The Magnitude of Hypothetical BiasT

The Magnitude of Hypothetical Bias

The extent to which WTP for new products, and the marginal WTP for nutrition information, can be accurately assessed in a hypothetical setting can be seen by

comparing the magnitudes of WTP in treatments 2 and 3, as shown in Table 4. It is

evident that in the absence of a cheap talk script, hypothetical bias is large: on average,

participants overstated their WTP by a factor of above 2 in hypothetical scenarios

compared to real scenarios. This supports the findings reported in List and Gallet (2001).

In addition, the marginal WTP estimates from the hypothetical choices are also

overstated. As might be expected, the degree of hypothetical bias for white varieties is

minimal, since this is the product with which consumers are familiar. Note that unlike the

case in most hypothetical elicitation experiments, consumers in Uganda had an

opportunity to taste the three products that they were being asked to assess. So the

difference in the “real” and “hypothetical” arms cannot be attributed to their experience

with the product, but entirely to the fact that in the real treatment respondents had to make

a purchase, whereas in the hypothetical arm, they did not have to do so.

The Mitigation of Hypothetical Bias through Cheap Talk

Turning to the effectiveness of cheap talk in mitigating hypothetical bias, the results

suggest that although the use of a cheap talk script does result in a reduction in the

magnitude of the willingness to pay, the bias remains substantial. This finding is of

interest because, unlike most other cheap talk scripts, such as those used by Cummings

and Taylor (1999) and List (2001) which explicitly indicate the direction of bias, the cheap

talk script used in this study was deliberately neutral in its wording. Nonetheless, the

cheap talk had the desired effect of lowering hypothetical bias, but it did not eliminate it.

For example, the WTP for deep orange in the hypothetical arm was UGS750 per

kilogram, more than double the UGS357 in the real arm, whereas with cheap talk the

WTP estimates drop to UGS553/kilogram; hence considerably lower than

750UGS/kilogram, but still much higher than the UGS357/kilogram. Upward biases are

also seen in the estimated marginal willingness to pay for orange and deep orange

varieties as well, although once again these biases are lower than was the case in the

hypothetical scenario without using a cheap talk script.