Similarly, data indicate that the increased incidence of roaming dogs can
be linked to both a logistical problem-inefficient impoundment and absence
of pound space and to weak regulatory enforcement, such as the absence of
owner incentives to license and control dogs (Lehan 1984, 74). ln the health care area, studies have shown that higher fees for hospital and clinic services are associated with reduced demand (a 1% increase in fees reduces demand by 0.3%, which is roughly the same price elasticity of demand as for urban transit) (The Economist 1997e, 84). But studies also link higher fees (plausible causes) to higher rates of certain diseases and the possible spread of contagious diseases. This suggests that the problem should be defined to include the plausible impact of fees on both system cost recovery and on poorer groups.
In their quest for plausible causes, analysts have had difficulty distinguishing causality from correlation. To deal with this problem, policy analysts search for an instrumental variable. 1bis variable acts as a proxy for one variable in statistical analysis but is unrelated to the others. For example, in examining the relationship between imprisonment and reduced crime, it is difficult to find much of an effect due to increased incarceration. Increased incarceration could even lead to increased crime through the unintended effect of a convict absorb ing collegial training during his or her sentence. But suppose there is another variable related to incarceration that is unrelated to crime rates? Use of an instrumental variable, in this case litigation over prison overcrowding, might clarify the causal relationship between incarceration and crime rates. Using data from 1973 to 1997, researchers found that where litigation is filed and prison populations fall. crime rates increase (The Economist 1998d). Finding an instrumental variable in such litigation over prison crowding can strengthen the relationship between the other variables. This technique illustrates the need for the creative use of both empirical analysis and multidisciplinary