The prospect of better physical devices, such as logic gates or solar cells, often generates huge interest and investment. Of course there are many more physical devices where improvements generate much less interest, because we haven’t yet found nearly as much use for those devices. But even so, for devices we often use, small improvements can be very big news.
Similarly, there are many widely used computer algorithms where small improvements also generate big interest and financial investments. Of course most gains aren’t like this. For example, there is less interest in techniques tied to very narrow contexts, such as ways to reorganize particular programs. But when wide use is plausible, algorithm gains can be big news.
We can do engineering and design not only with physical and software systems, but also with social systems. There should of course be less interest in designs tied to very particular contexts, such as reorganizing the management of a particular firm. But we often repeatedly use some simple social mechanisms, like voting. So we should be a lot more interest in improving the designs of these.
I started out in engineering, moved to physics, then to software, and then finally to economics. That last move was very much inspired by big apparent gains from better social institutions. I knew that in physical and software engineering we put in huge efforts to scour the vast space of possible designs to find even small gains on devices of moderate generality. Yet in economics it seemed that big gains could be found from very simple easy to find innovations on general mechanisms of wide applicability.
Over two decades later, I must admit that the world shows far less interest in better designs for institutions and social mechanisms, relative to better designs for physical and software systems. Few talk about them, and even fewer business ventures pursue them. Some say that physics and software designs are far more valuable because we know far less about economics; these proposed social designs just don’t work. But this claim seems just wrong to me.
Yes of course any particular argument for any particular social design will make convenient but questionable assumptions. But this is also true for our main arguments for physical or software designs. They also almost always neglect relevant considerations. Tractable analysis simplifies.
I recently posted on a new voting mechanism. Voting is a very general process whose main purposes are also pretty general. I’ve also posted for years about the very general advantages of prediction markets for the problem of info aggregation, which is a very general problem. (Scott Sumner sees their gains as so obvious he calls anything else “Stone Age Economics”.) I just heard a nice talk on better political institutions to promote urban density. And economic journals are full of articles describing new institution designs, and testing the effects of institutions that are not widely adopted.
Yes, proposed new social mechanisms often fail along the path from simple theory models to complex models to lab experiments to small field experiments to large field trials. But physical and software designs also often fail along this path. I don’t see social designs as failing much more often, except for the key failing of not generating much enthusiasm or interest. That is, most people just don’t seem to care how well social designs do in theory or lab or field tests. Even most social scientists don’t care much about design innovations outside their specialty areas.
Yes in the last decade or so there has been more enthusiasm for social innovations embodied in physical and software innovations, like smart phones or block chains. But this enthusiasm seems to be mainly an accidental side effect of tech enthusiasm. For example, while many are excited by Uber achieving new value in cheaper-if-nominally-illegal cab services, most of those gains could have come decades ago from just deregulating cabs, an option in which there was little interest. As another example, there is far more interest today in prediction markets build on block chains than in ordinary prediction markets, even though far more value could be achieved by the later.
I should admit that this all confirms Bryan Caplan’s claim that few people can generate much emotional enthusiasm for efficiency. Bryan says people are far more engaged by moral arguments. I’d say people are also far more engaged by following fashion and by us vs. them coalition politics. Most apparent interest in innovation in social designs can be attributed to these three sources; we explain little more by positing an additional direct interest in helping us all get more of what we want.
This seems mostly also true at the level of smaller organizations like firms. While people give lip service to increasing the efficiency or effectiveness of the organization as a whole, that in fact generates little passion. The passion we do see in the name of efficiency mostly advances particular factions and individual careers. Homo hypocritus is quite skilled at saying that he serves the great good, while actually serving far more personal ends.
Added 9a: Many of you seem to be stuck on the ideas that social innovations can’t be tested unless the entire world agrees to adopt them. Or an entire nation, or city. Yes, some innovations are like that. (There are also physical and software innovations like that.) But a great many social innovations can be tried out on very small scales, where regulations do not block them. And there is very little interest in pursuing these innovations.
Opinion polls that try to ascertain if the electorate is x/100-x on an issue use at least 1000 participants. A more common divide would be 55/45 and there a sample of 50 people would certainly not be considered statistically significant. But really, a parliament of 1000 people isn't really problematic: the US has ~435 people in its lower house, India has ~550. For the US with a 1000 person lower house, per-capita representation would still be lower than that of many mid-sized countries.
If the electorate were 70/30 on an issue, you wouldn't need 1000 voters to get it "right" reliably.
If the electorate is 49.999/50.00001 though, you might need hundreds of thousands of voters. But at this point, I kind of throw my hands up and say who the f*ck cares. The marginal EV (practical or philosophical) of getting the vote "right" is basically nil and not enough to outweigh the superiority of a concentrated electorate over a dispersed one.