13.11.2012 Views

Testing Distributional Dependence in the Becker-DeGroot-Marschak ...

Testing Distributional Dependence in the Becker-DeGroot-Marschak ...

Testing Distributional Dependence in the Becker-DeGroot-Marschak ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>in</strong>structions reported elicitations of $0 more often than those with basic <strong>in</strong>structions regardless of<br />

<strong>the</strong> distribution faced <strong>in</strong> round 2.<br />

4.2 Results on Mass<br />

The data provide compell<strong>in</strong>g evidence that <strong>the</strong> position of mass <strong>in</strong> <strong>the</strong> distribution of random<br />

prices can have a pronounced effect on <strong>the</strong> WTP elicitations across subjects. More mass to <strong>the</strong><br />

right often leads to higher WTP elicitations. This is suggested at first glance by <strong>the</strong> descriptive<br />

statistics given <strong>in</strong> Table 4. The pairwise comparisons that appear <strong>in</strong> Table 6, Table 7, and Table<br />

8, often show significant difference <strong>in</strong> average WTP across distributions. This is true even though<br />

<strong>the</strong>se tables are based only on <strong>the</strong> last elicitations, which should exhibit less distributional<br />

sensitivity than ei<strong>the</strong>r <strong>the</strong> first elicitations or <strong>the</strong> entire set of observations. Fur<strong>the</strong>rmore, <strong>the</strong><br />

patterns and conditions under which <strong>the</strong> null statement of Hypo<strong>the</strong>sis 2 is rejected are<br />

enlighten<strong>in</strong>g.<br />

Table 6 holds <strong>the</strong> support and <strong>the</strong> type (i.e., shape) of <strong>the</strong> distribution constant and focuses<br />

on symmetric shifts <strong>in</strong> <strong>the</strong> location of mass with<strong>in</strong> <strong>the</strong> fixed support. For five of <strong>the</strong> six result<strong>in</strong>g<br />

pairs, <strong>the</strong> null of Hypo<strong>the</strong>sis 2 is soundly rejected. Interest<strong>in</strong>gly, <strong>the</strong> effect disappears <strong>in</strong> all cases<br />

for <strong>the</strong> subjects who viewed detailed <strong>in</strong>structions.<br />

Table 7 holds <strong>the</strong> shape of <strong>the</strong> distribution and <strong>the</strong> location of <strong>the</strong> mass fixed and varies <strong>the</strong><br />

support across <strong>the</strong> pairs considered. For all of <strong>the</strong> seven pairs of distributions <strong>the</strong> null of<br />

Hypo<strong>the</strong>sis 2 is rejected outright. That <strong>the</strong> effect is particularly strong when <strong>the</strong> support is<br />

varied is unsurpris<strong>in</strong>g, s<strong>in</strong>ce <strong>the</strong> elicitations for <strong>the</strong> distributions with higher support would<br />

mechanically be higher if enough of <strong>the</strong> subjects rema<strong>in</strong>ed unsure about whe<strong>the</strong>r it would be<br />

possible or proper to submit values outside of a given distribution’s support. This table, <strong>the</strong>n,<br />

provides <strong>the</strong> clearest evidence that <strong>the</strong> detailed <strong>in</strong>structions had a substantial effect: for that set<br />

of subjects only two of <strong>the</strong> seven pairs had significantly higher elicitations for <strong>the</strong> distribution<br />

with higher support, and <strong>the</strong> level of significance was less pronounced.<br />

20

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!