Is It Better to Know than to Not Know?

A 1994 song titled “Positive” by Spearhead goes:

“I should-a done this a long time ago

A-lot of excuses why I couldn't go

I know, these things and these things, I must know

'Cause it's better to know than to not know!

But how am I gonna live my life if I'm positive?

Is it gonna be a negative?

…”

When it comes to HIV testing, the prevailing assumption has been that it is both privately optimal for individuals at risk and socially optimal due to the negative externalities, i.e. it is better to know your status.

In a paper forthcoming in the American Economic Review (online appendix here), a small team of two neurologists and an economist study the behavior of people who are at risk for Huntington disease (HD) and wonder (a) why they don’t get tested and (b) why they don’t behave differently.

The authors describe the behavior of people enrolled in a study of Huntington Disease who were at risk of developing the degenerative neurological disease (meaning that they had a 50% chance of inheriting the expanded copy of the Huntingtin gene from one of their parents) but had not been previously tested for the genetic expansion. They document the following:

1.    The incidence of individuals who took up predictive genetic testing, which is perfectly predictive of eventually developing the disease, was less than 10% over the approximately decade-long study period.

2.    Individuals with higher levels of objective probabilities (assessments by investigators based on clinical signs) were more likely to pursue predictive testing.

3.    Individuals’ perceived probabilities of having the disease were much lower than the objective probabilities and were, in many cases, extremely biased. For example, 11% of the people with certain signs of the disease thought there was no chance that they carried the gene mutation with the average probability perceived among this group being only 52%.

4.    While people who are certain that they carry the gene mutation behave differently in important domains of life such as marriage, pregnancy, retirement, financial planning, and recreation than those who are certain that they don’t carry the gene; those individuals who remain untested behave as though they don’t have the gene – even when objective assessments indicate that the probability of developing the disease is more than 90%.

The authors then suggest that the neoclassical model cannot accommodate these behaviors and propose that they can be better explained by an optimal expectations framework, in which beliefs impact utility directly. In the authors’ words when individuals avoid testing they “…are not making a mistake… they are avoiding testing because they prefer to consume happiness in the anticipation period.” (As an aside, Kaler and Watkins describe a similar reluctance to HIV testing in this paper.)

In the concluding section, the authors try to draw a few parallels to other diseases, including HIV, especially because HIC and cancer are much more common than HD. However, while I think that the hypothesis here may have merit, I was much less convinced of the implications for HIV testing.

First, the authors cite a couple of papers finding that testing positive for HIV leads to a reduction in risky behaviors. However, the problem is that these risky behaviors are self-reported (or increased purchase of condoms in the case of Thornton 2008) and may not be indicative of actual risk reduction – particularly because HIV testing itself may change the extent of reporting bias (see, e.g., Gong 2012).

Second, the authors are talking about information avoidance in the case of predictive testing for HD. This is not necessarily the case for HIV: door-to-door testing campaigns are very successful in getting people tested and learning their statuses. In our study in Malawi, the refusal rates for HIV testing were tiny and almost everyone who got tested learned their results from the counselors. What does happen in the case of HIV, now documented by multiple papers (see here, here, and here) is that a substantial share of people who receive HIV+ test results later report zero likelihood of being HIV+. This is a direct contradiction to the authors’ assumption that if one receives a perfectly predictive test result (such as for HIV or HD), then they cannot change their beliefs about the probability of developing the disease. Perhaps the denial of HIV+ test results that are almost 100% accurate can be seen as an extreme implication of the optimal expectations framework, where the individual is better off not believing the result of a fully accurate test.

The problem is that while these HIV “deniers” may be forming beliefs that are privately optimal, they may also subsequently act in ways that may make HIV testing not even socially optimal. In ongoing work, we are seeing signs that such deniers may actually increase their risky behavior and perhaps contribute to the transmission of the epidemic. In that case, our approach to follow-up work with those individuals who tested HIV+ becomes more complicated yet also much more important. Stay tuned for more on this here in the upcoming weeks…