2.4.3.1 Natural gwajen

Natural gwajen yi amfani da bazuwar abubuwan a duniya. bazuwar taron + yaushe-on data tsarin = halitta gwaji

The kewayawa don yi da ka sarrafawa gwajen kunna gaskiya kwatanta ne randomization. Duk da haka, lokaci-lokaci wani abu ya faru a duniya wanda da gaske sanya mutane da ka ko kusan da ka ga m jiyya. Daya daga mafi tsabta misalai na dabarun yin amfani da halitta gwaje-gwajen ya zo daga bincike na Angrist (1990) cewa ƙaddara sakamako na soja ayyuka a kamfata.

A lokacin yaki a Vietnam, da Amurka ya karu da girman da sojojin ta a daftarin. Domin yanke abin da 'yan ƙasa za a kira zuwa sabis, gwamnatin {asar Amirka da aka gudanar a irin caca. Kowane ranar haihuwa da aka wakilta a kan wani yanki da takarda, kuma waɗannan takardun da aka sanya a babban gilashi tulu. Kamar yadda aka nuna a Figure 2.5, wadannan jirgin ruwa ake ɗaukar takarda aka kõma daga jar daya a lokaci domin sanin da umurni cewa samari za a kira su zuwa ga bauta (mata matasa suka ba batun da daftarin). Bisa ga sakamakon, maza haifi on Satumba 14 da aka kira na farko, maza haifi on Afrilu 24 da aka kira na biyu, da sauransu. Daga qarshe, a cikin wannan irin caca, maza haifi on 195 daban-daban days aka kira su zuwa ga aiki, yayin da mutane haifi on 171 days aka ba da ake kira.

Figure 2.5: majalisar Alexander Pirnie (R-NY) jawo farko kwantena ga zababben Service daftarin on Disamba 1, 1969. Joshua Angrist (1990) a hade da daftarin irin caca da albashi bayanai daga Social Security Administration to kimanta sakamakon soja a kan albashi. Wannan misali ne na bincike ta amfani da halitta gwaji. Source: Wikimedia Commons

Figure 2.5: majalisar Alexander Pirnie (R-NY) jawo farko kwantena ga zababben Service daftarin on Disamba 1, 1969. Joshua Angrist (1990) a hade da daftarin irin caca da albashi bayanai daga Social Security Administration to kimanta sakamakon soja a kan albashi. Wannan misali ne na bincike ta amfani da halitta gwaji. Source: Wikimedia Commons

Ko da yake shi ba su nan da nan ya bayyana, wani daftarin irin caca yana da m kama zuwa yi da ka sarrafawa gwaji: a biyu yanayi mahalarta suna da ka sanya a yi maka wani magani. A cikin yanayin da daftarin irin caca, idan mun kasance mũ ne sha'awar koyo game da sakamakon daftarin-cancanta da soja a kan m aiki kasuwa albashi, za mu iya kwatanta sakamakon ga mutanen da wanda birthdates kasance kasa da irin caca cutoff (misali, Satumba 14, Afrilu 24, da dai sauransu) da sakamakon ga mutanen da wanda Birthdays kasance bayan cutoff (misali, Fabrairu 20, Disamba 2, da dai sauransu).

Ganin cewa wannan magani da ake tsara da aka da ka sanya, za mu iya to auna sakamako na wannan magani ga wani sakamako da aka auna. Alal misali, Angrist (1990) a hade da bayanai game da wanda aka da ka zaba a cikin daftarin da albashi data da aka tattara ta Social Tsaro Administration a kammala da cewa albashi farin Tsohon soji sun kasance game da 15% kasa da albashi na m ba Tsohon soji . Other masu bincike sun yi amfani da irin wannan zamba da. Alal misali, Conley and Heerwig (2011) a hade da bayanai game da wanda aka da ka zaba a cikin daftarin da iyali data tattara daga shekarar 2000 Census da 2005 American Community Survey kuma gano cewa, saboda haka dogon bayan daftarin, akwai kadan dogon lokaci sakamako na soja a kan da dama sakamakon irin gidaje wa'adin mulki (mallakan a kan hayar) da kuma na zama kwanciyar hankali (alama ne na koma a baya shekaru biyar).

Kamar yadda wannan misali ya nuna, wani lokacin zamantakewa, siyasa, ko na halitta sojojin halitta gwajen ko kusa-gwajen da za a iya leveraged da masu bincike. Sau da yawa na halitta gwajen ne hanya mafi kyau don kimanta hanyar-da-sakamako dangantaka a saituna inda ba da'a ko m gudu yi da ka sarrafawa gwaje-gwajen. Su ne wani muhimmin dabarun domin ganowa gaskiya kwatancen a ba-gwaji data. Wannan bincike dabarun da za a iya takaita da wannan lissafi:

\ [\ rubutu {bazuwar (ko kamar yadda idan bazuwar) taron} + \ rubutu {yaushe-on data rafi} = \ rubutu {halitta gwaji} \ qquad (2.1) \]

Duk da haka, bincike na halitta gwajen iya zama quite tricky. Alal misali, a cikin akwati na Vietnam daftarin, ba kowa da kowa wanda yake daftarin-m ƙare har bauta (akwai da dama exemptions). Kuma, a lokaci guda, wasu mutane da suke ba zayyana-m da yardar ransa domin sabis. Shi ne kamar yadda idan a wani asibiti fitina da wani sabon magani, wasu mutane da magani kungiyar bai riƙi magani da kuma wasu daga cikin mutanen da a cikin iko kungiyar ko ta yaya samu da miyagun ƙwayoyi. Wannan matsalar, da ake kira biyu mai gefe noncompliance, kazalika da sauran matsaloli suna bayyana a zurfafe a wasu daga cikin shawarar karatu a karshen wannan babi.

The dabarun shan amfani da sauƙi abin da ke faruwa bazuwar aiki Earsbe digital shekaru, amma ruwan dare na babban data sa wannan dabarun da sauƙin amfani. Da zarar ka yi wani magani da aka sanya da ka, babban data kafofin iya samar da sakamako data cewa, kana bukatar in domin kwatanta results for mutane da magani da kuma kula da yanayi. Alal misali, a cikin binciken na da sakamakon da daftarin da soja, Angrist yi amfani da albashi records daga Social Security Administration. ba tare da wannan sakamako data, ya yi karatu ba zai yiwu. A wannan yanayin, da Social Security Administration ne ko da yaushe-on babban data Madogararsa. Kamar yadda kuma da ta atomatik tattara data kafofin zama, za mu iya samun karin sakamako data da za a iya auna da sakamakon canje-canje halitta exogenous bambancin.

To kwatanta wannan dabarun a cikin dijital shekaru, bari mu yi la'akari da Mas da Moretti ta (2009) m bincike a kan sakamako daga takwarorina kan yawan aiki. Ko da yake a kan surface shi zai duba daban-daban fiye da Angrist ta binciken game da sakamakon da Vietnam Daftarin, a tsarin su biyu bi da juna a Eq. 2.1.

Mas da Moretti auna yadda takwarorina shafa aiki na ma'aikata. A daya hannun, da ciwon wuya aiki tsara zai kai ma'aikata don su kara yawan aiki saboda matsi na tsara. Ko kuma, a daya bangaren, a wuya aiki tsara zai kai wasu ma'aikatan ku yi rauni kashe fi. The sarari hanyar nazarin tsara effects on yawan aiki zai zama wani yi da ka sarrafawa gwaji inda ma'aikata suna da ka sanya wa canjawa da ma'aikata daban-daban yawan matakan, sa'an nan kuma sakamakon himmar aiki ne auna ga kowa da kowa. Masu bincike, duk da haka, ba su sarrafa jadawalin ma'aikata a cikin wani real aiki, don haka Mas da Moretti ya dogara a kan wani halitta gwaji abin da ya faru a cikin wani babban kanti.

Kamar Eq. 2.1, da binciken da sassa biyu. Na farko, su yi amfani da rajistan ayyukan daga kanti biya tsarin a yi daidai, mutum, kuma ko da yaushe-on ma'aunin yawan aiki: yawan abubuwa leka ta biyu. Kuma, na biyu, saboda hanyar da tanadi da aka yi a wannan kanti, suna da kusa bazuwar abun da ke ciki na takwarorina. A wasu kalmomin, kuma kõ da tanadi na cashiers ba m da wani irin caca, shi ne da gaske bazuwar. A yi, da amincewa da muke da a cikin halitta gwaje-gwajen da akai-akai hinges a kan plausibility wannan "as-idan" bazuwar da'awar. Shan amfani da wannan bazuwar bambancin, Mas da Moretti gano cewa aiki da hakan yawan takwarorina qara yawan aiki. Bugu da ari, Mas da Moretti amfani da size da richness da dataset don motsawa a hayin hakkin na hanyar-da-sakamako gano biyu mafi muhimmanci da kuma dabara al'amurran da suka shafi: heterogeneity wannan sakamako (ga abin da iri ma'aikata ne sakamako ya fi girma) da kuma inji baya da sakamako (dalilin da ya sa ya aikata ciwon high yawan aiki takwarorina kai ga mafi girma yawan aiki). Za mu mayar da wadannan biyu da muhimmanci al'amurran da suka shafi-heterogeneity magani illa da kuma sunadaran-a Babi na 5 a lokacin da muka tattauna gwaje-gwajen a more daki-daki.

Generalizing daga karatu a kan sakamako daga cikin Vietnam Daftarin kan albashi da kuma nazarin sakamakon takwarorina kan yawan aiki, Table 2.3 takaita wasu karatu da cewa suna da wannan ainihin wannan tsarin: ta yin amfani da ko da yaushe-on data source don auna da tasiri na wasu taron . As Table 2.3 ta bayyana, halitta gwajen ne a ko'ina, idan kun m san yadda za su kama su.

Table 2.3: Misalan halitta gwaje-gwajen yin amfani da babban data kafofin. Duk wadannan karatu bi wannan asali girke-girke: bazuwar (ko kamar yadda idan bazuwar) taron + yaushe-on data tsarin. Dubi Dunning (2012) don ƙarin misalai.
gudunmawata mayar da hankali Source na halitta gwaji Koyaushe-on data source lissafi
Peer effects on aiki tanadi tsari biya data Mas and Moretti (2009)
Friendship samuwar mahaukaciyar guguwa Facebook Phan and Airoldi (2015)
Yada motsin zuciyarmu ruwan sama Facebook Coviello et al. (2014)
Tsara don sa'a tattalin arziki yana canja wurin girgizar kasa mobile kudi data Blumenstock, Fafchamps, and Eagle (2011)
Personal amfani hali 2013 gwamnatin Amirka kashewa sirri finance data Baker and Yannelis (2015)
Tattalin tasiri na recommender tsarin daban-daban browsing data at Amazon Sharma, Hofman, and Watts (2015)
Effect danniya a kan unborn jariran 2006 Isra'ila-Hezbollah yaki Birth records Torche and Shwed (2015)
Reading hali a Wikipedia Snowden ayoyin wikipedia rajistan ayyukan Penney (2016)

A yi, masu bincike amfani biyu daban-daban dabarun samun halitta gwaje-gwajen, da abin da zai iya zama hayayyafa. Wasu masu bincike fara da yaushe-on data source kuma nemi bazuwar events a duniya. wasu fara da bazuwar abubuwan a cikin duniya, kuma nemi data kafofin cewa kama su bugu. A karshe, lura cewa da ƙarfin halitta gwajen zo ba daga sophistication na ilimin kididdiga analysis, amma daga care a ganowa a gaskiya kwatanta halitta a sa'a hatsari na tarihi.