We combine newly created data on per student government expenditure on children in government elementary schools across India, data on per student expenditure by households on students attending private elementary schools, and the ASER measure of learning achievement of students in rural areas. The combination of these three sources allows us to compare both the “accounting cost” difference of public and private schools and also the “economic cost”—what it would take public schools, at their existing efficacy in producing learning, to achieve the learning results of the private sector. We estimate that the “accounting cost” per student in a government school in the median state in 2011/12 was Rs. 14,615 while the median child in private school cost Rs. 5,961. Hence in the typical Indian state, educating a student in government school costs more than twice as much than in private school, a gap of Rs. 7,906. Just these accounting cost gaps aggregated state by state suggests an annual excess of public over private cost of children enrolled in government schools of Rs. 50,000 crores (one crore=10 million) or .6 percent of GDP. But even that staggering estimate does not account for the observed learning differentials between public and private. We produce a measure of inefficiency that combines both the excess accounting cost and a money metric estimate of the cost of the inefficacy of lower learning achievement. This measure is the cost at which government schools would be predicted to reach the learning levels of the private sector. Combining the calculations of accounting cost differentials plus the cost of reaching the higher levels of learning observed in the private sector state by state (as both accounting cost differences and learning differences vary widely across states) implies that the excess cost of achieving the existing private learning levels at public sector costs is Rs. 232,000 crores (2.78% of GDP, or nearly US$50 billion). It might seem counterintuitive that the total loss to inefficiency is larger than the actual budget, but that is because the actual budget produces such low levels of learning at such high cost that when the loss from both higher expenditures and lower outputs are measured it exceeds expenditures.