However I think choosing to use an economic example as your theme may be problematic. Economic studies are basically impossible to conduct rigorously. There can be no controlled, double blind, repeatable experiments. In addition, given the fundamental inteconnectedness of all things (Douglas Adams was a greater man than many people realize), it is extremely difficult to predict what the effects of a change actually are, and therefore what effects need to be measured (at least in order for any results to be usable for policy decisions, where it is necessary to consider all effects, not just a subset of them).
As a consequence at least one school of economics (Austrians) argues that all economic studies are bunk and that only economics based on deductive logic is valid. I’m not sure I completely agree with this position, but I can see where they are coming from, and it is interesting that the current state of economic “science”, based upon such studies, appears incapable of successfully making any useful prediction at all. The Austrians aren’t really any better: they are very limited in the predictions they are able to make because deductive logic limits their models; but at least in their case they know that they cannot make predictions.
For the sake of argument let us reject the Austrian view, and accept that evidence from flawed experiments is better than no evidence at all. Even so we must at least I think accept that all evidence from studies in such a field must be treated exceedingly cautiously and not given the same weight as a laboratory experiment in a “hard” science, or even a well designed psychology study or medical trial. And perhaps this means that we must also continue to give credence to well founded logical theory even when we have some real world evidence that conflicts with it.
So what are the implications of the above thinking when applied to the theme of minimum wage?
(1) Firstly I think that your assumption that the asymmetry of the funnel probably arises from publication bias is suspect. Asymmetry can also arise from a systematic difference between studies of higher and lower precision, or use of an inappropriate effect measure: exactly the sort of problems that we would expect to arise given the real world constraints on economic research. So I think you should interpret the asymmetry not as “probable publication bias”, but as invalidating your use of this funnel plot to draw conclusions without further detailed investigation of individual studies.
(2) The economic law of supply and demand is a very strong and well supported law, both by simple, elegant theory and a great deal of real world evidence. To suggest that it doesn’t apply to the price of labor is a staggering assertion which I suspect you are not making, although your blanket statement “all in all, I think there’s at least some evidence that the liberals are right on this one” seems somewhat misleading to me. To your credit you do identify the many shortcomings of this method, but my interpretation is that these shortcomings which result from attempting to amalgamate the results of studies on different populations, subject to different inputs and measuring different effects are sufficient to render any conclusion essentially meaningless.
(3) Even if we discovered that publication bias is the cause of the asymmetry (which seems unlikely to me and would take a great deal of investigative effort on our part), should we allow these studies to guide policy and raise minimum wages? Surely (at least if we are aiming to maximize utility rather than win votes) the answer is no? We have no idea what other potentially negative effects a change in minimum wage has on the economy, nor do we even have any specific information about the set of circumstances for which our tentative conclusion that raising minimum wage does not impact employment holds true.
Your overall argument is well made.
However I think choosing to use an economic example as your theme may be problematic. Economic studies are basically impossible to conduct rigorously. There can be no controlled, double blind, repeatable experiments. In addition, given the fundamental inteconnectedness of all things (Douglas Adams was a greater man than many people realize), it is extremely difficult to predict what the effects of a change actually are, and therefore what effects need to be measured (at least in order for any results to be usable for policy decisions, where it is necessary to consider all effects, not just a subset of them).
As a consequence at least one school of economics (Austrians) argues that all economic studies are bunk and that only economics based on deductive logic is valid. I’m not sure I completely agree with this position, but I can see where they are coming from, and it is interesting that the current state of economic “science”, based upon such studies, appears incapable of successfully making any useful prediction at all. The Austrians aren’t really any better: they are very limited in the predictions they are able to make because deductive logic limits their models; but at least in their case they know that they cannot make predictions.
For the sake of argument let us reject the Austrian view, and accept that evidence from flawed experiments is better than no evidence at all. Even so we must at least I think accept that all evidence from studies in such a field must be treated exceedingly cautiously and not given the same weight as a laboratory experiment in a “hard” science, or even a well designed psychology study or medical trial. And perhaps this means that we must also continue to give credence to well founded logical theory even when we have some real world evidence that conflicts with it.
So what are the implications of the above thinking when applied to the theme of minimum wage?
(1) Firstly I think that your assumption that the asymmetry of the funnel probably arises from publication bias is suspect. Asymmetry can also arise from a systematic difference between studies of higher and lower precision, or use of an inappropriate effect measure: exactly the sort of problems that we would expect to arise given the real world constraints on economic research. So I think you should interpret the asymmetry not as “probable publication bias”, but as invalidating your use of this funnel plot to draw conclusions without further detailed investigation of individual studies.
(2) The economic law of supply and demand is a very strong and well supported law, both by simple, elegant theory and a great deal of real world evidence. To suggest that it doesn’t apply to the price of labor is a staggering assertion which I suspect you are not making, although your blanket statement “all in all, I think there’s at least some evidence that the liberals are right on this one” seems somewhat misleading to me. To your credit you do identify the many shortcomings of this method, but my interpretation is that these shortcomings which result from attempting to amalgamate the results of studies on different populations, subject to different inputs and measuring different effects are sufficient to render any conclusion essentially meaningless.
(3) Even if we discovered that publication bias is the cause of the asymmetry (which seems unlikely to me and would take a great deal of investigative effort on our part), should we allow these studies to guide policy and raise minimum wages? Surely (at least if we are aiming to maximize utility rather than win votes) the answer is no? We have no idea what other potentially negative effects a change in minimum wage has on the economy, nor do we even have any specific information about the set of circumstances for which our tentative conclusion that raising minimum wage does not impact employment holds true.