table 1). All three measures were strongly and positively
” Spending data were taken from Elerrion Expenses Cmd. 426 (London, HMSO,1988).
’’The amount spent on the local campaign was a better predictor of reported canvassing activity
than was local party activism: a bivariate regression of LSP87PC and CANVASS yielded an R’
of 0.24. rather better than the equivalent values for the bivariate regressions of CANVASS and
ACTIVE (R = 0.17) and of LSP87PC and ACTIVE (R = 0.22). Local campaign spending is no
worse an indicator of canvassing than more direct measures of local activism!
I: Polilicdl Studies Asrooallon. 1994
414 Research Note
TABLE1. Campaign Effort:
Principal Components Analysis
Component I
LSP87PC 0.82
CANVASS 0.79
ACTIVE 0.78
Eigen value 1.91
YOvariance 63.8
explained
related to the component, indicating that they are all to a considerable extent
measuring the same thing. On the basis of these findings, then, it seems
reasonable to argue that local campaign spending is an acceptable measure of
campaign effort. This is not to say that spending is an accurate guide to
absolute levels of local campaigning: the data are not sufficiently reliable for
this. However, as the PCA results show, it is a good guide to relative activity.
While there are deviations from the 'true' level of campaigning these are
random across constituencies: systematic errors would have reduced the
loading of spending on the campaign component.
Measuring the Effectiveness of the Local Campaign
Unless the considerable party resources, in terms both of money and party
members' time and enthusiasm, expended in the constituency campaign have
a tangible and favourable effect upon the constituency election result, it is
largely wasted effort, contributing perhaps to party morale but doing little to
change the election outcome. In this section, the effectivenessof campaign effort
is estimated. Campaign effort is measured by the standardized component score
(CAMPAIGN) derived from the analysis reported in Table 1. Scores of greater
than zero on CAMPAIGN indicated above average constituency campaign
effort by the local Labour party; scores below zero indicate below average
effort.
Conventionally, estimates of the effectiveness of local campaign effort use as
dependent variables the percentage share of the vote won by a party and the
percentage point change in that share since the previous e1ecti0n.I~Such
analyses obscure important patterns of change. In particular, an emphasis on
net change in vote share between elections ignores the even larger gross shifts
in party support. For instance, it would be interesting to know not only whether
Labour campaign effort was a significant factor in accounting for the party's
vote in a constituency but also whether it was more effective in winning over
more former centre party supporters than former Conservative voters. In order
to allow the examination of the latter possibility, entropy maximizing estimates
of the constituency flow of the vote between 1983 and 1987 were used in the
analyses reported below.'
"J. M. Bochel and D. Denver, 'Canvassing, turnout and party support'; R. J. Johnston,
'Campaignexpenditure and the efficacy of advertising';Seyd and Whiteley, Labour's Grass
Roofs.
,C Polilial Studin Arurulion. I994
Research Note 475
TABLE2. Measuring the Effectiveness of the Local Campaign: Regression of Constituency
Flow-of-the-vote against Local Campaign Effort
FROM: 1983 CON LAB LAB LAB LAB ALL RNV
TO: 1987 LAB CON LAB ALL RNV LAB LAB
CAMPAIGN 0.66 -1.27 6.30 -2.55 -2.48 3.35 3.29
CONST 3.06 5.43 68.85 9.42 16.30 15.91 17.23
RZ 0.13 0.28 0.33 0.23 0.34 0.21 0.28
The vote flows which were of most concern here were to and from the
Labour party between 1983 and 1987. Specifically, attention ocused on flows
from Conservative in 1983 to Labour in 1987; from Labour to Conservative;
Labour to Labour; Labour to the Alliance; Labour to other parties and
non-voters; the Alliance to Labour; and other parties and non-voters to
Labour. These flows were estimated as percentages of the 1983 vote for the
party of origin: for instance, the percentage of 1983 Conservative voters who
supported Labour in 1987 is the flow from Conservative to Labour. The
purpose of Labour’s constituency campaigns was both to retain those voters
who voted Labour at the last election and to gain the support of those who
had previously voted for another party. The hypothesized relationship between
flow-of-the-voteand campaign effort in 1987is, therefore, that the more effort
a constituency party put into the local campaign in 1987, the greater should
be the flows from all parties to Labour and the lower should be flows from
Labour.
Straightforward tests of this hypothesis regressed the flow-of-the-vote
estimates against CAMPAIGN for the 191 constituencies. Campaign effort
proved to be significantly related to all the flow estimates (Table 2: only those
regression coefficients which were significant at the 95% level are reported
here). Furthermore, the relationships were all in the expected directions:
positive for flows to Labour, and negative for flows away from the party. The
greater the effort put into the local campaign by a constituency Labour party,
the higher were Labour’s gains and the lower its losses.
Local campaigning by the Labour party seems to have had a
more pronounced impact on the 1987 votes of those who had voted for
Labour, the Alliance or who had supported a minor party or abstained in
1983, than on the votes of those who had voted for the Conservatives at
that election. For instance, a one standard deviation increase in CAMPAIGN
meant that the flow from Conservativeto Labour would rise by 0.66 percentage
points. But the same increase in campaign effort would result in a rise of
3.35 percentage points in the flow from the Alliance to Labour, and of
6.30 points in the proportion of 1983 Labour supporters who stayed loyal
in 1987. But local campaign efforts did not account for a great deal of
the inter-constituency variation in the flow of the vote. Values of R 2 never
exceeded 0.35; typically, only between a fifth and a third of the variance was
explained.
mR.J. Johnston, The Geography of English Politics (London, Croom Helm, 1985); R. J.
Johnston, C. J. Pattie and J. G. Allsopp, A Marion Dividing? The Electoral Map of Great Brirain.
1979-1987 (London, Longman, 1988).
IPolitical Studies Assmintion, 1994
476 Research Note
Towards a Fuller Model
Local campaigning is just part of a ‘lifetime of learning’2’Taking those wider
influences on the vote into account requires a more detailed model than that
outlined in Table 2. Among the factors which might influence the geography
of the flow of the vote, apartfrom local campaign effects, are a number of longand
short-term factors: the class composition of the constituency; the previous
state of the parties; the marginality of the seat; its regional context; and so on.U
Most of these factors reflect longer term influences on the local vote and they
should be incorporated in a more fully specified model.
That specification was provided by adding the control variables employed by
Seyd and Whiteley.” The social composition of each constituency was
measured by the percentage of workers employed in manual occupations in
1981 (MANUAL) and the percentage of council tenants in the same year
(COUNCIL): we expect that flows to Labour would increase (and flows from
the party decrease) as the percentages of manual workers and of council tenants
in a constituency increased. The local context of the seat may have an
independent influence on the flow of the vote. We control for the percentage
of agricultural workers in the seat in 1981 (AGRIC): Labour tends to do worse
in rural seats, cereris paribus, than elsewhere.24The more marginal a constituency,
the more an individual voter’s actions might influence the result. We
include a measure of the marginality of each constituency for the Labour
candidate (MARG): where Labour won the seat, MARG is measured by the
absolute difference between the Labour vote and the party in second place;
where Labour did not win, it is the absolute difference between Labour and the
winning party. Furthermore, increases in turnout from one election to another
can influence the flow of the vote, especially where differential turnout takes
place. We include a measure of the change in turnout between 1983 and 1987
(TURNCH). Finally, we also control for the regional context. During the
I980s, several commentators identified a widening north-south divide, in which
Labour gained ground in the north and lost ground in the south.” A dummy
variable (REGION) is included in the model to take account of this: constituencies
in Scotland, Wales, Yorkshire, the North-West or the north, REGION
were coded as 1 and the remainder were coded zero.
Adding these controls to the models reported in Table 2 increased R values
by 45% on average, such that most of the models had R’ values over 70%
(Table 3). In general, then, the controls provided a better guide to the flow of
the vote than did campaign effort. The local campaign appears to have
influenced voters ‘at the margin’. However, in particular contexts, and especially
in closely fought contests, this ‘marginal’ influence might mean the
difference between winning and losing a seat.
‘‘R.Rose and 1. McAllister, The Loyalries o/ Voters:u Lijetime Leurning Model (London, Sage.
22 W. L. Miller, Elecrorol Dynomics (London, Macmillan, 1977); R. J. Johnston er al., A Nation
zJ Seyd and Whiteley, hbuur’s Grass Roors. pp. 190--3.
”1. Crewe and C. Payne. ‘Another game with nature: an ecological regression model of the
” J . Curtice and M. Steed, ‘Analysis’in D. Butler and D. Kavanagh (eds), The British General
1990).
Dividing?.
British two-party vote ratio in 1970’. British Journal of Poliiicul Science, 6 (1976). 43-81.
Election o/ 1987 (Basingstoke, Macmillan. 1988); R. J. Johnston et a/.. A Nation Diciding?
1 Polilia1 Sludm Assccia~lon. 1994
Research Note 477
Broadly speaking, the ‘long term’ factors behaved as expected. Flows to
Labour were higher in more working class seats and in seats with a large
proportion of council tenants. Furthermore, the party was less successful in
winning and keeping support in highly rural areas than in more urban places,
but did better in northern than in southern constituencies. Measures of the
political context in each seat provided more ambiguous results, however.
Changes in turnout had a negligible influence on most flows (the exceptions are
the flow from Labour to the Alliance - increasing turnout increased the flow;
and the flow from Labour to minor parties and abstention - increased turnout,
not surprisingly, decreased the flow). Labour loyalty was greatest in the party’s
safest seats, and in its most hopeless ones but there was more turmoil in the
Labour vote as the competition between Labour and other parties in a
constituency became more intense.
Encouragingly, even with extra controls in the equations, the relationships
between campaign effort and the flow of the vote were again significant and in
the expected directions. Where Labour put relatively large resources into the
local campaign, it enjoyed above average flows in its favour, and below average
flows away from it. An increase of one standard deviation in CAMPAIGN
meant: a fall of 0.29 percentage points in the flow from Labour to the
Conservatives; a fall of 0.88 percentage points in flows from Labour to minor
parties and to abstention; a rise of 1.72 in Labour loyalty; rises of 1.04 points
in flows from the Alliance and from ‘other parties and non-voting’ to Labour;
and a rise of 0.39 percentage points in flows from the Conservatives to Labour.
While these effects were not large, they could have been decisive in closely
fought constituency competitions.
The Effects of Campaigning by Other Parties
The above analyses establish the importance of Labour’s local campaigns in
1987. However, the party does not campaign in a vacuum. A Labour
constituency party’s efforts could have been offset by the similar efforts of its
Conservative and Alliance rivals. To take this into account, the models reported
in Table 3 were rerun, adding measures of campaign effort for the Conservatives
and the Alliance as well as for Labour (Table 4). In the absence of direct
measures of activism by local Conservative and Alliance party members, these
TABLE3. Local Campaign Effects:a Fuller Regression Model
FROM: 1983 CON LAB LAB LAB LAB ALL RNV
TO: 1987 LAB CON LAB ALL RNV LAB LAB
MANUAL 0.07
COUNCIL 0.03
TURNCH
MARG 0.02
AGRIC -0.09
REGION 1.66
CAMPAIGN 0.29
R ? 0.74
CONST -2.24
IPolltirdl Studies Asscaation. 1594
-0.08
-0.04
0.03
0.08
-1.98
-0.29
9.89
0.76
0.44 -0.16
-0.05
0.62
-0.16 0.07
-0.67 0.42
6.02 -1.72
1.72
49.62 15.18
0.71 0.56
-0.20 0.33 0.28
0.09
-0.43
0.06
0.17 -0.51 -0.34
-2.31 4.32 4.91
-0.88 1.04 1.04
25.30 -2.44 2.33
0.67 0.70 0.74
478 Research Note
TABLE4. Local Campaign Effects: Accounting for the Effects of Campaigns by Other
Parties
FROM: 1983 CON LAB LAB LAB LAB ALL RNV
TO: 1987 LAB CON LAB ALL RNV LAB LAB
MANUAL
COUNCIL
TURNCH
MARC
AGRIC
REGION
SDP
CSP87PC
LSP87PC
ASP87PC
CONST
R =
0.04 -0.03 -0.14
0.03 -0.02 -0.06
0.10 0.46 -0.48
0.02 0.03 -0.22 0.11 0.08
-0.05
1.58 -1.75 5.12 -1.05 -2.32
-2.28 1.44
-0.01 0.03 -0.06
0.02 -0.02 0.08 -0.04 -0.03
-0.10 0.08 0.02
-0.42 5.68 72.88 -0.26 21.68
0.76 0.82 0.77 0.70 0.67
0.12
0.09
-0.21
3.84
- 1.72
0.05
11.54
0.77
-0.08
0.14
-0.07
4.58
-1.19
-0.04
0.04
-0.04
13.10
0.78
used party spending only, on the assumption that for Conservative and Alliance
this bears a similar relationship to their local activism as spending does for
Labour. Spending by each party is expressed as a percentage of the legally
allowed maximum in each constituency: CSP87PC reports Conservative spending
and ASP87PC Alliance spending.
The analysis is further complicated by the fact that the Alliance vote in 1987
was really a vote for two parties: the Liberals and the Social Democratic party
(SDP). They entered into an electoral pact prior to the 1983election, dividing
the seats between themselves. The pact was still in force in 1987. However, the
local resources available to each party differed. The Liberals, with their much
longer tradition and their relatively strong grassroots, tended to be relatively
well organized in their constituencies, whereas the SDP, which was formed only
in 1981 and which had a far larger national than local profile, was comparatively
weak in constituencies. Furthermore, the SDP had formed largely as a
result of defections from Labour. As a result, the two Alliance parties might
reasonably be expected to have had different impacts on flows to and from
Labour in 1987. To take account of this possibility, a dummy variable
controlling for which party contested the seat for the Alliance was also added
to the regression model: the variable (SDP) took a value of 1 where the Liberals
fought the seat, and zero where the candidate was from the SDP.
In the results, the dummy variable for which Alliance party fought in a seat
was significant for four out of seven flows (Table 4). Where the Alliance
candidate was a Liberal, flows from Labour to the Alliance were on average
1.44 percentage points higher than where the SDP stood. However, with a
Liberal candidate in place, Labour loyalty was 2.28 percentage points lower
than where the SDP fought the seat, and flows from the Alliance and from
minor parties and abstention to Labour were also reduced. Facing a Liberal
rather than an SDP candidate meant that Labour lost more votes to, and
gained fewer votes from, the centre party but it also had impacts on Labour
loyalty and on the party's ability to win over previous non-voters.
I Poliliwl Sludm Aruxnlion. 19%
Research Note 479
Labour campaign spending was significant in all the regression models and
was always in the expected direction. The efforts made by the other parties also
had an effect upon flows to and from Labour. The more the Conservatives
spent, for instance, the lower the rate of defections from Conservative and from
‘rest/non-voters’ to Labour, the greater the defections from Labour to Conservative
and the lower Labour’s loyalty rate. Conservative campaigning did not
affect flows between Labour and the Alliance, however. Equally, Alliance
spending had no impact upon flows between Labour and the Conservatives.
But higher spending by the centre party reduced Labour loyalty rates, increased
flows from Labour to the Alliance (and decreased flows in the reverse
direction), and adversely affected Labour’s ability to ‘compete’ with minor
parties and abstention.
The evidence Seems clear. Not only does Labour party constituency campaigning
have a significant impact upon the party’s performance but campaigns
mounted by the other parties can also do so. In particular, campaigning by a
party can reduce its losses to and increase its gains from Labour. But the efforts
of all parties have an effect upon Labour loyalty and upon the ability of Labour
to win over past abstainers. The local campaign is a genuine contest with real
consequences for vote shifts.
Conclusions
The analyses reported here reveal that campaign spending is indeed a useful
surrogate for campaign effort. The finding opens up avenues for future
research. Analysis of the effectivenessof constituency campaigning need not be
constrained by the absence of survey data for a large number of seats. By law,
every party must report how much it spent on its constituency battles. This
provides a ready, and apparently fruitful, source of data for psephologists.
More substantively, such data demonstrate the continued importance of the
local campaign. The efforts of party members in each constituency at election
time are far from being a mere ‘ritual’.26They do have real though small
consequences for election outcomes. Elections are not simply lost and won on
the national campaign. Particularly in close-fought contests, an effective and
energetic local campaign can make the difference between winning and losing.
Political parties ignore the work of their local members at their electoral peril.
(First submitted: 27 November 1992;finally accepted: 13 April 1993)
26 D. Kavanagh, Constituency Electioneering, p. 41.
& Polma1 StudinAssmalion. I994