Methodology
By its nature, the extent of the garrison state cannot be observed directly. The actions and influence of what Lasswell called "specialists on violence" often occur behind the scenes and are deliberately undocumented. Many written materials are classified and many of the relevant officials from earlier eras are retired or dead. These facts render problematic any attempt to study the garrison state by means of archival or interview strategies. They also make it close to impossible to use budgetary data as a proxy for the size of the garrison state (in most countries, for example, the budget of intelligence agencies is deliberately undisclosed or dissimulated among dozens of other line items).
The Garrison State Project uses an alternative methodology. We start from the observation that in democracies, every state agency's budget (whether to pay employees, or to buy equipment, or to carry out operations) has to be authorized by parliament. Moreover, parliaments have at least the formal authority of forbidding certain types of operations, as well as of engaging in oversight of them. Of course, that authority is more often than not unexercised, just as the budgets are hidden. Nonetheless, for the garrison state to develop over a long period of time, a consensus on the part of members of parliament must develop that certain activities, and the budgets for those activities, are correct. This implies that over time, MPs will exhibit greater agreement on national security-related issues than at earlier times, and that that trend is more marked than for other, non-national security-related issues, such as development assistance. We therefore look at the differential trends in consensus, on the two sorts of issues, within each of the 7 countries in our sample of democracies.
But how to look for consensus? The usual method for studying legislative agreement -- cross-party alignment on votes -- is not open to us: first, many important issues are precisely not voted on, and second, it is possible for MPs to vote the same way on a given issue without agreeing on the broader questions, or for them to vote on opposite sides on a given issue while agreeing on those broader issues. What we want, instead, is a sense of the reasoning used by MPs: why they are voting in a certain way. Hence, we look at speeches in particularly acrimonious debates, coding each speech for the (many) reasons put forward by the speaker in favor of his/her position, as well as the ways in which those reasons are chained together ("vote for the bill because of A, and A is true because of B"). These reasons may be blatant falsehoods or non sequiturs, but to the extent that speakers both within a side and across sides use the same reasons and chains, there is agreement. If that agreement grows over time, in the differential way referred to above, we can infer that there is consensus among elites on the garrison state.