COIN 1.0 holds that establishing a set of observable measures is vital to assessing the effectiveness of various efforts. This is often easier said than done, however, as measures are fraught with inaccuracy and largely based on guesswork. For instance, the United States often measures stability by gauging levels of violence. In most cases, this is uninformative: U.S. commanders don't know whether violence is occurring because of their programs, because there are fewer victims, or because of a change in insurgent tactics. For example, in Afghanistan, civilian casualties by insurgents increase violence against coalition forces, leading many to conclude that protecting the population is critical to counterinsurgency success. However, civilian casualties by insurgents in Iraq actually resulted in less violence against coalition forces. Simply measuring levels of civilian casualties may be informative for policies in Afghanistan but under other conditions (such as those in Iraq) other more nuanced measures may be required.
The other common approach is to measure progress using operational output, conflating measuring the process of counterinsurgency activity with the impact of those activities. Process measures are most easily understood as the direct results of COIN programs. For example, a process measure for security-training programs may be the number of trained police. An assessment like this is clearly appealing -- it's easy to count -- but this approach is incomplete and often misleading. In this example, "more police" is not the desired end; "more police" increases public safety, which in turn generates government legitimacy -- but we do not know the number or quality of police required for public safety or what level of public safety is required for this increase in legitimacy to occur.
Additionally, the models and measures of progress in COIN 1.0 do not account for third-party problems. Simply stated, the United States usually fights insurgencies by supporting a host-nation government, which means it is constrained by that government's preferences. In many cases, this is a significant constraint, as U.S. officials have learned through their often-problematic partnerships with Iraqi Prime Minister Nouri al-Maliki and Afghan President Hamid Karzai.
This presents us with what social scientists call the "principal-agent problem." In this case, the United States serves the role of "principal" -- stripped of academic jargon, that means an actor that wants things done -- with a set of objectives related to international security. The host-nation government, as the "agent" (commissioned to act for the "principal") does not receive the full benefit of the objectives, but does face costs and risks for its efforts to conform to U.S. desires.
Broadly speaking, the United States has two options to induce the host-nation government to act in accordance with its objectives. First, it can try to convince the host nation that it should change its preferences. Secondly, it can change the incentives -- through positive (carrots) or negative (sticks) means.
In the international arena, the first approach very seldom works. Politicians in weak or failed states have typically risen to the top by knowing and exploiting local power dynamics, and they don't need an international power to inform them of their interests. The second approach, while perhaps more useful, has the risk of incentivizing the host nation to sustain conflict in order to be rewarded for resolving it. That is what social scientists call a moral hazard -- a government may be inclined to play the role of an arsonist in order to get hired for the job of firefighter. This moral hazard tends to exacerbate the problems with process measures. For example, do police still increase legitimacy if they are corrupt and foster instability, thus ensuring future aid and support?