������������������
In both scenarios , the number of contaminated restaurants declines as the percentage of compliant restaurants increases , in part because of the model ’ s rules : compliant restaurants are less likely to become contaminated . This effect can be seen in Figure 5 . The differences between scenarios were not statistically significant ( � > 0.05 ).
Table 4 sums up the comparisons between scenarios , divided by percentage of compliant restaurants .
Policy Implications and Conclusions
This study represents the application of a method for policy analysis that can be useful for shedding light on risks that typify the risk society , including foodborne illness . The results from the model indicate that , overall , having access to restaurant inspection scores results in a slightly higher mean number of sick consumers , but much less variation in the overall number of sick consumers , over 100 realizations . This also holds true for the number of sick , at-risk consumers . For both scenarios , more compliant restaurants results in fewer sick consumers . As well , although there tended to be more naïve consumers when the consumers did not have access to inspection scores and therefore did not act in a risk averse way following an illness , these distributions were generally leptokurtic , which indicates a higher probability of outcomes from the extremes ; the results from the second scenario , when consumers did have access to inspection scores and behaved in a risk averse manner for 15 time steps following an illness , tended to be much closer to a normal distribution , with more predictable outcomes . Again , in both scenarios , a higher percentage of compliant restaurants leads to more naïve consumers .
The basic findings have implications for policy systems . First , although the mean number of sick consumers was slightly lower when consumers did not have access to re-inspection scores , the high degree of variability in outcomes could present a challenge for inspectors ; it may be preferable to have a slightly higher average number of illnesses , but more predictable outcomes , rather than large swings which could contribute to public panic and negative risk perceptions . Given that many statistical procedures rely on assumptions of normality , outcomes that conform more closely to a normal distribution could contribute to better understanding of what is going on within the system and better predictive capacity from a policy perspective , which would be an advantage despite the slightly higher average number of illness .
Second , having a higher number of restaurants comply with regulations results in fewer sick consumers ; empirical evidence shows that disclosing inspection results leads to increases in inspection scores ( Jin & Leslie , 2003 ). Although a mechanism for restaurants choosing to comply with regulations was not explicitly considered in the model , there is evidence to suggest that disclosing scores would increase compliance , while improving transparency in the inspection system which would provide consumers with more information to make choices .
Restaurants Canada , an industry lobby group , has indicated that they do not support the use of grades or scores to inform consumers about the hygiene of restaurants , indicating that “ complex inspection findings based on subjective interpretations by individual inspectors cannot accurately or fairly be reduced to a single grade ” ( Griffith-Greene , 2014 , para . 33 ). The response from industry indicates that government intervention would be necessary to en-
152