4. ESSAY 3: THE EFFECT OF NOISY CERTIFICATION ON
4.4.5 Robustness Check of Previous Models
We have also run some robustness checks for our empirical analysis in Essay 2 by running similar regressions with our experimental data. Although Essay 2 has used number of bidders as the dependent variable and run a count data regression, we instead use purchase probability as the dependent variable and run a logistic regression to check whether those variables in Essay 2 again show significant effects with this data. The results from Table 4.13 shows that most of the key variables from Essay 2 have also shown the same significant effects in the experimental data. Whether the seller uses the certification or not, claimed value, and suggested price are all significant in the basic models for both the 50 percent certification and the 100 percent certification cases. Moreover, the interaction effects in the alternative specification are also all significant in both the 50 percent and the 100 percent cases. These results suggest that the empirical results from Essay 2 are feasible in our experimental data, and the findings of Essay 2 regarding the economic incentive for low-type sellers to disclose quality information are all robust and authentic. Moreover, the significance of the interaction effect shows that the
asymmetric effect of certification also exists in our experimental data and thus sellers with higher than average quality gain more from revealing weaknesses.
Table 4.13:Regression of Previous Model Specifications with Experimental Data
50% Certification 100% Certification
Basic Model Alternative Model Basic Model Alternative Model
Dummy for Certification Usage 1.529 *** 0.117 3.435 *** 1.714 ***
(0.254) (0.561) (0.359) (0.635)
Claimed Value 0.012 ** 0.006 0.023 *** 0.018 ***
(0.005) (0.005) (0.005) (0.006)
Suggested Price -0.028 *** -0.029 *** -0.037 *** -0.046 ***
(0.006) (0.006) (0.007) (0.008)
Interaction Between Certification
Usage and Claimed Value N/A
0.014 *** N/A 0.019 *** (0.005) (0.006) Observations 435 435 435 435 Chi-squared 46.8 *** 49.14 *** 92.5 *** 86.45 *** d.f. 3 4 3 4 * p<0.1. ** p<0.05. *** p<0.01. 4.5 Discussion
In this essay, we have tried to determine the effect of noisy certification on various market outcomes such as seller profit, buyer profit, and the level of information disclosure. The analytic model has predicted that seller profit will be the lowest with 50 percent certification and the highest with no certification mostly due to the certification cost, and that buyer profit will always be zero. This is somewhat surprising result as certification helps neither sellers nor buyers, suggesting that there is no need to use certification in a market with information asymmetry when we consider the certification cost. However, our experimental analysis has shown different results and suggested that there should be more factors determining the effect of certification on market outcomes than certification costs. We have come up with the following observations.
First, the seller profit is actually the highest with 100 percent certification and the lowest with 50 percent certification. The reason behind this pattern is twofold. Seller profit is highest with 100 percent certification because buyers purchase more frequently with 100 percent certification than with 50 percent or no certification. Seller profit is also the lowest with 50 percent certification as sellers have to pay for the certification cost although buyers do not purchase more with 50 percent certification than with no certification. Second, buyer profit is the highest with 100 percent certification and the lowest with no certification. This is again due to the fact that buyers purchase more frequently with 100 percent certification, and that the level of the seller’s dishonesty is much higher with no certification. In other words, buyer profit is the lowest with no certification because sellers exaggerate the quality more (i.e., are more dishonest) than when certification of either type is available.
Therefore, we have found that inaccurate certification does not help sellers, as it provides the lowest profit. On the other hand, inaccurate certification actually helps buyers, as sellers disclose more information even with inaccurate certification and thus buyer profit is higher than with no certification. Sellers thus might want to have either a highly accurate certification or no certification at all, as inaccurate certification is worse than no certification. However, it is different for buyers, as even an inaccurate certification mechanism would help buyers by providing higher profit. In terms of a public policy point of view, we have found that overall social welfare increases only with 100 percent certification, and 50 percent certification does not really affect social welfare compared with no certification. Therefore, policy makers or other third-party certification providers should carefully consider the quality of the certification system before introduction, as inaccurate certification may actually decrease entire social welfare, especially when considering the cost to establish the new system. This essay also provides
specific strategies for marketing managers on how to deal with information asymmetry in markets. When there is no certification available in a market under information asymmetry, it is better for sellers to focus on pricing strategy then on overstating the quality, as the quality claim is not trusted by buyers (“cheap talk”). When there is certification available, using the
certification and disclosing more information contributes a lot toward increasing the profit regardless of its accuracy, since any quality claim without certification does not affect buyer behavior (“cheap talk,” again), and the use of certification significantly increases purchase probability. Other than these main findings, this essay has also provided various interesting implications about market outcomes under information asymmetry, such as the effect of certification on social welfare and irrational behaviors of sellers and buyers. We have also confirmed that the findings of Essay 2 are replicated with our experimental data.
Overall, the findings from the analytic model and the experimental analysis agree that an inaccurate certification does not help sellers as much as an accurate certification does, and is even worse than no certification. This finding is somewhat counterintuitive, as people believe that certification generally helps sellers, and it may be able to provide some important
implications to certain sellers who attempt to manipulate the certification system to increase short-term profit, as any noise added to the certification system will eventually work against the sellers. The result of this essay that it is better for everyone (i.e., both sellers and buyers) to have a solid and accurate certification system is consistent with the findings from Essay 1 in that the certification may provide economic incentives for sellers to fully disclose quality information and increase social welfare. Moreover, the variation in the effects of different certification systems found from this essay may also provide an explanation about the empirical results of Essay 2 where the economic incentive for information disclosure differs across different product
categories. Future research may extend the findings from this essay and provide more evidence regarding the effect from the noise of the certification on various market outcomes by exploring other observational market data.