Fiscal conservatism in inaugural speeches
(Imbeau, Louis M. 2009. «Dissonance in Fiscal Policy : A Power Approach». In L.M. Imbeau (ed.) Do They Walk Like They Talk? Dissonance in Policy Processes. New-York/Berlin: Springer, pp. 178-181.)
Three speeches mark the budgetary process at the provincial level: the inaugural speech in which the premier introduces the legislative program of his or her government at the beginning of every parliamentary session, the budget speech in which the minister of Finance presents the details of the government budget at the beginning of every fiscal year and several speeches in which ministers of various departments defend the government budget in committees of the legislative assembly. Each of the speakers delivering those speeches plays a precise role in the budgetary process following Wildavsky’s theory (1964; 1988). The minister of Finance is the guardian of the treasury while ministers of large program departments (mostly Health and Education at the provincial level) play the role of advocates of program and, consequently, they are advocates of spending. The premier plays the role of an arbiter who may sometimes support the guardians, sometimes the spenders, so as to influence the decisions of his cabinet in the direction that he prefers. Following Allison’s famous maxim «Where you stand depends on where you sit» (Allison 1969: 711), it is assumed that guardians support policy positions that systematically differ from those supported by program advocates. Thus minister of Finance tends to support fiscal conservatism, emphasizing restraint and control while ministers of Health or Education focus more on programs and therefore express more fiscal liberalism. The premier’s speech fluctuates between the two, now conservative, now liberal, thus accompanying the actions of his government as regards fiscal discipline.
For the purpose of this empirical illustration the issue is to assess to what extent the premier speaks like his minister of Finance or, conversely, like his ministers of Health or Education. Therefore we want to know if the inaugural speech is closer to the budget speech delivered by the minister of Finance than to the remarks made by ministers of Health or Education in parliamentary committees. It is assumed that the closer an inaugural speech is to the budget speech, the more conservative it is; and the closer it is to speeches by Health or Education ministers, the more liberal it is.
To assess the distance between the premier’s speech and that of his ministers, I used the Wordscores technique developed by Laver, Benoit and Garry (2003). This content analysis method compares the vocabulary used in various speeches in order to determine their respective position on a given continuum. Here we are interested in assessing the position of inaugural speeches on a continuum ranging from fiscal liberalism to fiscal conservatism. Speeches by Education or Health ministers represent the liberalism end of the continuum while the budget speech represents the conservatism end. These are the «reference texts» whose policy positions are determined a priori. The inaugural speech is the «virgin text» to be compared to the reference texts in order to determine its position relative to them:
Paraphrasing Laver and his colleagues (2003: 313), let us say that all we know about the inaugural speeches are the words that they contain. We compare those words to those we find in speeches of which we «know» the position on the liberalism-conservatism scale. The inaugural speech delivered at the beginning of a session is thus compared to the budget speech (arbitrarily coded +1) and to the budget remarks of ministers of Education or Health (coded -1) delivered in the same session. A computer program gives each word a score between -1 and +1 according to its relative frequency in the reference texts. For example, if the word «deficit» appears 10 times in a 1000-word speech delivered by the Health minister, and 90 times in a Budget speech of equal length, it is given a score of 0.08 (that is, 0.01*-1 + 0.09*1). Then, if the same word appears 5 times in a 1000-word inaugural speech, it gets a score of 0.0004 (that is, 0.08*0.005). Adding these scores for each non-unique word found in the inaugural speech, we get a conservatism score for that speech.
Laver and his colleagues propose to consider the issue in another perspective. On the basis of the frequency distribution of each word in the reference texts, we can estimate the probability of reading one reference text while reading a given non-unique word in the virgin text. In the example above, we know that the probability that we are reading the Budget speech while reading the word «deficit» is 0.9. If we assign a score of +1 to the Budget speech and -1 to the Health or Education speeches, it is logical to give the virgin text we are reading a score of 0.8 each time we read the word «deficit». After doing this for every non-unique word in the reference text, we divide the sum of these scores by the number words. This mean corresponds to the conservatism score of the text.
 Wildavsky’s work is about the American budgetary process. For an application on Canada at the federal level, see Savoie 1990 and, at the provincial level, see Imbeau 2000.
 For an empirical test of this proposition, see Imbeau 2006.