Abstract
Forecasting the benefits of management interventions intended to improve ecological conditions requires a causal understanding of the factors that lead to system change. The causal attribution of a factor is defined as the difference between the outcome observed in the presence of the factor and the outcome that would have been observed in the factor’s absence, that is, the counterfactual condition. Estimating this contrast is relatively straightforward, where matched or randomized controls are available to approximate the counterfactual condition. However, researchers must reason retrospectively from observational data where matched or randomized controls are not available. In this case, the challenge of establishing causal attribution is in estimating the true counterfactual, that is, the outcome that would have resulted from the absence of the factor, given that it was present. Causal analysis permits the estimation of counterfactuals from observational data, assuming that the model captures all common causes between exposure and outcome, that the exposure is independent of other factors in the model (i.e., exogenous), and that the exposure causes the same directional change for all units (i.e., monotonic). I estimated retrospectively the causal attribution of habitat-related factors to recruitment rates in Canada’s boreal population of woodland caribou (Rangifer tarandus caribou). Aggregate habitat disturbance had low causal attribution (17.6%). Attribution was greater (29.5%) when habitat disturbance was disaggregated into different factors associated with different pathways of caribou decline. The causal attribution of all habitat factors considered nevertheless rarely exceeded 50%, suggesting that there are other systematic and/or stochastic factors that can limit the effectiveness of current habitat-related recovery actions. More effort is required to understand these factors and how they might be managed to improve the probability of successful caribou recovery.