Advanced

Institutions, Inequality and Societal Transformations

Moricz, Sara LU (2019)
Abstract (Swedish)
Institutioner har betydelse för ekonomisk utveckling. Denna avhandling består av tre fristående artiklar som bidrar till olika grenar inom institutionell ekonomi.

I den första artikeln undersöks kortsiktiga förändringar av könsnormer. Artikeln tar tillvara på senare tids utveckling av maskininlärningsalgoritmer för att studera normförändringar i svenska tweets. Mer specifikt så tränas en LSTM neural network- modell till att definiera könsnormer på ett omfattande och datadrivet sätt. Modellen används som ett verktyg för att ta fram ett mått på könsnormer som människor möter i vardagslivet, vilket används för att studera normförändringar i samband med Metoo-rörelsen. Metoo startade den 17 Oktober 2017 efter att en amerikansk... (More)
Institutioner har betydelse för ekonomisk utveckling. Denna avhandling består av tre fristående artiklar som bidrar till olika grenar inom institutionell ekonomi.

I den första artikeln undersöks kortsiktiga förändringar av könsnormer. Artikeln tar tillvara på senare tids utveckling av maskininlärningsalgoritmer för att studera normförändringar i svenska tweets. Mer specifikt så tränas en LSTM neural network- modell till att definiera könsnormer på ett omfattande och datadrivet sätt. Modellen används som ett verktyg för att ta fram ett mått på könsnormer som människor möter i vardagslivet, vilket används för att studera normförändringar i samband med Metoo-rörelsen. Metoo startade den 17 Oktober 2017 efter att en amerikansk skådespelerska uppmuntrade kvinnor att dela med sig av sina erfarenheter av sexuellt ofredande. Datumet för händelsen är orelaterat till svenska könsnormer och därmed tillåter kontexten att en effekt skattas utan jämförelseproblem. Tweets som innehåller Metoo-hashtagen eller relaterade hashtags tas bort från analysen för att inte effekten av en mer intensiv könsdebatt ska mätas. Artikeln finner att svenska tweets reflekterar könsnormer i en mindre utsträckning sex månader efter Metoo-händelsen i jämförelse med fem månader innan. Föregående års normförändringar används som en jämförelsegrupp. Resultatet är robust gentemot att använda fixa effekter för kalenderdagar och Twitter-användare. Resultatet får även stöd genom att istället undersöka en orelaterad norm som inte borde förändras med placebo-test. Studien ger ett exempel på att normer kan förändras snabbt, vilket visar på att tidigare litteratur borde omvärdera sitt synsätt på normer som relativt oföränderliga.

Den andra artikeln undersöker effekten på ekonomisk tillväxt av att byta från ett indirekt (parlamentariskt) till ett direkt (presidentiellt) demokratiskt valsystem i Indonesien. Under indirekta val väljs den lokala ledaren av distriktsparlamenten och under direkta val väljs den lokala ledaren av befolkningen. Denna artikel avhjälper de jämförelseproblem som existerat i den tidigare litteraturen genom att studera förändringen i en specifik politisk institution. Förändringen av valsystemet skedde vid olika år för olika distrikt. Vilket år ett distrikt bytte valsystem beror på när den förre diktatorn Suharto installerade lokala ledare, och därmed förväntas inte distrikten ha några skillnader mellan sig som påverkar tillväxt utöver förändringen i valsystem. I artikeln jämförs tillväxt i distrikt som bytte valsystem tidigt gentemot de som bytte senare. Artikeln finner inga tillväxteffekter av förändringen. En möjlig förklaring till resultatet är att valsystem har ringa betydelse för förekomsten av pålitliga, opartiska och effektiva samhällsinstitutioner (governance). Artikeln finner stöd för detta genom att undersöka ett stort antal governance-indikatorer (t.ex. om den lokala ledaren anses korrupt). Resultatet pekar på att det inte finns några skillnader i tillväxt mellan indirekta och direkt valsystem, vilket kan användas vid framtida beslut angående institutionella ramverk.

Den tredje och sista artikeln estimerarar mått på inkomstojämlikhet i svenska kommuner för 1871 och 1892. Utvecklingen av ojämlikhet under 1800-talet är relativt okänd, speciellt med avseende på inkomst, och artikeln bidrar med en deskriptiv pusselbit från den tidiga fasen i den svenska industrialiseringsprocessen. De nya estimaten på kommunal nivå konstrueras från data på distributionen av röster vid kommunala val. I det kommunala valsystemet fick människor röster efter hur mycket de betalade i skatt, vilket gör det möjligt att härleda personers inkomst från datakällan. Artikeln finner att den rikaste procentenhetens andel av totalinkomsten inom den industriella sektorn ökar i mediankommunen, medan densamma inom jordbrukssektorn är oförändrad. Artikeln kompletterar den tidigare forskningen på förmögenhetsojämlikhet på nationell nivå genom att stärka resultatet angående ökande ojämlikhet under den senare delen av 1800-talet. Forskare kan använda de nya estimaten, tillgängliga under en open-source licens online, i vidare studier. (Less)
Abstract
Institutions matter for economic development. This thesis consists of three self-contained articles which provide different contributions to institutional economics.

The first article studies short-run changes in gender norms. It takes advantage of recent developments in machine learning algorithms to study changes in norms in Swedish tweets. More specifically, gender norms are defined in a comprehensive data-driven manner by training an LSTM neural network model. The model functions as a tool to create a measure on gender norms experienced by people in their everyday life, which is used to study norm-changes in relation to the Metoo movement. After an American actress encouraged women to share their experiences of sexual... (More)
Institutions matter for economic development. This thesis consists of three self-contained articles which provide different contributions to institutional economics.

The first article studies short-run changes in gender norms. It takes advantage of recent developments in machine learning algorithms to study changes in norms in Swedish tweets. More specifically, gender norms are defined in a comprehensive data-driven manner by training an LSTM neural network model. The model functions as a tool to create a measure on gender norms experienced by people in their everyday life, which is used to study norm-changes in relation to the Metoo movement. After an American actress encouraged women to share their experiences of sexual harassment under the Metoo hashtag, the Metoo movement spread to Sweden the 17th of October 2017. The event’s date of occurrence is unrelated to the Swedish gender norm environment and, thus, provides a credible context for identifying an effect. Tweets with the Metoo-hashtag and related hashtags are removed from the analysis to ensure not to capture the effect of a more intensive gender debate. The article shows that Swedish tweets reflect gender norms less six months after the Metoo event, compared to five months before. The norm changes of the previous year function as a comparison group. The result is robust to the inclusion of calendar day and Twitter-user fixed effects. The result is also supported with placebo tests on an unrelated norm which is not expected to change. The article provides an example of norms changing rapidly, raising questions about previous literature’s conceptualization of norms as being constant.

The second article investigates the effect of changing from an indirect (parliamentary) to a direct (presidential) democratic system on economic growth in Indonesia. In an indirect system, the local leader is appointed by the local parliament, whereas, in a direct system, the local leader is elected by the people. The main contribution of the article is to have a credible identification strategy to investigate the effect of a change in one specific political institution. The date of switching from the indirect to the direct democratic system differed for the various districts. The date of the switch is dependent on when the former dictator Suharto installed district heads, and therefore one expects no inherent differences in the growth rates of the districts other than from the change of system. The article relies on comparing districts that had direct elections earlier to districts that had them later. It shows no effect on economic growth. The results suggest that the form of democracy has a limited impact on governance, which is confirmed by examining a large number of indicators on governance (such as, the district head being corrupt). The result points to no effects of the forms of democracy on growth, which can inform policy regarding the choice of institutional design.

The third and last article presents new estimates of municipality-level income inequality for Sweden in 1871 and 1892. There exists scant information on the evolution of inequality during the nineteenth century, especially on income inequality, and the article contributes with a descriptive piece of evidence from the early phase of the Swedish industrialization process. The new municipality-level estimates are based on data on the vote distribution. In the electoral system at this time, votes were allocated in proportion to people’s income, which renders it possible to retrieve income figures from the data source. The article shows that the income share of the top one percent richest in the industrial sector in the median municipality increases between 1871 and 1892, whereas inequality in the agricultural sector is stable. The article complements prior research on national-level wealth inequality by strengthening the interpretation of increasing inequality in the second half of the nineteenth century. Researchers can use the new estimates, available under an open-access license online, in further studies. (Less)
Please use this url to cite or link to this publication:
author
supervisor
opponent
  • Associate Professor Kotsadam, Andreas, University of Oslo
organization
publishing date
type
Thesis
publication status
published
subject
keywords
Gender Norms, Metoo, Sweden, LSTM Neural Network, Asia, Indonesia, Democracy, Elections, Economic Growth, Income Inequality, Economic Development
pages
182 pages
publisher
Department of Economics, Lund University
defense location
Holger Crafoord Centre, EC3:207
defense date
2019-06-05 14:15
ISBN
978-91-7895-085-0
978-91-7895-084-3
language
English
LU publication?
yes
id
484dbdcf-2b2a-41da-9ee8-5eae2eee9807
date added to LUP
2019-05-10 10:45:14
date last changed
2019-05-13 10:34:24
@phdthesis{484dbdcf-2b2a-41da-9ee8-5eae2eee9807,
  abstract     = {Institutions matter for economic development. This thesis consists of three self-contained articles which provide different contributions to institutional economics.<br/><br/>The first article studies short-run changes in gender norms. It takes advantage of recent developments in machine learning algorithms to study changes in norms in Swedish tweets. More specifically, gender norms are defined in a comprehensive data-driven manner by training an LSTM neural network model. The model functions as a tool to create a measure on gender norms experienced by people in their everyday life, which is used to study norm-changes in relation to the Metoo movement. After an American actress encouraged women to share their experiences of sexual harassment under the Metoo hashtag, the Metoo movement spread to Sweden the 17th of October 2017. The event’s date of occurrence is unrelated to the Swedish gender norm environment and, thus, provides a credible context for identifying an effect. Tweets with the Metoo-hashtag and related hashtags are removed from the analysis to ensure not to capture the effect of a more intensive gender debate. The article shows that Swedish tweets reflect gender norms less six months after the Metoo event, compared to five months before. The norm changes of the previous year function as a comparison group. The result is robust to the inclusion of calendar day and Twitter-user fixed effects. The result is also supported with placebo tests on an unrelated norm which is not expected to change. The article provides an example of norms changing rapidly, raising questions about previous literature’s conceptualization of norms as being constant.<br/><br/>The second article investigates the effect of changing from an indirect (parliamentary) to a direct (presidential) democratic system on economic growth in Indonesia. In an indirect system, the local leader is appointed by the local parliament, whereas, in a direct system, the local leader is elected by the people. The main contribution of the article is to have a credible identification strategy to investigate the effect of a change in one specific political institution. The date of switching from the indirect to the direct democratic system differed for the various districts. The date of the switch is dependent on when the former dictator Suharto installed district heads, and therefore one expects no inherent differences in the growth rates of the districts other than from the change of system. The article relies on comparing districts that had direct elections earlier to districts that had them later. It shows no effect on economic growth. The results suggest that the form of democracy has a limited impact on governance, which is confirmed by examining a large number of indicators on governance (such as, the district head being corrupt). The result points to no effects of the forms of democracy on growth, which can inform policy regarding the choice of institutional design.<br/><br/>The third and last article presents new estimates of municipality-level income inequality for Sweden in 1871 and 1892. There exists scant information on the evolution of inequality during the nineteenth century, especially on income inequality, and the article contributes with a descriptive piece of evidence from the early phase of the Swedish industrialization process. The new municipality-level estimates are based on data on the vote distribution. In the electoral system at this time, votes were allocated in proportion to people’s income, which renders it possible to retrieve income figures from the data source. The article shows that the income share of the top one percent richest in the industrial sector in the median municipality increases between 1871 and 1892, whereas inequality in the agricultural sector is stable. The article complements prior research on national-level wealth inequality by strengthening the interpretation of increasing inequality in the second half of the nineteenth century. Researchers can use the new estimates, available under an open-access license online, in further studies.},
  author       = {Moricz, Sara},
  isbn         = {978-91-7895-085-0},
  keyword      = {Gender Norms,Metoo,Sweden,LSTM Neural Network,Asia,Indonesia,Democracy,Elections,Economic Growth,Income Inequality,Economic Development},
  language     = {eng},
  month        = {05},
  pages        = {182},
  publisher    = {Department of Economics, Lund University },
  school       = {Lund University},
  title        = {Institutions, Inequality and Societal Transformations},
  year         = {2019},
}