input use, production, and income of farmers in the program. They also attempted to explain why farmers default on formal loans. Their model is a significant improvement over those used in most previous studies aimed at measuring credit impact. David and Meyer provide an extensive review of these studies. The major RH findings are that credit and fertilizer subsidies caused, at most, a 21% to 30% increase in yields on these farms, and they predicted that default rates should average 10% per year. Despite its strengths, this study suffers from several weaknesses inherent in most credit impact research. These flaws are the focus of this critique. We also suggest how to improve future research on rural financial markets in less developed countries. Our first concern is that RH treat loans as if they were production inputs, and they largely ignore the nonrice, nonfarm, and consumption activities of the farm household. They do not clarify that a formal loan is just one of various sources of near-cash liquidity available to farm households, and that purchases of inputs for rice production are only one of a number of uses of liquidity. RH do not account for the fungible nature of credit, substitution of borrowed funds for own funds, and possible diversion of borrowed funds to nonfarm uses or consumption. We feel it is useful to define loans as claims on resources, not as production inputs. Viewing credit as a production input rather than as a claim on resources is a weakness in most credit impact studies. The essential characteristic of credit is its fungibility. Borrowed funds can be used to buy any good or service available in the market, irrespective of the stated loan purpose. Hence, the effect of credit on rice output is not directly measurable. To assign cause and effect, a researcher would need comprehensive information on all sources and uses of liquidity by the farm household concurrent with the loan. While RH do acknowledge that credit is fungible, they do not integrate this notion into their analysis. Substitution of borrowed for owned funds is generally ignored by researchers, including the fact that a farmer's own funds and his borrowed funds are equal in purchasing power and can be substituted for each other. One needs to have information on the marginal changes in liquidity use occurring in all farm household activities concurrent with borrowing. For example, if additional beer drinking yields a higher satisfaction to the farm household than the return from increased fertilizer use in rice, the net effect from a fertilizer credit program will, undoubtedly be some increased beer consumption. RH did not analyze the diversion of liquidity to consumption or nonfarm enterprise activities (but they do note that it is difficult to detect). These alternative uses of funds are available to all farm households and need to be recognized. There is also the question of additionality. Additionality refers to the behavior and expenditure patterns of the farm household when additional liquidity is acquired. To account for this, one needs to know how additional liquidity is spent. RH assume that additional liquidity will be spent on rice enterprises. This leads to an overestimation of any credit