4.4.2 Heterogeneity magani effects

Gwaje-gwajen kullum auna talakawan sakamako, amma sakamako na iya zama daban-daban ga mutane daban-daban.

Na biyu key ra'ayin domin motsi bayan sauki gwajen ne heterogeneity magani effects. The gwajin na Schultz et al. (2007) powerfully misalta yadda wannan magani zai iya samun daban-daban effects a kan daban-daban irin mutane (Figure 4.4), amma wannan bincike na heterogeneity ne ainihin quite sabon abu ga wani analog shekaru gwaji. Mai analog shekaru gwajen unsa karamin yawan mahalarta da cewa suna bi da matsayin m "Widgets" domin kadan game da su da aka sani pre-jiyya. A digital gwaje-gwajen, duk da haka, wadannan data constraints ne ƙasa da na kowa, domin masu bincike ayan da karin mahalarta da kuma sani game da su. A cikin wannan daban-daban data yanayi, za mu iya kimanta heterogeneity magani effects domin samar da alamu game da yadda magani ke aiki, yadda za a iya inganta, da kuma yadda za a iya yi niyya ga waɗanda mafi yawa wata ila ya amfana.

Misalai biyu na heterogeneity magani effects a cikin mahallin zamantakewa norms da makamashi amfani zo daga m bincike a kan Home Energy Rahotanni. Na farko, Allcott (2011) sun yi amfani da manyan sample size (600,000 gidaje) don kara raba sample kuma kimanta sakamakon da Home Energy Report da decile na pre-jiyya makamashi amfani. Duk da yake Schultz et al. (2007) sami bambance-bambance a tsakanin nauyi da haske users, Allcott (2011) gano cewa akwai kuma bambance-bambance a cikin nauyi da haske mai amfani kungiyar. Alal misali, heaviest users (waɗanda suke a cikin top decile) rage m makamashi amfani sau biyu kamar yadda wani a cikin tsakiyar nauyi mai amfani tara (Figure 4.7). Bugu da ari, kimantawa da sakamako daga pre-jiyya hali kuma bayyana cewa akwai wani boomerang sakamako har ga lightest users (Figure 4.7).

Figure 4.7: Heterogeneity magani effects a Allcott (2011). The karu a makamashi amfani ya daban-daban domin mutane daban-daban a deciles na baseline cutarwa.

Figure 4.7: Heterogeneity magani effects a Allcott (2011) . The karu a makamashi amfani ya daban-daban domin mutane daban-daban a deciles na baseline cutarwa.

A cikin wata related binciken, Costa and Kahn (2013) Jita-jitar da tasiri na Home Energy Report iya bambanta dangane da wani ɗan takara ta siyasa akidar da kuma cewa magani zai zahiri sa mutane da wasu akidu ƙara da wutar lantarki amfani. A takaice, suka lissaftuwa cewa Home Energy Rahotanni iya samar da wani boomerang sakamako ga wasu iri mutane. Don tantance wannan yiwuwar, Costa da Kahn garwaya da Opower data tare data saya daga wani ɓangare na uku aggregator cewa hada bayani kamar jam'iyyar siyasa rajista, kyaututtuka, don muhalli kungiyoyi, da kuma iyali hallara a sabunta makamashi shirye-shirye. Da wannan merged dataset, Costa da Kahn gano cewa Home Energy Rahotanni samar broadly m effects ga mahalarta da daban-daban akidu. babu shaidar da cewa duk wani rukunin nuna boomerang effects (Figure 4.8).

Figure 4.8: Heterogeneity magani effects a Costa da Kahn (2013). The kiyasta talakawan magani sakamako ga dukan sample ne -2,1% [-1.5%, -2,7%. By hada bayanai daga gwaji tare da bayani game da gidaje, Costa da Kahn (2013) ya yi amfani da jerin ilimin kididdiga model to kimanta lura sakamako sosai musamman kungiyoyin jama'a. Biyu kimomi aka gabatar ga kowane rukuni saboda kimomi dogara ne a kan covariates suka hada da a cikin ilimin kididdiga model (duba Model 4 da model 6 a Table 3 Table 4 a Costa da Kahn (2013)). Kamar yadda wannan misali ya nuna, magani effects iya zama daban-daban ga mutane daban-daban da kuma kimomi na lura effects da ya zo daga ilimin kididdiga model iya dogara a kan cikakken bayani game da wadanda model (Grimmer, messing, kuma Westwood 2014).

Figure 4.8: Heterogeneity magani effects a Costa and Kahn (2013) . The kiyasta talakawan magani sakamako ga dukan sample ne -2,1% [-1.5%, -2,7%. By hada bayanai daga gwaji tare da bayani game da gidaje, Costa and Kahn (2013) ya yi amfani da jerin ilimin kididdiga model to kimanta lura sakamako sosai musamman kungiyoyin jama'a. Biyu kimomi aka gabatar ga kowane rukuni saboda kimomi dogara ne a kan covariates suka hada da a cikin ilimin kididdiga model (duba Model 4 da model 6 a Table 3 Table 4 a Costa and Kahn (2013) ). Kamar yadda wannan misali ya nuna, magani effects iya zama daban-daban ga mutane daban-daban da kuma kimomi na lura effects da ya zo daga ilimin kididdiga model iya dogara a kan cikakken bayani game da wadanda model (Grimmer, Messing, and Westwood 2014) .

Kamar yadda wadannan misalai biyu kwatanta, a cikin digital shekaru, za mu iya motsawa daga kimantawa talakawan magani effects zuwa kimantawa da heterogeneity magani effects domin mun iya samun mutane da yawa more mahalarta da kuma mu sani game da wadanda mahalarta. Koyo game da heterogeneity magani effects iya taimaka niyya mai magani inda shi ne mafi inganci, samar facts cewa ta sabon ka'idar raya kasa, da kuma samar da alamu game da m inji, da topic to abin da na yanzu juya.