Separating the roles of theory and direct empirical evidence in belief formation: the examples of minimum wage and anthropogenic global warming

I re­cently asked two ques­tions on Quora with similar ques­tion struc­tures, and the similar­i­ties and differ­ences be­tween the re­sponses were in­ter­est­ing.

Ques­tion #1: An­thro­pogenic global warm­ing, the green­house effect, and the his­tor­i­cal weather record

I asked the ques­tion here. Ques­tion state­ment:

If you be­lieve in An­thro­pogenic Global Warm­ing (AGW), to what ex­tent is your be­lief in­formed by the the­ory of the green­house effect, and to what ex­tent is it in­formed by the his­tor­i­cal tem­per­a­ture record?

In re­sponse to some com­ments, I added the fol­low­ing ques­tion de­tails:

Due to length limi­ta­tions, the main ques­tion is a bit sim­plis­ti­cally framed. But what I’m re­ally ask­ing for is the rel­a­tive im­por­tance of the­o­ret­i­cal mechanisms and di­rect em­piri­cal ev­i­dence. The­o­ret­i­cal mechanisms are of course also em­piri­cally val­i­dated, but the em­piri­cal val­i­da­tion could oc­cur in differ­ent set­tings.

For in­stance, the green­house effect is a mechanism, and one may get es­ti­mates of the strength of the green­house effect based on an un­der­stand­ing of the un­der­ly­ing physics or by do­ing lab­o­ra­tory ex­per­i­ments or simu­la­tions.

Direct em­piri­cal ev­i­dence is ev­i­dence that is as close to the situ­a­tion we are try­ing to pre­dict as pos­si­ble. In this case, it would in­volve look­ing at the his­tor­i­cal records of tem­per­a­ture and car­bon diox­ide con­cen­tra­tions, and per­haps some other con­found­ing vari­ables whose role needs to be con­trol­led for (such as so­lar ac­tivity).

Say­ing that your be­lief is largely grounded in di­rect em­piri­cal ev­i­dence is ba­si­cally say­ing that just look­ing at the time se­ries of tem­per­a­ture, car­bon diox­ide con­cen­tra­tions and the other vari­ables can al­low one to say with fairly high con­fi­dence (start­ing from very weak pri­ors) that in­creased car­bon diox­ide con­cen­tra­tions, due to hu­man ac­tivity, are re­spon­si­ble for tem­per­a­ture in­creases. In other words, if you ran a re­gres­sion and tried to do the usual tricks to in­fer causal­ity, car­bon diox­ide would come out as the culprit.

Say­ing that your be­lief is largely grounded in the­ory is ba­si­cally say­ing that the sci­ence of the green­house effect is suffi­ciently con­vinc­ing that the his­tor­i­cal tem­per­a­ture and weather record isn’t an im­por­tant fac­tor in in­fluenc­ing your be­lief: if it had come out differ­ently, you’d prob­a­bly just have thought the data was noisy or wrong and wouldn’t up­date away from be­liev­ing in the AGW the­sis.

I also posted to Face­book here ask­ing my friends about the push­back to my use of the term “be­lief” in my ques­tion.

Ques­tion #2: Effect of in­crease in the min­i­mum wage on unemployment

I asked the ques­tion here. Ques­tion state­ment:

If you be­lieve that rais­ing the min­i­mum wage is likely to in­crease un­em­ploy­ment, to what ex­tent is your be­lief in­formed by the the­ory of sup­ply and de­mand and to what ex­tent is it in­formed by di­rect em­piri­cal ev­i­dence?

I added the fol­low­ing ques­tion de­tails:

By “di­rect em­piri­cal ev­i­dence” I am refer­ring to em­piri­cal ev­i­dence that di­rectly per­tains to the re­la­tion be­tween min­i­mum wage raises and em­ploy­ment level changes, not em­piri­cal ev­i­dence that sup­ports the the­ory of sup­ply and de­mand in gen­eral (be­cause trans­fer­ring that to the min­i­mum wage con­text would re­quire one to be­lieve the trans­fer­abil­ity of the the­ory).

Also, when I say “be­lieve that rais­ing the min­i­mum wage is likely to in­crease un­em­ploy­ment” I am talk­ing about min­i­mum wage in­creases of the sort of­ten con­sid­ered in leg­is­la­tive mea­sures, and by “likely” I just mean that it’s some­thing that should always be se­ri­ously con­sid­ered when­ever a pro­posal to raise the min­i­mum wage is made. The be­lief would be con­sis­tent with be­liev­ing that in some cases min­i­mum wage raises have no em­ploy­ment effects.

I also posted the ques­tion to Face­book here.

Similar­i­ties be­tween the questions

The ques­tions are struc­turally similar, and be­long to a gen­eral ques­tion type of con­sid­er­able in­ter­est to the LessWrong au­di­ence. The com­mon fea­tures to the ques­tions:

  • In both cases, there is a the­ory (the green­house effect for Ques­tion #1, and sup­ply and de­mand for Ques­tion #2) that is foun­da­tional to the do­main and is sup­ported through a wide range of lines of ev­i­dence.

  • In both cases, the quan­ti­ta­tive speci­fics of the ex­tent to which the the­ory ap­plies in the par­tic­u­lar con­text are not clear. There are prima fa­cie plau­si­ble ar­gu­ments that other fac­tors may can­cel out the effect and there are ar­gu­ments for many differ­ent effect sizes.

  • In both cases, peo­ple who study the broad sub­ject (cli­mate sci­en­tists for Ques­tion #1, economists for Ques­tion #2) are more fa­vor­ably dis­posed to the be­lief than peo­ple who do not study the broad sub­ject.

  • In both cases, a sig­nifi­cant part of the strength of be­lief of sub­ject mat­ter ex­perts seems to be their be­lief in the the­ory. The data, while con­sis­tent with the the­ory, does not seem to paint a strong pic­ture in iso­la­tion. For the min­i­mum wage, con­sider the Card and Krueger study. Bryan Ca­plan dis­cusses how Bayesian rea­son­ing with strong the­o­ret­i­cal pri­ors can lead one to con­tinue be­liev­ing that min­i­mum wage in­creases cause un­em­ploy­ment to rise, with­out ad­dress­ing Card and Krueger at the ob­ject level. For the case of an­thro­pogenic global warm­ing, con­sider the draft by Kesten C. Green (ad­dress­ing whether a warm­ing-based fore­cast has higher fore­cast ac­cu­racy than a no-change fore­cast) or the pa­per AGW doesn’t coin­te­grate by Been­stock, Re­ingew­ertz, and Pal­dor (ad­dress­ing whether, look­ing at the data alone, we can get good ev­i­dence that car­bon diox­ide con­cen­tra­tion in­creases are linked with tem­per­a­ture in­creases).

  • In both cases, out­siders to the do­main, who nonethe­less have ex­per­tise in other ar­eas that one might ex­pect gives them in­sight into the ques­tion, are of­ten more skep­ti­cal of the be­lief. A num­ber of weather fore­cast­ers, physi­cists, and fore­cast­ing ex­perts are skep­ti­cal of long-range cli­mate fore­cast­ing or con­fi­dent as­ser­tions about an­thro­pogenic global warm­ing. A num­ber of so­ciol­o­gists, lawyers, and poli­ti­ci­ans of­ten are dis­parag­ing of the be­lief that min­i­mum wage in­creases cause un­em­ploy­ment lev­els to rise. The crit­i­cism is similar: namely, that a ba­si­cally cor­rect the­ory is be­ing over­stretched or in­cor­rectly ap­plied to a situ­a­tion that is too com­plex, is similar.

  • In both cases, the de­bate is some­what poli­ti­cally charged, largely be­cause one’s be­liefs here af­fect one’s views of pro­posed leg­is­la­tion (cli­mate change miti­ga­tion leg­is­la­tion and min­i­mum wage in­crease leg­is­la­tion). The an­thro­pogenic global warm­ing be­lief is more com­monly as­so­ci­ated with en­vi­ron­men­tal­ists, so­cial democrats, and pro­gres­sives, and (in the United States) with Democrats, whereas op­po­si­tion to it is more com­mon among con­ser­va­tives and liber­tar­i­ans. The min­i­mum wage be­lief is more com­monly as­so­ci­ated with free mar­ket views and (in the United States) with con­ser­va­tives and Repub­li­cans, and op­po­si­tion to it is more com­mon among pro­gres­sives and so­cial democrats.

Look­ing for help

I’m in­ter­ested in thoughts from the peo­ple here on these ques­tions:

  • Thoughts on the speci­fics of Ques­tion #1 and Ques­tion #2.

  • Other pos­si­ble ques­tions in the same refer­ence class (where a be­lief arises from a mix of the­ory and data, and the the­ory plays a fairly big role in driv­ing the be­lief, while the data on its own is very am­bigu­ous).

  • Other similar­i­ties be­tween Ques­tion #1 and Ques­tion #2.

  • Ways that Ques­tion #1 and Ques­tion #2 are dis­analo­gous.

  • Gen­eral thoughts on how this re­lates to Bayesian rea­son­ing and other modes of be­lief for­ma­tion based on a com­bi­na­tion of the­ory and data.