Introduction
The fulfillment of campaign promises is at the heart of democratic accountability. If parties are to be responsive to societal demands, there should be a substantial level of congruence between the policies they implement in government and the promises found in their election platforms. A strong program-to-policy linkage is central to the mandate theory of democracy and the responsible party model (Downs, Reference Downs1957; Manin, Reference Manin1997; Mansbridge, Reference Mansbridge2003). Whether parties keep their campaign promises is not just a theoretical issue. The question has become an important aspect of the general political debate both during and between elections. Elected leaders often claim to have a mandate to carry out their election programs, and their specific campaign promises receive considerable media attention during and between election campaigns.
The fulfillment of election promises in Canada has been the subject of published scholarship in the past (Flynn, Reference Flynn2011; Monière, Reference Monière1988; Rallings, Reference Rallings, Budge, Robertson and Hearl1987). Some of these studies date back several decades and their methodologies differ so their findings are difficult to compare. This paper deploys the methodology now commonly in use in more recent studies of fulfilling campaign promises in a comparative perspective under the auspices of the Comparative Party Pledge Group (CPPG) (Artés, Reference Artés2013; Artés and Bustos, Reference Joaquin and Bustos2008; Costello and Thomson, Reference Costello and Thomson2008; Håkansson and Naurin, Reference Håkansson and Naurin2016; Kostadinova, Reference Kostadinova2013; Mansergh and Thomson, Reference Mansergh and Thomson2007; Moury and Fernandés, Reference Moury and Fernandés2016; Naurin, Reference Naurin2014; Pétry and Duval, Reference François and Duval2015; Praprotnick, Reference Katrin2017; Thomson et al., Reference Thomson, Royed, Naurin, Artés, Costello, Ennser-Jedenastik, Ferguson, Kostadinova, Moury, Pétry and Praprotnik2017; Toros, Reference Toros2015). Our findings can therefore be compared with their findings without loss of validity. Specifically, we measure the fulfillment of campaign promises by seven successive Canadian governments from 1993 to 2015.
This study not only replicates existing findings on the determinants of governments fulfilling campaign promises in other countries, it also extends them by analyzing two determinants that have not been examined in recently published studies, namely, the ideology of the party in office and the state of the budget. We also innovate by tracking the point during a mandate at which each promise is fulfilled. We track time because we expect that the effect of our predictor variables is not randomly distributed through time but that it will vary depending on the moment at which a promise is declared fulfilled. Traditional methods of logistic and linear regression are not well suited to accommodate time. That is why we use a discrete time duration—more specifically a binary time-series-cross-section (BTSCS)—approach to promise keeping. The approach includes a so-called “time spell” variable, which counts the number of quarters it takes from the start of a mandate until the campaign promise is fulfilled. In other words, we are looking at the “status” (fulfilled against unfulfilled) of each promise at any given time, in this case each quarter, for the duration of a mandate.
Part 1 reviews the theory behind the various factors expected to influence the fulfilling of a campaign promise. Part 2 reviews the method used to record party promises and the fulfillment data and explains the BTSCS scheme that we use to analyze these data. Part 3 analyses the multivariate effects of our explanatory factors on fulfilling campaign promises. We show that these effects vary substantially depending on the time at which a promise is declared kept. The conclusion highlights the main findings on keeping campaign promises in Canada and their implications and proposes some avenues for future research.
Theory and Hypotheses
We examine the theory behind seven distinct expectations about keeping campaign promises. We expect first that keeping a campaign promise is influenced by the type of change proposed by each promise. Promises of change may be more difficult to fulfill than promises to maintain the status quo, given the incrementalism of large governments in general. There is evidence from past studies that promises to initiate change are less likely to be fulfilled than pledges to maintain the status quo (Costello and Thomson, Reference Costello and Thomson2008; Kostadinova, Reference Kostadinova2013; Mansergh and Thomson, Reference Mansergh and Thomson2007; Royed, Reference Royed1996). This is true except for promises to increase public spending and/or cut taxes, which are kept at a higher rate than other change-oriented ones. Therefore, it is expected that promises to increase public spending and/or cut taxes are more likely to be kept than other promises to initiate change. The variable for the type of proposed change has four categories other than to maintain the status quo (the reference category): 1) to increase government spending and/or to increase taxes, 2) to reduce government spending and/or to cut taxes, 3) to introduce other non-budget related changes, and 4) to achieve any type of policy outcome not related to government spending or taxation (for example, to create jobs). These four factors are expected to decrease the likelihood of keeping campaign promises relative to promises to maintain the status quo, although the odds of fulfilling promises to increase government spending are expected to be comparatively similar to the odds of fulfilling status quo pledges.
Inter party agreement
When there is agreement between the governing party and the opposition on an issue, it indicates that a promise is probably a response to widely shared societal demands, rather than reflecting narrow partisan interests. It seems likely that a government will find less parliamentary opposition when time comes to enact these sorts of promises, which could be particularly important for minority governments. Previous research has found that campaign promises by the party in government are more likely to be kept when they are similar to promises made by opposition parties (Naurin, Reference Naurin2014; Thomson et al., Reference Thomson, Royed, Naurin, Artés, Costello, Ennser-Jedenastik, Ferguson, Kostadinova, Moury, Pétry and Praprotnik2017). We expect agreement among parties to increase the likelihood of fulfilling a campaign promise. Inter party agreement is a binary coded 1 when the governing party agrees with at least one party and 0 when there is no agreement or disagreement. Agreement is measured between the electoral pledges of the Liberal party (LPC), the Conservative party (CPC) and the New Democratic party (NDP).
Majority government
According to conventional wisdom, minority governments have less capacity to make policy in general and less freedom to carry out their campaign promises than majority governments. It is true that minority governments must compromise with one or more opposition parties in order to govern, something that is not required of majority governments. However, the argument that the lack of freedom incapacitates minority governments is overstated (Strøm, Reference Strøm1990). Godbout and Høyland (Reference Godbout and Høyland2011) show that even in Canada where single-party governments are the rule, informal voting coalitions are a common occurrence, allowing minority governments to conduct their business effectively. There is evidence that single-party minority governments are capable of keeping as many if not more campaign promises as single-party majority governments while they last (Artés and Bustos, Reference Joaquin and Bustos2008; Mansergh and Thomson, Reference Mansergh and Thomson2007). However, the fact remains that minority governments appear to fulfill fewer campaign promises than majority governments because majority governments last longer (Pétry and Birch, Reference Pétry and Birch2016). Therefore, it is expected that majority governments fulfill more promises than minority governments, although this expectation is derived more from considerations of government duration than of freedom of action. The majority government variable is a binary coded 1 for majority governments and 0 for minority governments.
Newly elected (first mandate) government
Newly elected governments are poorly positioned relative to re-elected ones to keep their campaign promises. This is due in part to the fact that re-elected governments have the opportunity to obtain public service advice and expertise in the process of drafting their election platforms and subsequently in the day-to-day operations of government (Flynn, Reference Flynn2011: 249). Re-elected governments may also take advantage of projects already in the legislative “pipeline” that died on the order paper when the election was called, an option that is normally not available to newly elected governments. These elements benefit re-elected parties once they return to government relative to newly elected governments (Pétry and Birch, Reference Pétry and Birch2016). The variable, a binary coded 1 for newly elected governments and 0 for re-elected ones, is expected to make keeping campaign promises less likely.
The budget
Governments’ capacity to deliver policy programs increases (decreases) as they have more (less) financial resources at their disposal (Huber and Stephens, Reference Huber and Stephens2001). Governments’ financial capacity to fulfill their election promises is particularly sensitive to changes in whether there is a budget surplus or deficit. One clear illustration is the first Liberal government of Jean Chrétien elected in 1993. Chrétien's “Red Book” election program contained several promises to expand social programs (child care, social security). However, the fact that the Liberal government was facing a budget deficit (a consequence of the economic recession of 1992) forced Jean Chrétien to renege on the fulfillment of several of these promises during his first mandate (Prince, Reference Prince, Harder and Patten2006). The state of the budget gives an indication of a government's financial capacity to fulfill its election promises. We expect that a surplus (or deficit) budget (expressed in $B) will increase (decrease) the likelihood of pledge fulfillment. The data for the state of the budget are drawn from a statscan series on Revenue, expenditure and budgetary balance - General government (Statistics Canada 2018). The variable is lagged one year because we expect that its effect will be delayed.
Interaction between type of change and party ideology
There is evidence from the public finance literature to show that centre/right parties are more likely to make and implement proposals to reduce public spending whereas centre/left parties are more likely to make and implement proposals to increase public spending (Cusack, Reference Cusack1997; Imbeau and Pétry, Reference Imbeau and Pétry2004). In Canada, the Conservative party is positioned to the right of the Liberal party on left-right scales based on expert surveys (Benoit and Laver, Reference Benoit and Laver2006; Pétry et al., Reference Pétry, Collette and Klingemann2012), on the content of party manifestos (Koop and Bittner, Reference Koop, Bittner, Bittner and Koop2013; Pétry, Reference Pétry, Erikson and Laycock2014) and on results from national election surveys (Cochrane, Reference Cochrane2015).
In line with this evidence, we expect that promises to reduce public spending are more likely to be fulfilled by the centre/right Conservatives whereas promises to increase public spending are more likely to be fulfilled by the left-of-centre Liberals.
Interaction between type of change and the state of the budget
As argued above, we expect that promises to increase public spending are more likely to be fulfilled than promises to reduce public spending. This, however, poses the question as to whether the differential effects of these two types of promises interact with the budget throughout a government's mandate. It is plausible to assume that a surplus in the budget would be positively associated with the fulfillment of promises to increase public spending but negatively associated with keeping a promise to lower public spending. It is therefore expected that the presence of a favourable budget will increase the likelihood of keeping a promise to increase government spending but it will decrease the likelihood of fulfilling a promise to reduce government spending.
One word about endogeneity is in order before concluding the theoretical section. It might be argued that a loop of causality exists between fulfilling a promise and several of our explanatory factors. For example, if a rational party holds fulfilling its promises as its key goal, it would take into account the possibility of having only minority support when making campaign promises. Parties might make more modest promises if they do not expect to have a majority. Similar arguments can be made regarding the effect of a balance budget and agreement with opposition parties. On the other hand, rational parties may have an incentive to promise things that they know to have a slight probability of success (Davis and Ferrantino; Reference Davis and Ferrantino1996; Harrington, Reference Harrington1993; see also Aragonés et al., Reference Aragonès, Postlewaite and Palfrey2007). “Over-promising” might make sense in terms of creating negotiating space within the federal legislature or with other levels of government. In addition, promises have an important signaling function indicating commitment to key supporters during the campaign and to prospective ruling partners inside or outside government in case a party is elected. Calculating fulfillment odds may well take a back seat to such concerns.
Our main expectations regarding fulfillment of campaign promises by government parties are summarized in the following hypotheses.
H1 Type of proposed change. Promises to introduce change are less likely to be fulfilled than promises to maintain the status quo. Promises to increase public spending and cut taxes are more likely to be fulfilled than pledges to reduce public spending and to increase taxes.
H2 Inter party agreement. Government party commitments made in agreement with opposition are more likely to be fulfilled than other pledges.
H3 Majority government. Majority governments are more likely to fulfill election promises than minority governments.
H4 First mandate government. Governments in their first mandate are unlikely to fulfill as many pledges as re-elected governments.
H5 Balanced budget. Governing parties are more likely to fulfill their campaign promises when they can expect a balanced budget.
H6 Type of change × party ideology. The Conservatives are more likely to fulfill promises to reduce government spending; the Liberals are more likely to fulfill promises to increase government spending.
H7 Type of change × a balanced budget. A balanced budget is positively associated with increasing government spending and negatively associated with government spending reduction.
Data and Method
This paper analyzes the records of fulfilling campaign promises by the Liberal majority governments of PM Jean Chrétien, elected in 1993, 1997 and 2000 respectively, the Liberal minority government of PM Paul Martin, elected in 2004, the Conservative minority governments of PM Stephen Harper, elected in 2006 and 2008, and Stephen Harper's majority government, elected in 2011. First, we manually recorded individual election promises contained in officially sanctioned election programs. Election promises are defined as statements that contain unequivocal support for proposed government policy actions or outcomes that are testable (Thomson, Reference Robert2001). All members of the CPPG employ this definition (see Thomson et al., Reference Thomson, Royed, Naurin, Artés, Costello, Ennser-Jedenastik, Ferguson, Kostadinova, Moury, Pétry and Praprotnik2017). The “unequivocal support” part of the definition stipulates that a statement must imply an explicit promise to do something. Some language is too soft to be considered “unequivocal support.” Statements in which parties promise to “consider” or “look into” specific policy actions do not qualify as promises under this definition. The second part of the definition stipulates that promises contain “proposed government policy actions or outcomes” that are testable. This requires that the program writer, not the researcher, provide the criteria used to judge the fulfillment of promises.Footnote 1
The coding scheme is the same as the CPPG's. At first, pledges are classified in three categories: “fulfilled,” “fulfilled in part” and “not fulfilled.” To be classified as kept, a promise has to be followed by a subsequent government action (a law, a regulation, a treaty or an agreement) that has been passed. A promise is rated “fulfilled in part” when the corresponding action is a compromise or it is in the process of being completed. A promise is classified as “not fulfilled” when it is not followed by a government action. Those categories are then regrouped under “at least fulfilled in part” [1] or “not fulfilled” [0] for analysis purposes as well as comparability with other members of the CPPG. All of this is done manually for every government in the study. Promises are classified and matched with government actions by teams of trained graduate research assistants. Intercoder reliability tests are conducted on the identification of promises, on their classification as “fulfilled” or “not fulfilled” on the type of change proposed and on the timing of the verdicts. Intercoder agreement varies from a low of 82 per cent on the identification of what constitutes a promise to a high of 96 per cent on the type of change proposed. Each verdict is supported by one or more citations drawn from government press releases, bills, official documents and media sources. The reliability of coding increases with recency, as more recent verdicts can be supported by more sources.
Previous studies implicitly assume that fulfilling campaign promises is a uniform (or random) process over time. However, a uniformly distributed fulfillment probability over time seems unlikely, if only because sooner or later diminishing returns set in. Many (although by no means all) campaign promises that are fulfilled early in a government mandate tend to be the ones that are “easy” to keep, the proverbial low-hanging fruit. Promises not fulfilled by the middle of a mandate are the most “difficult” ones to keep; they have a lesser likelihood of being kept before the end of the mandate precisely because they touch on complex or controversial issues or because they face serious political and administrative obstacles (Duval and Pétry, Reference Dominic and Pétry2018). This leads us to expect that the probability of a promise being fulfilled through time is higher at first and then decreases either constantly or even faster as the end of the mandate draws near.
Following Beck and colleagues (Reference Beck, Katz and Tucker1998), we model the fulfillment of campaign promises under a time-series-cross-section analysis with a binary dependent variable (BTSCS). In essence, modeling the fulfillment of promises this way allows us to show not only when a promise is fulfilled but also when the series “switches” from not fulfilled to fulfilled, how long this takes and what covariates are associated with this change. Following Carter and Signorino's recommendation (Reference Carter and Signorino2010), we used a cubic polynomial approximation to model time dependence. That method consists in including a time spell variable, the same variable squared and again cubed (ts, ts2, ts3). The time spell is a count of the number of quarters since the date at which each successive mandate started or since the fulfillment of a promise occurred, whichever is most recent. In other words, we are looking at the “status” (fulfilled against unfulfilled) of each promise at any given time, in this case each quarter, for the duration of a mandate.Footnote 2
Findings
The dependent variable is a binary for the likelihood that a campaign promise is fulfilled at least in part (fulfilled at least in part = 1; unfulfilled = 0). Out of a total of 828 promises recorded over the period, 565 (68.2%) were at least partly fulfilled and 253 (31.8%) remained unfulfilled. These numbers are a bit lower but still comparable with the proportions found by Monière (Reference Monière1988) and by Rallings (Reference Rallings, Budge, Robertson and Hearl1987) for Canadian governments of the past (74% and 72% respectively).Footnote 3 The Liberals made 392 promises in the four elections between 1993 and 2004, and fulfilled 252 of them at least in part (average fulfillment rate of 64.3%). The Conservatives made 436 campaign promises in the three elections between 2006 and 2011, and fulfilled 313 of them at least in part (average fulfilment rate of 72.5%).
Figure 1 displays the proportion of promises fulfilled at least in part by each successive governing party between 1993 and 2015. The rate of promises kept by the Liberals has changed from 53 per cent for the government elected in 1993 to 66 per cent in 1997 to 77 per cent in 2000 and 72 per cent in 2004. The rate for the Conservatives has changed from 66 percent in 2006 to 62 per cent in 2008, and 84 per cent in 2011.Footnote 4 The low Liberal score in 1993 is an outlier. Its occurrence will become clear once we have a better understanding of the factors that shape the likelihood of promise fulfilment. The average promise keeping score of the CPC (72.5%) is significantly higher than that of the LPC (64.3%). But this has less to do with party ideology than with a trend toward an increase over time in the number of promises kept by Canadian governments. Regressing the average percentage of promises kept per month in office by each government against a counter produces a linear slope of 0.21, meaning that fulfilling promises by each successive government increases by 2.1 percent on average. Note finally the small percentage of promises kept in part for all the governments in the sample. A similar pattern of comparatively low frequency of partially fulfilled promises is observed in all the countries for which we have data (Thomson et al., Reference Thomson, Royed, Naurin, Artés, Costello, Ennser-Jedenastik, Ferguson, Kostadinova, Moury, Pétry and Praprotnik2017).

Figure 1 Pledges Fulfilled at Least in Part (1993–2015).
The multivariate model is constructed based on three successive blocks of explanatory variables with each block entered sequentially in the analysis. Model 1 includes variables for the type of proposed change, inter party agreement, majority governments, first mandate governments, balanced budget and Conservative party ideology. Models 2 and 3 include additional variables for interaction effects between type of change and party ideology and between type of change and state of budget, respectively. Table 1 displays the odds ratios and z-scores (in parentheses). Table A1 in the appendix displays the logit regression coefficients and confidence intervals in parentheses, including those for the time spell variables. Table A2 displays the results for the model without the time variables (N = 828) and Table A3 displays the results with the state of the budget expressed in percentage of GDP rather than in $B.Footnote 5
Table 1 Determinants of Party Promise Fulfillment in Canada 1993–2015 (odds ratios with z-scores in parentheses)

Note: *p < 0.05; **p < 0.01; ***p < 0.001. Odd ratios with z statistics in parentheses. Approximation terms (ts, ts2, ts3) and the constant are omitted from this table.
The first model in Table 1 includes the variables related to the type of change that each campaign promise proposes (H1) and the variable for interparty agreement (H2). H1 predicts that promises to introduce a change will have a smaller likelihood of fulfillment than promises to maintain the status quo (the reference category). H1 also predicts that the likelihood of fulfillment of promises to increase public spending and cut taxes will be higher than the likelihood of keeping promises to reduce public spending. Figure 2 displays the number and the rate of fulfillment of promises by type. The data partly support H1: Fulfillment is 25 per cent more likely for status quo pledges (84.5%) than for promises to make a change (67.3%). However, contrary to expectation, the rate of keeping promises to increase public spending (74%) is not statistically different from the rate for pledges to reduce spending (73%). Looking at the results in the multivariate setting of Table 1, the relevant odds ratios are significantly lower than 1 in model 1, indicating that, other things being equal, the likelihood of keeping promises to initiate a policy change is significantly lower than the likelihood of maintaining the status quo. The prediction that inter party agreement increases the likelihood of keeping promises (H2) is supported by the data. Inter party agreement approximately doubles the odds of fulfilling promises.

Figure 2 Pledge Fulfillment by Type (1993–2015).
Having a majority increases the odds of fulfilling promises by 300 per cent as expected (H3). The odds ratio for the first mandate variable is significantly smaller than 1, indicating that newly elected governments are less likely than re-elected ones to fulfill their campaign promises as predicted by H4.Footnote 6 A budget surplus significantly increases the odds of keeping promises. A $1 billion increase in last year's budget coincides with a 14 per cent increase in the odds of fulfilling campaign promises. This highlights the importance of having a favourable financial position in budget year t-1 for keeping promises in year t.
The odds of fulfilling campaign promises increase threefold when the Conservatives are in power. One should not read too much in this result, however. Indeed, the odds of keeping promises when the Conservatives are in power are considerably reduced in Model 2 (and they are no longer statistically significant) after the term for the interactions between the type of promise and the party in power has been added to the equation. The high odds associated with the Conservatives in power in Model 1 are the spurious effect of a trend toward fulfilling larger percentages of promises over time.
Model 2 adds variables for the interactions between the Conservatives in power and the type of promise kept. This model shows what type of promise is most likely to be kept by the Conservatives and how this compares with the type of promise most likely to be kept by the Liberals. The expectation is that, faithful to its centre/right conservative ideology, the CPC has favoured keeping promises to reduce government spending and to cut taxes. The expectation is supported by the data. The multiplicative part of the model indicates a sixfold increase in the odds of fulfilling promises to reduce government spending when the Conservatives are in power. Note also that the odds ratio for the Conservative party variable in the additive part of model 2 is not statistically significant. The results of Model 2 support the expectation that promises kept by the Conservatives have differed from the promises kept by the Liberals. In particular, the odds of fulfilling promises to reduce government spending are much higher under the Conservatives than under the Liberals. This is in line with the expectation that the different patterns of keeping promises by the Liberals and the Conservatives are due in part to ideological differences between them.
Do variations in budget surplus or deficit affect promise keeping? In the affirmative, we would expect that the presence of a budget surplus decreases the likelihood of fulfilling pledges to reduce government spending while increasing the likelihood of fulfilling promises to increase government spending. Model 3 tests these expectations by adding variables for the interaction between the state of the budget and type of promises made. The odds ratio relating to the state of the budget with a reduction of public spending supports the first expectation. However, the odds ratio relating to the state of the budget with expansion of public spending fails the statistical significance test, indicating that a surplus budget does not increase the odds of fulfilling promises to increase spending beyond those observed in the additive part of the model. Note that a budget surplus does not affect the odds of fulfilling promises in the “other” category. This is in contrast with the highly positive impact of the CPC in power on the odds of fulfilling promises in the “other” category as displayed in Model 3.
Our multivariate model helps elucidate the unusually low promise keeping “score” (53%) of the first Chrétien government. Even though this government had a majority, it was the only one in our sample under the simultaneous influence of two variables associated with lower odds of keeping promises in the model: 1) a budget deficit and 2) a newly elected government. The second-lowest score (62%) belongs, not surprisingly, to the second government of Stephen Harper, a minority government facing substantial budget deficits. Another factor impacting the percentage of promises kept by the second Harper government was the hyper-partisanship in Parliament which made it difficult for the government to find support for its legislative agenda across the aisle. It has been claimed that the hyper-partisan parliamentary environment was largely of Prime Minister Harper's own doing (Martin, Reference Martin2010; Wells, Reference Wells2014). The Conservatives’ first act after winning the 2008 election was to eliminate the government subsidy to political parties, a measure intended to cripple their opponents’ party machines. The government prorogued Parliament twice during the period before it was toppled by a non-confidence vote in March 2011.
Our models are predicated on the assumption that the probability of a campaign promise being kept takes the form of a decreasing function (f ’ < 0). Figure 3 displays a series of diagrams intended to test the validity of that assumption one predictor at a time. The bar graph of Figure 3.1 shows the variation in the rate of promises kept over the mandate. Clearly, the fulfilling of promises during a mandate is not uniform. Out of 828 campaign promises in the sample under study, 282 (34.5%) were fulfilled in the first 12 months of their respective government; out of the remaining 542 promises unfulfilled after one year, 165 (30.4%) were fulfilled in the second year; out of the 377 promises left unfulfilled after two years, 86 (22.8%) were fulfilled in the third year and 29 (10%) of the 291 promises left unfulfilled after three years were fulfilled after that. The bars in Figure 3.1 show an interesting pattern of periodic alternations of ups and downs, which clearly corresponds to the cycle of annual budgets. There are also unusually high levels of promises kept at the very beginning of a mandate.

Figure 3 Effect of Time on Probability of Pledge Fulfillment.
The plot of Figure 3.2 displays the impact of time on the probability of fulfilling promises of different types. Overall, promises to increase public spending and to cut taxes are more likely to be fulfilled than promises to reduce public spending and to increase taxes. However, the decrease in probability over time is so large that the probability of fulfilling promises to increase spending at the end of a mandate (0.65) is substantially lower than the probability of fulfilling promises to reduce spending at the beginning (0.86). Figures 3.3, 3.4, 3.5, and 3.6 illustrate how time affects the impact of inter party agreement of a majority government in its first mandate and of party ideology on the probability of fulfilling promises, respectively. Although there are some differences, all diagrams display similar probability distributions in the shape of a slanted S, dropping first at a slow rate, then at a faster rate and finally shifting back to a slow drop rate as the end of the mandate draws near. For example, in diagram 3.6, the predicted probability of a Conservative promise being kept hovers around 0.95 for the first five quarters, then drops to 0.87 during the next five quarters finally settling at 0.52 at the last time point (Q17). The predicted probability of a Liberal promise being kept starts at 0.89 at the beginning of a mandate and drops by 0.62 during the length of a full theoretical mandate to end at 0.27 at the last point (Q17).
Figure 4 tracks changes in the probability of promise keeping by the Conservatives and by the Liberals subject to the effect of time passing and the changing state of the budget. The comparison allows us to visualize the strength of the influence of party ideology on the likelihood of promise keeping over time. The probability for the Conservatives is higher than for the Liberals at the start and it stays higher until the end of a mandate. The change over time in the likelihood of promise keeping is also associated with the state of the budget. The hazard plots for both parties change shape as the state of the budget changes. The closer the budget is to deficit, the more likely the plot line assumes the shape of a slanted S. This suggests that a budget deficit encourages governments to fulfill promises without delay at the start of their mandate even if this means a more rapid drop in keeping promises later in the mandate.

Figure 4 Effect of Budget Balance and of Time on the Probability of Pledge Fulfillment.
Conclusion and Discussion
In conclusion, while highlighting our main findings, we examine some of the questions left unanswered by our analysis and we discuss the implications of our results for understanding the functioning of representative democracy in Canada.
All things remaining equal, the interaction of the ideology of the party in power with the type of campaign promise influences whether that promise is fulfilled. In particular, promises to reduce government spending and to initiate other changes are significantly more likely to be kept when the Conservative party is in power. Although statistically significant, the impact of party ideology on the type of promise kept is not very strong. Furthermore, the likelihood of fulfilling promises to increase public spending remains the same under Liberal and Conservative governments. This study only investigates whether parties with distinct ideologies keep different types of promises. We do not look at the relationship between the ideology of parties and those campaign promises. Whether and to what extent campaign promises reflect party values and ideology awaits further research.
The likelihood of promises being kept is affected by institutional factors. Minority and newly elected governments are significantly less likely to keep their election promises than majority or re-elected governments. This suggests that parties do not always adjust their campaign promises to what they expect the political situation will be after the election. Expecting to win a minority will not moderate a party's ambition during the campaign even if it means making adjustments in the Throne Speech at the opening of Parliament following the election. An opposition party does not adjust its campaign promises to the expectation that if it wins the election it will face the handicap of being a first-mandate government.
The likelihood of keeping campaign promises is also affected by the availability of financial resources. Budget surpluses have a significantly positive effect on the likelihood of keeping campaign promises. The effects are not only additive but also multiplicative. The likelihood of fulfilling promises to reduce public spending is even less likely when the previous budget is positive. The findings about the impact of minority and newly elected governments and the impact of the state of the budget on keeping campaign promises raise the question of whether parties make promises in accordance with a coherent ideology or whether they adjust their pledges to expected conditions. Further research is needed on this question in connection with the earlier discussion about endogeneity and over-promising.
This study uses the discrete time approach (Duval and Pétry, Reference Dominic and Pétry2018) to measure how the various determinants of keeping campaign promises are impacted by time. The evidence shows that, although there are some differences, the passage of time has a remarkably similar impact on the various determinants of keeping the promises that we analyze in our models. Promises are more likely to be kept at the start than at the end of a mandate, irrespective of the institutional structure of government. The likelihood of keeping different types of promises—whether to increase or reduce public spending or to make other changes—seems to be affected by similar patterns of diminishing returns. These findings are largely in line with results from the policy agenda approach which show that most policy changes occur soon after an election and then things tend to stabilize (Bevan et al., Reference Bevan, John and Jennings2011; Jennings et al., Reference Jennings, Bevan, Timmermans, Breeman, Brouard, Chaqués-Bonafont, Green-Pedersen, John, Mortensen and Palau2011; Mortensen et al., Reference Mortensen, Green-Pedersen, Breeman, Chaqués-Bonafont, Jennings, John, Palau and Timmermans2011). This contradicts the political budget cycle approach which holds that government policy outputs (at least those that are popular with the electorate) should accelerate as the next election draws closer (see Campbell and Garand, Reference Campbell and Garand2000, and Lewis-Beck and Stegmaier, Reference Lewis-Beck and Stegmaier2000, for reviews).
One unexpected finding of this study is the small but noticeable increase in the rate of promise keeping over time. The increase is reflected in the higher rate of promises kept by the Conservative party in the multivariate analyses (Models 1 and 3) of Table 1. This demonstrates that the trend is not due to an imbalance in the distribution of factors over time associated with a higher likelihood of promise keeping (for example, by majority governments, re-elected governments, surplus budgets). The trend in keeping more campaign promises over time is real, not the spurious effect of other factors unrelated to time, although we do not have enough cases in our sample to make a definitive affirmation. The suggestion is reinforced by the fact that after two and a half years in office, the Liberal government of Justin Trudeau elected in 2015 had fulfilled at least in part 81 per cent of its campaign promises.Footnote 7
What is the significance of keeping campaign promises in terms of evaluating democratic representation and efficient policy making in a Westminster system like Canada? The mandate model and the responsible party model (Downs, Reference Downs1957; Klingemann et al. Reference Klingemann, Hofferbert, Budge, Bergman, Keman, Pétry and Strøm1994, Manin, Reference Manin1997; Mansbrige, Reference Mansbridge2003) assume that only the promises of the winning party (or parties in a coalition government) are carried out by the subsequent government. However, there are other models of representation. In particular, the agenda model assumes that parties elected to office also carry out some of the promises of the opposition parties, and not just those of the winning party (Pétry Reference Pétry1995, Louwerse Reference Louwerse2011). The fact that the winning party fulfills a large percentage of its campaign promises, and that this percentage is growing over time is “good” from the viewpoint of the mandate model; but it is debatable whether it is always good from the viewpoint of the agenda model.Footnote 8 One should also keep in mind that the extent to which parties keep their promises is only one criterion of efficient policy making. The fact that some governments keep most of their promises does not necessarily mean that they are somehow “better” at policy making than other governments. The desire to keep campaign promises should not turn to an obsession; sometimes, as Schedler (Reference Schedler1998) reminds us, there are legitimate excuses for breaking your word.
Supplementary materials
To view supplementary material for this article, please visit https://doi.org/10.1017/S0008423918000379.