Geför­dert durch:


Abs­tract: Bren­tel, I. (2018, 28.06.). How to Make Expe­ri­ments Repli­ca­ble – A Case Stu­dy of Online Poli­ti­cal Stock Mar­kets (OPSM) in Ger­ma­ny. WAPOR 71st Annu­al Con­fe­rence, Worl Asso­cia­ti­on for Public Opi­ni­on Rese­arch (WAPOR). Mar­ra­k­ech, Marok­ko. [Tan­dem 1]

Rele­van­ce & Rese­arch Ques­ti­on:
Elec­tion rese­arch stu­dies have asked whe­ther sur­veys influ­ence the electorate’s expec­ta­ti­ons of elec­tion results and what effects have risen from the­se expec­ta­ti­ons of voting decisi­ons. Theo­re­ti­cal assump­ti­ons have been empi­ri­cal­ly exami­ned with dif­fe­rent expe­ri­men­tal designs. In this con­text among others the fol­lowing points are cri­ti­cal­ly seen:

  1. Metho­do­lo­gi­cal aspects: the electorate’s expec­ta­ti­ons of elec­tion results via ques­tio­ning; the mea­su­re­ment of the electorate’s expec­ta­ti­ons of elec­tion results takes place only once, but is a dyna­mic process.
  2. Inter­sub­jec­ti­vi­ty of expe­ri­ments: pro­blems in docu­men­ta­ti­on (esp. the inten­ti­on of the theo­re­ti­cal base regar­ding ope­ra­tio­na­li­sa­ti­on, mea­su­re­ments and tre­at­ments); This leads to a lack of repli­ca­ti­on and com­pa­ra­ti­ve studies.

With this paper we sug­gest a solu­ti­on for the­se pro­blems. We will dis­cuss how we can make expe­ri­ments repli­ca­ble and acces­si­ble for other rese­ar­chers and fur­ther for repli­ca­ti­ons as well as com­pa­ra­ti­ve stu­dies. This will be done by showing an alter­na­ti­ve data resour­ce for public opi­ni­on data. The case study’s ques­ti­ons whe­ther expe­ri­men­tal designs wit­hin OPSMs open up fur­ther metho­do­lo­gi­cal pos­si­bi­li­ties to inves­ti­ga­te the impact of sur­veys on the expec­ta­ti­ons of elec­tion results using a lon­gi­tu­di­nal data source. Ther­eby, OPSMs ope­ra­te accord­ing to the fol­lowing princip­le: The con­ver­gence of many par­ti­ci­pants leads to rea­listic pri­ces of objects (can­di­da­tes and par­ties) which can be com­pa­red with poll results and used as elec­tion forecast.

Methods & Data:
In order to eva­lua­te the rese­arch ques­ti­on of our case stu­dy, we con­duc­ted an expe­ri­ment on the influ­ence of publis­hed poll data on voting beha­viour using an OPSM on the Bun­des­tag elec­tions 2013 pro­vi­ded by a Ger­man natio­nal news­pa­per. Bet­ween May 1 and July 17, 2013, 413 acti­ve par­ti­ci­pants were ran­do­mi­zed and divi­ded into four groups. The­se groups were shown dif­fe­rent pre­sen­ta­ti­ons of the par­ties’ sur­vey results (short and long sur­vey trend, dai­ly cross sec­tion, no sur­vey). The dai­ly num­ber of tra­ding acti­vi­ties, the kind of tran­sac­tion, and the tran­sac­tion pri­ce gene­ra­ted the depen­dent varia­bles.
In making expe­ri­men­tal design (mea­su­re­ment con­si­de­ra­ti­ons, tre­at­ments, etc.)  visi­ble and thus data acces­si­ble for aca­de­mic rese­arch we used a high­ly inno­va­ti­ve digi­tal docu­men­ta­ti­on and har­mo­niz­a­ti­on soft­ware, Charm­Stats, to reach the second goal of our pro­ject. Charm­Stats allows us to docu­ment expe­ri­ments and other data in high qua­li­ty stan­dards of data documentation.

One tar­get of the pro­ject is to make the com­plex pro­cess of data with expe­ri­men­tal designs most trans­pa­rent and repli­ca­ble. Charm­Stats offers the pos­si­bi­li­ty to ful­fil the project´s goals to enab­le rese­ar­chers to repro­du­ce the pro­ce­du­re of an expe­ri­men­tal design with the deve­lo­p­ment of defi­ni­ti­ons to ope­ra­tio­na­li­sa­ti­on towards mea­su­re­ments and tre­at­ments. Addi­tio­nal­ly, Charm­Stats out­puts a report for docu­men­ta­ti­on. Moreo­ver, syn­ta­xes for har­mo­ni­sa­ti­ons, inde­xes and aggre­ga­ti­on of varia­bles can be com­pi­led for dif­fe­rent sta­tis­ti­cal soft­ware.
Final­ly, the results of the stu­dy show that the tra­ding acti­vi­ties dif­fer bet­ween the groups and tra­ding acti­vi­ties depen­ded on chan­ges in polls. With OPSM expec­ta­ti­ons of elec­tion out­co­me could be mea­su­red in form of real actions as a result of know­ledge about sur­vey results. Bes­i­des, tem­po­ral dyna­mics of dif­fe­rent expec­ta­ti­ons could also be map­ped longitudinal.