I argue that the incumbent is more credible in making promises of local club goods than the challenger, because he can underscore his determination to realize a proposed project by making costly investments already at the time of the campaign. I hence expect the incumbent to turn to the strategy of making local promises more widely instead of only disseminating the national promises entailed in his party’s manifesto. Indeed, in many cases in which the incumbent John Dramani Mahama made a local promise he emphasized that work related to the promise was already ongoing or that it was to begin within a short delay. In a speech at the Pusiga constituency on November 10, 2012, cited on page 53, he promised residents in the constituency to connect them to the national electricity grid. In order to underline the credibility of this promise, he assured the audience that a memorandum of understanding with the Chinese government for the funding of the project had been signed. In addition, he drew people’s attention to the fact that some poles had already been delivered. The rally in Pusiga by no means stands out as an exceptional example as the following quote illustrates:
“All the communities in this area that are on the electrification project are going to get electricity. This is because we have secured the counterpart funding. [...] We have secured money from the Chinese government credit. We were to pay a 15% counterpart funding which we have done; I spoke to SADA and SADA has helped and so all the communities are going to get their lights. As president I am assuring you, the poles have started coming and you are going to get your lights very soon.”14
Figure 6.1: Rallies at the which candidates made local promises
(a) Incumbent’s rallies
(b) Challenger’s rallies
Notes:The black dots represent a rally event held in a constituency, the red dots indicate that the candidate made at least one local promise at at least one rally in the constituency.
Figure 6.1 maps the campaign rallies of the incumbent and the challenger across Ghana (black dots), and shows those rallies, where they made local promises (red dots). It is already evident from these descriptive statistics that the incumbent recurred to promising local club goods at many more of his rallies, than the challenger. In fact, while the challenger only made two promises of local club goods at two campaign rallies, the incumbent made as many as 41 promises at 20 campaign rallies, out of the 47 and 51 rallies recorded in the sample. Figure 6.1a shows the proportion of rallies at which the incumbent made local promises. The incumbent made local promises at 20 of his 50 rallies (40%). The proportion of rallies where the challenger promised local club goods, shown in Figure 6.1b, is as low as (4%) for the challenger. He only made local promises at 20 of his 47 campaign events. This difference in the use of local promises by the incumbent and the challenger is also illustrated in Figure 6.2a. The difference is not only substantial in size, but also statistically significant at the 1%-level, as reported in Table 6.4. This finding is robust to restricting the sample to speeches of full length, as shown in Table C.1 Appendix C.15
The same trend is evident when considering the average number of promises made per rally event which is .82 (+/- .19) for the incumbent and only .04 (+/- .03) for the challenger. See Figure 6.2b. The results concerning the determinants of the average number of promises made per rally are reported in Table 6.5.16
Figure 6.2: The use of local promises by the incumbent and the challenger
(a) Rallies with and without local promises
0 10 20 30 40 50 Number of rallies Challenger Incumbent
Local promises No local promises
(b) Average number of local promises per rally by the incumbent and the challenger
0 .2 .4 .6 .8
Number of local promises per rally
Challenger Incumbent
The WMW test suggests that this difference is also statistically significant at the 1%-level. While these results could be influenced by the differences in the parties’ programmatic or ideological profiles and not only by their incumbency status, this is unlikely to account for this striking difference in the use
15
As a number of observations of five and smaller per cell might distort results of the chi square test (e.g. Conover, 1971), I also ran a Fisher’s exact test. The difference remained significant at a p-value of 0.000.
of local promises.
In order to rule out the possibility that the variation in the use of local promises across the incumbent and the challenger is driven by other differences between the parties than the fact that one was in government and the other was in opposition, I would ideally compare the 2012 election campaign to the one in 2000 or in 2004, where the NPP was in government and the NDC was in opposition. In fact, the present study is the first attempt to systematically analyze the allocation of promises of local club goods. Consequently, there are no previous findings to compare the 2012 election campaign to. Evaluating another strategy where an incumbent candidate is similarly expected to have an advantage over the challenger, we can largely alleviate potential concerns. Such a case is the distribution of individual benefits. Concerning such clientelistic targeting of voters, incumbents are also expected to have an advantage over the challenger for similar reasons as in the case of local campaign promises. As they have access to state funds, they have more resources at hand to distribute during election campaigns than opposition parties. Hence they not only have more resources to distribute, but are also believed to be more credible in promising the delivery of private benefits after the election (e.g. Wantchekon, 2003). Indeed, a regional campaign manager of the opposition party emphasized how the ruling party made use of this advantage in the 2012 campaign, distributing benefits such as “money, soap,[...], lanterns, mattresses, roofages,”17and even giving out cars to party faithfuls:
“And you know then from then [from September 2012 onwards, four months to the election] they [the ruling NDC] are even buying cars for some ladies. Those who can convince their colleagues. Assuming you are somebody who a lot of people follows (sic!) you. So they buy you a car, so you can talk to the other people that if they vote for your party what they did to you, they can also do it for them. Because you, they are using as a big example, an [...] example. So if they follow that train, they follow that train. That disturb (sic!) us.”18
Asked, why his party did not distribute benefits in the same way, the campaign manager explained that his party had less money in the 2012 election than the ruling party. He alluded that his party had also more frequently relied on distributing benefits to individual voters in the 2008 campaigns when it was in government.
“We don’t have the money like they have. They are in government. Don’t forget this is the second time Nana Addo [NPP’s candidate] is going. But this time, you know, we don’t have a lot of money. We don’t have the money to match them, you see. [...] In the last campaign, in the last campaign I can say that Nana Addo have (sic!) used a lot of money, a lot of money.”19
17
Interview with a member of a regional campaign of the NPP, December 15, 2012.
18Interview with a member of a regional campaign of the NPP, December 15, 2012. 19Interview with a member of a regional campaign of the NPP, December 15, 2012.
This anecdotal evidence supports the assumption that parties adopt their strategy to whether they are in government or in opposition, and that the incumbent party has an advantage in using campaign strategies that involve the promising or distribution of benefits to small groups of the electorate and to individual voters.
Table 6.4: Hypothesis testing, proportion of rallies with local promises
Hypothesis Implication Significant Type of test Test statistic p-value
H1 Proportion of rallies with local Yes Chi2 17.65 (1) 0.000
promises incumbent > challenger
Proportion of rallies with local promises No 1.0870(2) 0.581 differs across levels of past support
H2 Rallies with local promises in NDC No Marascuilo 14.5 (dif), –
strongholds > swing constituencies 39 (cv)
Rallies with local promises in NDC No Marascuilo 14.5 (dif), –
strongholds > NPP strongholds 42 (cv)
Rallies with local promises if ethnic partisans No Chi2 0.4076 0.523 majority > ethnic partisans minority
H3 Rallies with local promises in NDC strongholds No Chi2 0.0154 (1) 0.901 if past turnout low > turnout high
Notes: Results are the difference in proportion of rallies with local promises across different values of the independent vari- ables. Values displayed in column 5 are Pearson’s Chi-values for Chi2 tests, degrees of freedom in brackets; differences (dif) and critical values (cv) for Marascuilo tests. Except for the test of the effect of candidate identity, where rallies of both candidates are studied, all tests are performed for campaign events by the incumbent.
These findings with the regard to the use of local promises by the incumbent and the challenger provide strong support for Hypothesis 1. Whether the causal mechanism at work is that the incumbent was more credible than the challenger in promising local club goods, will be tested in Chapter 7, which reports the results of the survey experiment. In the remainder of this chapter, I restrict the analysis of the use of local club goods to the incumbent, as the results, presented so far, show clearly that the challenger did not make use of local promises.
Table 6.5: Hypothesis testing, number of promises per rally
Hypothesis Implication IV Significant z-value p-value
H1 Promises per rally incumbent > challenger Candidate Yes 4.259 0.000
H2 Local promises per rally if Ethnic partisans No -0.528 0.5977
ethnic partisans majority > minority
Local promises per rally in NDC strongholds NDC stronghold No -0.818 0.4136 > swing const. or NPP strongholds
H3 Local promises in NDC strongholds if past Turnout No -0.343 0.7315 turnout low > past turnout high
Notes: Results of Wilcoxon-Mann-Whitney tests of the difference in proportion of rallies with local promises across different values of the independent variables. Except for the test of the effect of candidate identity, which uses the full sample, all tests are performed for the incumbent.
6.2.2 Did the incumbent focus local promises on his partisans?
I have posited that candidates should focus most of their promises of local club goods in areas where potential supporters are concentrated, because such promises are likely to be most credible and thus effective with these voters. In the following, I analyze to what extend the incumbent targeted con- stituencies where his partisans were concentrated with local promises. I first apply a direct measure of partisanship, reflecting how many times a constituency has been won by the NDC and the NPP over the 2000, 2004, and 2008 presidential elections. As in Chapter 5, strongholds of the NDC (NPP) are those where the NDC (NPP) has won a plurality in all these three elections. Swing constituencies are areas where not the same party has gained a plurality of votes across all these elections. Subsequently, as in Chapter 5, I also apply a more indirect measure of partisanship, based on the share of ethnic partisans living in the various constituencies.
Local promises in NDC and NPP strongholds, and in swing constituencies
Figure 6.3 maps rallies at which the incumbent made local promises, over levels of past support for the NDC party in the 2000, 2004, and 2008 presidential elections.
Figure 6.3: Rallies by the incumbent, over past support for the ruling NDC
Notes: The red dots represent local promises, the larger the dots, the more local promises the incumbent made at rallies in a constituency. Dark gray shading indicates ruling party strongholds. Dotted shading reflects swing constituencies. NPP strongholds are shaded white.
The map suggests that the incumbent targeted both his party’s strongholds (dark-gray shaded) with local promises and swing constituencies (dotted shading). As expected, he made least promises in op- position strongholds (white shaded). The incumbent made 41 promises of local club goods in total. Out of all the promises, he made 19 (46%) in his party’s strongholds, 11 (27%) in swing constituencies, and 11 (27%) in the rival party’s strongholds.
In line with Hypothesis 2, the incumbent did not only allocate most promises to his strongholds. Also, the proportion of rallies at which he made local promises is higher for constituencies that have consist- ently supported the NDC party in the past than for swing constituencies and opposition strongholds, as Figure 6.4a illustrates. He made local promises at 48% (11 out of 23) of the rallies held in ruling party strongholds, only at 33% (5 out of 15) of the rallies held in swing constituencies, and at 33% (4 out of 12) of the rallies organized in NPP strongholds. However, the overall difference in the proportion of rallies with local promises across these past levels of support does not reach conventional levels of stat- istical significance, as the results reported in Table 6.4 show. The same is true for the restricted sample, including only speeches of full length, as reported in Table C.1 in Appendix C. While it is in line with my argument that the incumbent made local promises at more of the rallies that were held in his strong- holds, than in other constituencies, I would have expected him to make least promises in the opposition strongholds. Instead, he made local promises at as many of his rallies held in swing constituencies as at rallies hosted in opposition strongholds. This is likely due to the fact that he avoided campaigning in opposition strongholds in the first place, which the findings reported in Chapter 5 show. In the few instances where the incumbent did host rallies in NPP strongholds, the campaign might have regarded it as particularly important to offer some local public goods to the voters living there. In doing so, the incumbent’s aim most likely was to avoid to activate voters’ predispositions to vote for the opposition NPP. Such an “activation” (Lazarsfeld, Berelson and Gaudet, 1968) might have been the consequence of the incumbent campaigning in NPP strongholds without convincing these opposed voters with local promises. I turn to this point again below.
The Marascuilo test on the difference in the proportion between NDC strongholds and swing constitu- encies and between these strongholds and opposition strongholds paints a similar picture. According to this test, the proportion of rallies where the incumbent made local promises is indeed 14.5 percentage points higher than in swing constituencies. However, this difference does not cross the critical value of 39 percentage points defined by the test and hence does not reach statistical significance. The test suggests that the proportion of rallies with local promises which the incumbent held in NDC strong- holds was 14.5 percentage points higher than that in NPP strongholds. Again, this difference does not reach the critical value of 42 percentage points, defined by the test. The same is true for the comparison
between swing constituencies and opposition strongholds, as reported in Table 6.4. The results are ro- bust to restricting the sample to those speeches that are recorded full length, as shown in Table C.1 in Appendix C.
Concerning the average number of local promises the incumbent made across NDC and NPP strong- holds and swing constituencies consider Figure 6.4b. While the incumbent made most of his promises to his strongholds in absolute terms, the average number of promises per rally is highest for strongholds of the rival party. This unexpected finding is not that surprising at a second look. While the incumbent seems to have avoided holding rallies at opposition strongholds – by holding only 12 rally events out of the 50 rallies in the sample in such constituencies – once he visited such hostile crowds, he might have been compelled to promise his rival’s supporters some local projects.
Figure 6.4: Allocation of local promises by the incumbent across past support for the ruling party
(a) Rallies with and without local promises
0 5 10 15 20 25 Number of rallies
NPP stronghold Swing NDC stronghold
Local promises No local promises
(b) Average number of local promises per rally
0 .2 .4 .6 .8 1
Number of local promises per rally
NPP Stronghold Swing NDC Stronghold
A member of a regional campaign team of the ruling party explained that he tries to avoid sending the president to areas where the party does not intend to make a local promise and more importantly, start delivering on it before the election. Asked why the campaign in the region of Greater Accra had focused so much on holding rallies in strongholds, and if his party had traveled more to swing constituencies, had it had more time,20he provided the following answer:
“Yes, we would have gone to swing states more than necessary. We would have visited, whatever would be their demands – we didn’t have time on our side. Because when you go, they will make demands. We are in government and we have to make sure that those things are addressed, before you go again. Because if you go again, you came, we asked, we said this, this is no good. At least we need to be able to do some, then you can go back.”21
20
As mentioned in Chapter 4, the original candidate, late president John Evans Atta Mills passed away unexpectedly on 25 July 2012, so that the NDC began its official campaigns only on October 5, 2012. Not only did the party have to endorse a new candidate, but also respect a period of mourning, before officially entering the campaigning period.
The strategy this campaign manager describes is similar to the attempt of avoiding to polarize an un- friendly crowd against supporting the party that is campaigning, as Fenno (1978) describes for US election campaigns. The finding that the number of promises per rally held in opposition strongholds is relatively high should not be overemphasized, however, as the Kruskal-Wallis test does not suggest that the difference between the average number of local promises made across these different types of constituencies is statistically significant. The same is true for the sample restricted to speeches recorded in full length, as reported in Table 6.5 and in Table C.2 in Appendix C.
Did the incumbent target his ethnic partisans with local promises?
Besides using the direct measure of support for the incumbent to assess its effect on the local promises he makes, I use a more indirect proxy. Here, I investigate to what degree the concentration of ethnic partisans affected where the incumbent promised local club goods. The classification of the Ewe, the Ga and the speakers of the Northern languages, as partisans of the ruling NDC, is the same as in Chapter 5. The results of the chi square test, reported in Table 6.4, show that there is no significant difference in the proportion of rallies with local promises between constituencies which are dominated by the incumbent’s ethnic partisans and others (chi value of -0.4076, p-value of 0.523). I neither find a difference in the number of promises the incumbent made at those of his rallies held in localities where his ethnic partisans were in the majority compared to those where they constituted a minority. The result is reported in Table