Saturday, June 04, 2016

Left vs Right Science vs Risk vs Propensity to Regulate

Jayson Lusk has an interesting post on his blog related to an article in the Journal of Agricultural and Resource Economics finding an interesting relationship between left leaning voters and their willingness to support GMO labeling initiatives:



“One distinction, which I think is missing, is the greater willingness of those on the left to regulate on economic issues, such as GMOs, than those on the right. Stated differently, there are questions of science: what are the risks of climate change or eating GMOs. And then there are more normative questions: given said risk, what should we do about it? Even if the left and the right agreed on the level of risk, I don’t think we should expect agreement on political action.”


If I understand this correctly, I think this implies that if both those on the left and right agreed that there was some 'day after tomorrow' scenario (in terms of climate change) that warranted some type of government intervention, and they agreed that the science says there is a 3% chance of it happening without the intervention, then those on the right might object to the intervention for that given level of risk while a more left leaning person would support it. A right leaning person might suggest more market based alternatives or taking the gamble. But perhaps if the risk were higher, they might support doing more. In other words there might be different thresholds for the level of risk required to support a given policy interventions across the political spectrum.

Of course, the scientific consensus on climate change may not really even be strong enough to know for sure, i.e. the science isn't settled on exactly what scenarios are likely to play out and the probabilities that they will occur. There's a lot of science to support a wide range of probabilities and scenarios based on a number of assumptions. (see herehere, here, and here). So really, I think even the science, risk, and potential outcomes or scenarios are largely based on perceptions and these might actually differ significantly across the political spectrum. Maybe its really about perceived risk.

Just thinking about this a little more what if we specified a model of preferences toward government intervention like that below (this is more an illustration than a serious attempt to look at this empirically):

Pr(SUPPORT POLICY) =  B0 + B1 PERCEIVED RISK + B2 KNOWLEDGE

So if we estimated simple linear probability models as specified above for democrats and republicans (as short hand for political preferences) according to the story line above B1 would be higher for democrats than republicans. (I'm ignoring the use of interaction terms on purpose for simplicity) I wonder if this would also be true for B2, for a given level of knowledge, would B2 be higher for democrats/liberals? I also wonder if PERCIEVED RISK is really a function of KNOWLEDGE? Maybe a different specification would look something like:

Pr(SUPPORT POLICY) = B0 + B1 PERCEIVED RISK(KNOWLEDGE)

where PERCEIVED RISK = f(KNOWLEDGE)

So in this case perhaps B1 would still be higher for those with more left leaning politics. Still I wonder, besides this effect, what if its the case that the level or mean of PERCEIVED RISK is in general higher for those on the left? So you have this effect of a greater inclination for a preference for government intervention given a level of PERCEIVED RISK (via B1) but also a population of left leaning voters with a PERCEIVED RISK levels that are on average some magnitude higher. Both of these effects would likely increase the propensity of supporting government intervention.

Consider also....if PERCEIVED RISK = f(KNOWLEDGE), is the level of KNOWLEDGE about GMOs or climate change the same for those on the left and right and is this really what is partly determining different levels of PERCEIVED RISK?  I'm not sure....how often do we hear arguments from the left that drastic actions or mitigating policies to combat climate change are necessary because of the scientific consensus on climate change when in fact the consensus as it is is pretty weak. Too weak to offer much guidance on actions, or very precise estimates of actual risks. (again see herehere, here, and here). And even some of the world's leading experts in risk modeling tend to have some ideas about GMO risks that can be seriously questioned (see here). There was a really good book a few years back discussing voter preferences and systemic bias regarding economic policy that addressed similar issues (see The Myth of the Rational Voter).

If preferences toward policy can be modeled in this way, an interesting and maybe promising feature is that perhaps the level of knowledge feeding into PERCEIVED risk can be altered. We often hear that science and evidence rarely will change minds when it comes to biotechnology or climate change, however, in a paper recently published by the Journal of the Federation for American Societies for Experimental Biology (FASEB) Jayson and Brandon McFadden observed the following:

1) consumers, as a group, are unknowledgeable about GMOs, genetics, and plant breeding and, perhaps more interestingly

 2) simply asking these objective knowledge questions served to lower subjective, self-assessed knowledge of GMOs (i.e., people realize they didn't know as much as they thought they did) and increase the belief that it is safe to eat GM food. 

I'm not  a PhD Economist or Psychometrician but I would think an approach similar to the structural equation modeling framework I discussed before (depicted below) might get closer to specifying and measuring all of the causal paths and connections between latent constructs around risk perception and the policy environment for GMOs or climate change. Of course that would also require a solid data set and valid survey instruments. Jayson's work seems to be leading the way. These are just my initial thoughts prior to even reading the Jayson and McFadden article or the JARE article mentioned above and honestly I have not reviewed much of the actual literature or survey analysis related to risk and perceptions or policy preferences since graduate school. Maybe a lot of this has been done already.

(click to enlarge)