When no single party obtains the absolute majority of seats it is usually the case that various parties sign an agreement either to form a coalition government or to commit to a stable legislative coalition. However, we still know little about the relationship between the policies contained in these agreements and the preferences of the signatory parties. The few works having explored this issue have used party manifestos to assess how parties influence the policy output of government as expressed in the coalition agreement or government declaration. Yet, the evidence provided has been, at best, mixed. This paper tries to make a twofold contribution to this literature. First, unlike other studies that rely on other techniques and make commitments on policy or ideological positions (such as expert codings), we defend the use of a policy-blind scaling method to automatically evaluate how similar/different are the documents measuring party preferences and the coalition agreement, and obtain the distances between them. Second, in addition to party manifestos, we also use parliamentary speeches in the investiture debate to capture the preferences of parties. Using data from the Spanish regional coalition governments between 2005 and 2011, we find that i) the use of an automatic scaling method does lead to the emergence of significant relationships between party preferences and coalition agreements, and ii) the use of parliamentary speeches proves a fruitful strategy to underscore the relevance of investiture speeches in the study of coalition dynamics. In particular, we show that being a member of the coalition instead of remaining in the opposition has a decisive influence to lower the distance between the preference of the party and the content of the coalition agreement. Finally, we find that there is a specially strong connection between the candidate speech in the investiture debate and the coalition agreement, which offers promising avenues for further research.