there is a high correlation between what voters as a whole want on any single narrow issue, and what the outcome of the black box produces
I certainly agree that if I compare the results of the black-box output and reported public opinion in various U.S. jurisdictions, I’ll find that shifts in the two are correlated on certain kinds of issues and not on others. I’ll even find, often, that shifts in the latter predict shifts in the former (though, as you say, there’s significant lag time).
I’m not convinced that the relevant distinction is “narrow” vs. “meta”, but that’s a longer conversation than I feel like having.
In those cases where such a correlation does exist, how would you go about distinguishing between the following theories to account for that correlation?
shifts in reported public opinion reflect changes in public opinion, and changes in public opinion drive changes in black-box output
shifts in reported public opinion drive changes in black-box output (e.g., because legislators take those reports seriously), regardless of actual public opinion
pollsters are able to predict shifts in black-box output, and alter their reporting of public opinion accordingly
the public is able to predict shifts in black-box output, and alter their opinions accordingly
some other set of factors we don’t yet know about controls both public opinion shifts and black-box output shifts, and causes them to be correlated
some other set of factors we don’t yet know about controls both shifts in public opinion reporting and black-box output, and causes them to be correlated
I would say that a combination of the first two (Public opinion changes black-box output) and (changes in polls change black box output) are the most likely. I can think of some common political observations that seem to match those theories (for example: a President that wins an election by a large margin is considered to have more “political capital”, it’s easier for him to get things done in Congress in the near future.)
I guess the best way to distinguish between, say, “Public opinion changes cause political change” and “Political change causes public opinion change” is to look at which happens first. Looking at some historical examples, espcally on really big issues that people feel strongly about, it seems that there is a lead time where first people feel strongly about an issue and then, a few years later, the political outcome changes. For example, look at Prohibition. Popular opinion in favor of Prohibition grew for several years before it actually happened (it was a significant part of the progressive movement). Then, after it actually happened and people saw the flaws, popular opinion changed over time, but it took several years before Prohibition became unpopular before it was actually repealed. So there seems to be a lag time in most cases, which implies the direction of causality.
It’s harder to rule out the possibility that there is one third factor that explains both. Perhaps in some cases there is some kind of general culture or media shift that affects both the popular opinion and the politician’s opinion at the same time? Even if that is so, though, it still seems like a system that keeps the attitudes of the population and the end-result of the govenrment in synch is likely to be more stable then one that doesn’t, so that still might not be a bad thing.
Anyway, the point of my post was just that, despite all the flaws of our democracy, that it appears that the attitudes and votes of the population as a whole seem to have a greater influence on the outcome then Eliezer was suggesting.
I agree that it’s difficult to rule out the presence of common causal factors, especially in complex systems where there really are a lot of causal factors.
To my mind, that’s a good reason not to assume simple causality from correlation between two factors in a complex system.
I certainly agree that if I compare the results of the black-box output and reported public opinion in various U.S. jurisdictions, I’ll find that shifts in the two are correlated on certain kinds of issues and not on others. I’ll even find, often, that shifts in the latter predict shifts in the former (though, as you say, there’s significant lag time).
I’m not convinced that the relevant distinction is “narrow” vs. “meta”, but that’s a longer conversation than I feel like having.
In those cases where such a correlation does exist, how would you go about distinguishing between the following theories to account for that correlation?
shifts in reported public opinion reflect changes in public opinion, and changes in public opinion drive changes in black-box output
shifts in reported public opinion drive changes in black-box output (e.g., because legislators take those reports seriously), regardless of actual public opinion
pollsters are able to predict shifts in black-box output, and alter their reporting of public opinion accordingly
the public is able to predict shifts in black-box output, and alter their opinions accordingly
some other set of factors we don’t yet know about controls both public opinion shifts and black-box output shifts, and causes them to be correlated
some other set of factors we don’t yet know about controls both shifts in public opinion reporting and black-box output, and causes them to be correlated
Interesting question.
I would say that a combination of the first two (Public opinion changes black-box output) and (changes in polls change black box output) are the most likely. I can think of some common political observations that seem to match those theories (for example: a President that wins an election by a large margin is considered to have more “political capital”, it’s easier for him to get things done in Congress in the near future.)
I guess the best way to distinguish between, say, “Public opinion changes cause political change” and “Political change causes public opinion change” is to look at which happens first. Looking at some historical examples, espcally on really big issues that people feel strongly about, it seems that there is a lead time where first people feel strongly about an issue and then, a few years later, the political outcome changes. For example, look at Prohibition. Popular opinion in favor of Prohibition grew for several years before it actually happened (it was a significant part of the progressive movement). Then, after it actually happened and people saw the flaws, popular opinion changed over time, but it took several years before Prohibition became unpopular before it was actually repealed. So there seems to be a lag time in most cases, which implies the direction of causality.
It’s harder to rule out the possibility that there is one third factor that explains both. Perhaps in some cases there is some kind of general culture or media shift that affects both the popular opinion and the politician’s opinion at the same time? Even if that is so, though, it still seems like a system that keeps the attitudes of the population and the end-result of the govenrment in synch is likely to be more stable then one that doesn’t, so that still might not be a bad thing.
Anyway, the point of my post was just that, despite all the flaws of our democracy, that it appears that the attitudes and votes of the population as a whole seem to have a greater influence on the outcome then Eliezer was suggesting.
I agree that it’s difficult to rule out the presence of common causal factors, especially in complex systems where there really are a lot of causal factors.
To my mind, that’s a good reason not to assume simple causality from correlation between two factors in a complex system.