On normal days, the temperature decreases with altitude, allowing air pollutants to rise and disperse. During inversion episodes, a warmer air layer at higher altitude traps pollutants close to the ground. We show how readily available NASA satellite data on vertical temperature profiles can be used to measure inversion episodes on a global scale with high spatial and temporal resolution. Then, we link inversion episode data to ground level pollution monitors and to daily in- and outpatient records for the universe of children in Sweden during a six-year period to provide instrumental variable estimates of the effects of air quality on children's health. The IV estimates show that the respiratory illness health care visit rate increases by 8 percent for each 10 μm/m³ increase in PM10; an estimate four times higher than conventional estimates. Importantly, by linking the health care data to detailed records of parental background characteristics, we show that children from low-income households suffer significantly more from air pollution than children from high income households. Finally, we provide evidence on the importance of several mechanisms that could contribute to the difference in the impact of air pollution across children in rich and poor households.