border is not a new phenomenon, with previous surges occurring in the 1920s and 1950s, persis tent mass migration did not take hold until late in the twentieth century. Among the factors contributing to emigration from Mexico are weak growth in the country's labor demand and strong growth in its labor sup ply. Mexico's economy stagnated in the 1980s and 1990s, such that per capita GDP in the early 2000s was unchanged from two decades before. During periods of wage decline in Mexico, emi gration from the country spiked. Perhaps less appreciated is that the 1980s were also a period of accelerated growth in Mexico's relative labor supply. With the US baby boom peaking in 1960, the number of US native born individuals coming of working age actually declined in the 1980s. Adding in the secular increase in US educational attain ment, the number of native born American workers with less than a high school education has dropped sharply. In Mexico, high levels of fertility in the 1960s and 1970s meant that two decades hence the country had large numbers of young adults entering the labor force. While educational attainment has also increased in

We report on a field experiment using several methods for collecting deposits made in formal bank accounts in rural areas in Sri Lanka. We find that only frequent, face-to-face collection increases aggregate household savings. Collection involving community lock boxes increases balances at the collecting bank, but not overall household savings. Only community box collection appears to have the possibility of being financially viable. The various collection methods allow us to unbundle the role of frequency, salience and habit formation in deposit decisions. We find that frequency and salience affect the number of transactions, but not the level of savings.