Table 1.
Formal model comparison results
| Model | n | LL | Δ(AIC) | w(AIC) | Δ(BIC) | w(BIC) |
|---|---|---|---|---|---|---|
| 1. Trait only | 1 | − 365.66 | 47.00 | 0.00 | 43.18 | 0.00 |
| 2. In-domain | 2 | − 347.37 | 12.41 | 0.00 | 10.50 | 0.01 |
| 3. Monetary fungibility | 2 | − 349.16 | 15.99 | 0.00 | 14.08 | 0.00 |
| 4. Negative spillover | 3 | − 347.37 | 14.41 | 0.00 | 14.41 | 0.00 |
| 5. Spillover | 3 | − 340.17 | 0.00 | 1.00 | 0.00 | 0.99 |
| 6. State-only | 2 | − 348.44 | 14.56 | 0.00 | 12.65 | 0.00 |
LL is the log likelihood of the data given the maximum likelihood parameters. Higher (i.e., smaller negative) values indicate better fits to the data. ΔAIC is the AIC value, relative to the best (lowest) AIC value, similarly for ΔBIC. w(AIC) and w(BIC) are the probabilities of each model being the best. Each model has n parameters, and all models have a scale parameter for the Cauchy-distributed measurement error of change in log discount rates