Objective: To assess whether the strategy of "a midwife in every village" in Indonesia achieved its aim of increasing professional delivery care for the poorest women. Methods: Using pooled Demographic and Health Surveys (DHS) data from 1986-2002, we examined trends in the percentage of births attended by a health professional and deliveries via caesarean section. We tested for effects of the economic crisis of 1997, which had a negative impact on Indonesia's health system. We used logistic regression, allowing for time-trend interactions with wealth quintile and urban/rural residence. Findings: There was no change in rates of professional attendance or caesarean section before the programme's full implementation (1986-1991). After 1991, the greatest increases in professional attendance occurred among the poorest two quintiles - 11% per year compared with 6% per year for women in the middle quintile (P = 0.02). These patterns persisted after the economic crisis had ended. In contrast, most of the increase in rates of caesarean section occurred among women in the wealthiest quintile. Rates of caesarean deliveries remained at less than 1% for the poorest two-fifths of the population, but rose to 10% for the wealthiest fifth. Conclusion: The Indonesian village midwife programme dramatically reduced socioeconomic inequalities in professional attendance at birth, but the gap in access to potentially life-saving emergency obstetric care widened. This underscores the importance of understanding the barriers to accessing emergency obstetric care and of the ways to overcome them, especially among the poor.