Dans ma thèse, je me sers de modèles de recherche solides pour répondre à des questions importantes de politique publique. Mon premier chapitre évalue l’impact causal de l’allégeance partisane (républicain ou démocrate) des gouverneurs américains sur le marché du travail. Dans ce chapitre, je combine les élections des gouverneurs avec les données du March CPS pour les années fiscales 1977 à 2008. En utilisant un modèle de régression par discontinuité, je trouve que les gouverneurs démocrates sont associés à de plus faibles revenus individuels moyens. Je mets en évidence que cela est entrainée par un changement dans la composition de la main-d’oeuvre à la suite d’une augmentation de l’emploi des travailleurs à revenus faibles et moyens. Je trouve que les gouverneurs démocrates provoquent une augmentation de l’emploi des noirs et de leurs heures travaillées. Ces résultats conduisent à une réduction de l’écart salarial entre les travailleurs noir et blanc. Mon deuxième chapitre étudie l’impact causal des fusillades qui se produisent dans les écoles secondaires américaines sur les performances des éléves et les résultats des écoles tels que les effectifs et le nombre d’enseignants recruté, a l’aide d’une stratégie de différence-en-différence. Le chapitre est coécrit avec Dongwoo Kim. Nous constatons que les fusillades dans les écoles réduisent significativement l’effectif des élèves de 9e année, la proportion d’élèves ayant un niveau adéquat en anglais et en mathématiques. Nous examinons aussi l’effet hétérogene des tueries dans les écoles secondaires entre les crimes et les suicides. Nous trouvons que les fusillades de natures criminelles provoquent la diminution du nombre d’inscriptions et de la proportion d’élèves adéquats en anglais et mathématiques. En utilisant des données sur les élèves en Californie, nous confirmons qu’une partie de l’effet sur la performance des élèves provient des étudiants inscrits et ce n’est pas uniquement un effet de composition. Mon troisième chapitre étudie l’impact des cellulaires sur la performance scolaire des élèves. Le chapitre est coécrit avec Richard Murphy. Dans ce chapitre, nous combinons une base de données unique contenant les politiques de téléphonie mobile des écoles obtenues à partir d’une enquète auprès des écoles dans quatre villes en Angleterre avec des données administratives sur la performance scolaire des éleves. Nous étudions ainsi l’impact de l’introduction d’une interdiction de téléphonie mobile sur le rendement des éleves. Nos résultats indiquent qu’il y a une augmentation du rendement des éleves après l’instauration de l’interdiction des cellulaires à l’école, ce qui suggère que les téléphones mobiles sont sources de distraction pour l’apprentissage et l’introduction d’une interdiction à l’école limite ce problème. / In my thesis, I use compelling research designs to address important public policy issues. My first chapter estimates the causal impact of the party allegiance (Republican or Democratic) of U.S. governors on labor market outcomes. I match gubernatorial elections with March CPS data for income years 1977 to 2008. Using a regression discontinuity design, I find that Democratic governors are associated with lower average individual earnings. I provide evidence that this is driven by a change in workforce composition following an expansion in employment of workers with low and medium earnings. I also find that Democratic governors cause a reduction in the racial earnings gap between black and white workers through an increase in the annual hours worked by blacks relative to whites. My second chapter analyze how shootings in high schools affect schools and students using data from shooting databases, school report cards, and the Common Core of Data. The chapter is co-written with Dongwoo Kim. We examine schools’ test scores, enrollment, and number of teachers, as well as graduation, attendance, and suspension rates at schools that experienced a shooting, employing a difference-in-differences strategy that uses other high schools in the same district as the comparison group. Our findings suggest that homicidal shootings significantly decrease the enrollment of students in Grade 9, and reduce test scores in math and English. We find no statistically significant effect for suicidal shootings on any outcome variables of interest. Using student-level data from California, we confirm that some of the effects on student performance occur as a result of students remaining enrolled and not only due to changes in student body composition. My third chapter investigates the impact of school mobile phone policy on student performance. The chapter is co-written with Richard Murphy. Combining a unique dataset on autonomous mobile phone policies from a survey of schools in four cities in England with administrative data, we investigate the impact of imposing a mobile phone ban on student performance. Our results indicate an improvement in student results after a school bans the use of mobile phones; this suggests that mobile phones distract learning and imposing a ban limits this problem.

<p>This dissertation consists of four self-contained essays.</p><p>Essay I (with Cecilia Enström Öst) investigates whether housing allowance affects recipients’ tenure choice in Sweden. A two-stage conditional maxi-mum likelihood probit (2SCMLP) model is applied in a panel data setting to simultaneously control for individual heterogeneity, state dependence and endogeneity. The empirical study is based on administrative data of housing allowance recipients between the years 1994 and 2002. Our results indicate that the housing allowance positively affects recipients’ homeownership propensity in Sweden. </p><p>Essay II investigates whether the Swedish housing allowance system creates dependence on welfare in recipients. Using longitudinal data from Swedish micro database-LINDA, this paper found that there is no evidence of nega-tive duration dependence among the Swedish housing allowance spells. This finding is consistent across different model specifications and various con-trols of the heterogeneity issue. </p><p>Essay III analyzes the impacts of the 1997 reform of Swedish housing al-lowance system on affected recipients’ exit hazards using the DD (differ-ence-in-difference) estimation strategy. This paper found strong evidence that the 1997 reform positively shifted up the conditional exiting probability of the couple with children recipient group, and the estimated magnitude of impact is sizable.</p><p>Essay IV extends the VECM (Vector Error Correction Cointegration Model) and PT (permanent-transitory) variance decomposition framework proposed by Lettau & Ludvigson (2004) to a situation in which total wealth is disag-gregated into housing wealth and financial wealth. The empirical studies are based on the Swedish aggregate quarterly data spanning from 1980q1 to 2004q4. We found strong statistical evidence that the movements of total consumption expenditures, disposable income, housing wealth and financial wealth are tied together. It is also shown that the movements of housing wealth in Sweden contain a large proportion of transitory component. </p>

This dissertation consists of four self-contained essays. Essay I (with Cecilia Enström Öst) investigates whether housing allowance affects recipients’ tenure choice in Sweden. A two-stage conditional maxi-mum likelihood probit (2SCMLP) model is applied in a panel data setting to simultaneously control for individual heterogeneity, state dependence and endogeneity. The empirical study is based on administrative data of housing allowance recipients between the years 1994 and 2002. Our results indicate that the housing allowance positively affects recipients’ homeownership propensity in Sweden. Essay II investigates whether the Swedish housing allowance system creates dependence on welfare in recipients. Using longitudinal data from Swedish micro database-LINDA, this paper found that there is no evidence of nega-tive duration dependence among the Swedish housing allowance spells. This finding is consistent across different model specifications and various con-trols of the heterogeneity issue. Essay III analyzes the impacts of the 1997 reform of Swedish housing al-lowance system on affected recipients’ exit hazards using the DD (differ-ence-in-difference) estimation strategy. This paper found strong evidence that the 1997 reform positively shifted up the conditional exiting probability of the couple with children recipient group, and the estimated magnitude of impact is sizable. Essay IV extends the VECM (Vector Error Correction Cointegration Model) and PT (permanent-transitory) variance decomposition framework proposed by Lettau &amp; Ludvigson (2004) to a situation in which total wealth is disag-gregated into housing wealth and financial wealth. The empirical studies are based on the Swedish aggregate quarterly data spanning from 1980q1 to 2004q4. We found strong statistical evidence that the movements of total consumption expenditures, disposable income, housing wealth and financial wealth are tied together. It is also shown that the movements of housing wealth in Sweden contain a large proportion of transitory component.

北京市為解決嚴重的道路堵塞問題，自2010年12月23日實施汽車限購政策。北京市於2010年10月傳出將要實施汽車限購政策，而本研究使用差中差方法估計消息傳出至開始實施日期間，政策消息對於汽車銷售量之影響。 研究中所使用之汽車銷售資料來源為中國「車輛登記管理所」中的「全中國登錄各月新增車輛」，以及「太平洋汽車網」的汽車詳細特性資料整理而成。以北京市為實驗組，其他非北京城市，與重慶市、天津市等為對照組，在控制了汽車特性以及加入了時間趨勢之後，可從實證結果得知，不論是以哪些城市當作對照組，在這段期間內，北京市的汽車銷售量皆有顯著的增加，與預測的結果相符。 在研究最後進一步使用穩定性檢驗，證明北京市汽車銷售量的突然增加是否真的為政策消息影響。最後結果亦與預期相符，證實北京市的汽車銷售量突然增加是由汽車限購政策消息所影響。 / To solve the serious road congestion problems, Beijing has implemented the restrictions on car ownership. In October, 2010, news started to spread that Beijing would bring the automobile purchase limit policy into practice. This study, using a method named “difference-in-difference”, estimated the policy’s impact on automobile sales volume in the period from when the news was launched to the day the policy was actually implemented. The data in this paper come from “Monthly Increase in Registered Vehicles Across China” of “Vehicle Registration and Management Bureau of China”, and the detailed characteristic data of vehicles from “Pacific Automobile Network”. We set Beijing as the experimental group, and the non-Beijing cities such as Chongqing and Tianjin as the control group. After controlling the characteristics of vehicles and adding time trend, we can learn from the experimental result that no matter which city is used as the control group, during the period, the vehicle sales volumes in Beijing increase significantly, which meets our expectation. In the last step of the research, stability test is used to prove whether the sudden increase of Beijing’s automobile sales volume is affected by the spread news. And the final result is also in accord with our expectation: the sudden increase in Beijing’s automobile sales volume is affected by the spread news of the automobile purchase limit policy.

本文旨在評估所有本國銀行自 1999 年至 2011 年的績效，以檢驗我國兩次金融改革的成效。研究方法主要採用 Olley and Pakes (1996) 的三階段模型，再者運用差異中之差異法 (Difference in Difference，DID)。有別於相對於過去文獻中使用非計量的方式，或是以計量方式評估但卻未考慮到銀行進出的情況，本文考量因新設或是合併而解散所造成的銀行家數變化影響，並試圖歸納金融改革對於整體本國銀行之成效，以及對於同一本國銀行的生產力提升是否有所助益。

Comme un déterminant de la survie d'une banque durant la crise financière de 2007/2008, la liquidité dans le secteur bancaire a depuis récemment représenté un défi pour les communautés financières et universitaires. Les trois articles présentés dans cette thèse portent sur les deux principales facettes de la liquidité dans le secteur bancaire: la détention d'actifs liquides (à savoir, la trésorerie et les ressources assimilées) et le processus de création de la liquidité dans les banques utilisé pour financer des prêts. Comme on le verra dans les articles, ces deux aspects de la liquidité peuvent être considérés comme les deux faces d'une même pièce. Je reconnais que la liquidité dans le secteur bancaire est liée à la création monétaire; cependant, cette thèse se concentre sur les deux précités aspects de la liquidité. Tout d'abord, cette introduction présente comment le concept de la liquidité a évolué dans la pensée économique dominante. La seconde partie considère le renouveau de la détention de cash qui a été observée depuis la crise financière de 2007/2008 dans le secteur bancaire. La troisième section examine les propriétés de liquidité. La quatrième section explore ce que nous ne savons pas sur la liquidité. La cinquième section identifie et sélectionne trois problèmes fondamentaux relatifs à liquidité et qui sont analysés dans les trois articles présentés dans thèse. La sixième et dernière section présente la méthodologie utilisée dans les trois articles pour répondre à ces questions. Chapitre 1 : “Why do banks hold cash ?". La détention de cash et assimilé cash par les banques détiennent est devenue un enjeu majeur depuis la crise financière de 2008 qui a démontré que la trésorerie retenue est un déterminant majeur dans les chances de survie des banques. Cet article examine les déterminants de la détention de cash banque en utilisant des données internationales pour la période 1981-2014. Sur la base d'un grand échantillon, nous documentons une augmentation séculaire de la détention de cash par les banques pendant une période de 35 ans. Nous apportons la preuve que la nature optimale dynamique de la détention de cash est rejetée dans le secteur bancaire. Ces résultats contrastent avec le secteur non bancaire, où la nature optimale dynamique de trésorerie est observée. Chapitre 2: “Does an increase in capital negatively impact banking liquidity creation?”. A partir d'un ensemble de données composé d'un panel de 940 banques cotées des pays européens, américains et asiatiques, cet article documente l'évolution de la création de la liquidité bancaire au cours d'une période de 35 ans (1981-2014). La preuve empirique confirme que les niveaux de risque et de capital jouent un rôle significatif et négatif dans la création de liquidité par les banques. Dans l'ensemble, les effets négatifs de l’augmentation de capital sur la création de la liquidité bancaire sont plus importants que les effets positifs sur la gestion du risque correspondant, ce qui suggère que les exigences de fonds propres imposées pour soutenir la stabilité financière affectent négativement la création de liquidités. Ces résultats ont de larges implications pour les régulateurs bancaires. Chapitre 3: “Positive effects of Basel III on banking liquidity creation”. Ce document évalue l'effet du cadre réglementaire de Bâle III sur la création de liquidité bancaire. Les résultats sont basés sur un ensemble de données de panel de banques américaines qui représentent environ 60% des prêts et dépôts américains sur une période de 7 ans (2009-2015), en plus de différence dans la différence et les méthodes de survie standard. Tous les composants de Bâle III pris ensemble, il existe des preuves empiriques que Bâle III a un effet positif sur la création de liquidité bancaire sur le marché américain, en particulier pour les grandes banques. Ces résultats ont de larges implications pour les régulateurs bancaires. / As one determinant of a bank’s survival during the financial crisis of 2007-2008, liquidity in the banking sector presents a challenge for the financial and academic communities and has recently become a central point of interest. The three articles presented in this thesis focus on the two main facets of liquidity in the banking sector: the holding of liquid assets (i.e., cash and assimilated resources) and the process of liquidity-creation in banks used to fund loans. As will be discussed in the articles, these two aspects of liquidity can be viewed as two sides of the same coin. I acknowledge that liquidity in banking is linked to the creation of money; however, this thesis focuses on the aforementioned two aspects of liquidity. First, this section presents how ideas about liquidity in the banking sector have evolved in mainstream economic thought. Second, it considers the revival of cash-holding that has been observed since the financial crisis of 2007-2008. Third, it discusses the properties of liquidity. Fourth, it explores what we do not know about liquidity. Fifth, it identifies the fundamental issues analyzed in the three articles. Finally, it presents the methodology used in the articles to address these issues. Chapter1: “Why do banks hold cash ?”. This paper investigates the determinants of bank cash holding by using international data for the period 1981-2014. The results do not seem to provide support for the substitutability hypothesis regarding the substitutive relation between cash and debt levels. Further, using the GMM-system estimation method, we find no support for the dynamic optimal cash model, suggesting that cash management in the banking sector is bounded by number of constraints that make it difficult for the agents to optimize their utility. Chapter 2: “Does an increase in capital negatively impact banking liquidity creation?”. From a dataset composed of a panel of 940 listed banks based in European, American and Asian countries, this paper documents the evolution of bank liquidity creation over a 35-year period (1981-2014). The empirical evidence confirms that risk and equity levels play a significant and negative role. Overall, the negative effects of equity increases on bank liquidity creation are more significant than corresponding positive effects on risk management, suggesting that capital requirements imposed to support financial stability negatively affect liquidity creation. These findings have broad implications for policymakers. Chapter 3: “Positive effects of Basel III on banking liquidity creation”. This paper estimates the effect of the Basel III regulatory framework on banking liquidity creation. The results are based on a panel data set of U.S. banks that represent approximately 60% of U.S. loans and deposits over a 7-year period (from 2009 to 2015) in addition to difference-in-difference and standard survival methods. All components of Basel III taken together, there is empirical evidence that Basel III has a positive effect on banking liquidity creation in the US market in particular for major banks. These findings have broad implications for policy makers.

近年臺北市政府試圖藉由興建品質良好的公共住宅，解決住宅價格高漲影響人民居住的問題，然附近居民往往認為公共住宅會造成周圍住宅生活環境品質低落及房價下跌，反對其於住家鄰近興建。過去國內缺乏以實證方式計算公共住宅對周圍住宅價格之影響，亦尚未針對新推出的公營出租住宅個案深入分析；而國外相關研究雖有利用量化實證公共住宅與周圍房價關係，然多針對個別公共住宅社區為研究對象，對不同類型公共住宅及其對不同高低房價住宅之影響與差異，缺乏更全面的探討分析，且亦未釐清住宅價格的空間自相關特性及公共住宅設立的不同階段等，對周圍房價影響效果的差異。 本文應用特徵價格理論傳統迴歸模型、空間迴歸模型及分量迴歸模型，以2012年7月至2015年9月之臺北市不動產實價登錄交易資料為對象，研究整建住宅、出售國宅、出租國宅、公營出租住宅等四類不同時期公共住宅，於100~300公尺不同影響範圍下，分別對周圍住宅價格的影響效果。另外並選取興隆公共住宅作為個案研究對象，以臺北市文山區不動產實價登錄交易資料為對象，應用差異中之差異法結合空間迴歸模型，分析興隆公共住宅的興建完工與住戶入住對於周圍住宅價格之外溢效果。實證結果顯示，附近有早期興建之整建住宅及出租國宅者，其住宅總價皆呈現顯著下跌現象，分別下跌約1.7%至8.9%及1.5%至7.7%；附近有出售國宅之住宅總價漲跌情形不顯著；而近期興建之公營出租住宅則造成周圍住宅價格顯著上漲約1.7%至7.7%；又前述影響程度皆隨影響範圍擴大而隨之減輕。進一步分析不同影響範圍下對周圍高低房價住宅的影響，公營出租住宅對周圍低價住宅正向提升效果為6.2%至7.1%，皆大於高價住宅之無顯著影響至3.2%。而興隆公共住宅的興建完工確實能使周圍住宅價格較外圍地區多提升約6.5%至15.8%，惟住戶入住興隆公共住宅則未具顯著正向的價格外溢效果。由新興辦的公營出租住宅對周圍房價造成正面影響的實證結果，應可釐清公共住宅外部性之正向效果，有助後續公共住宅的興建。 / In recent years, government has tried to build the public housing with high quality to solve the living problems caused by the soaring of the house price. However, residents beside the public housing often regard them as Nimbys that will detrimentally affect property values and oppose public housing to be built nearby. There were short of empirical studies that focus on the effects of public housing on nearby property, as well as the case study of social housing newly built in Taiwan. Besides, the empirical studies in foreign only select a specific public housing as their subject, which few of them analyze different effects caused by different types of public housing, ignore different effects on nearby property with different values, and have not clarify whether the spatial autocorrelation of property values as well as the construction and operation of the social housing leads to different results. This study analyzes the effect of social housing, resettled tenement, public housing for rent and for sale on nearby property values in different ranges separately by using hedonic price theory OLS, spatial and quantile regression as model, and selecting the sale price of real estate in Taipei City from July 2012 to September 2015 as sample. Besides, Xing-Long Social housing and sale prices of real estate in Wenshan district of Taipei City are also selected for case study applying Difference in Difference method with spatial regression, in order to analyze the nearby property values changes after the construction and the move in of the residents of Xing-Long social housing. The result shows that although resettled tenement and public housing for rent detrimentally affect the property values by 1.7% to 8.9% and 1.5% to 7.7% separately, public housing for sale does not affect the property values significantly, and social housing positively affects the property values by 1.7% to 7.7% on the contrary. In addition, these effects decline with distance from public housing. With regard to the impact on high and low priced property in different ranges, social housing has positive effect on low-priced property by 6.2% to 7.1% and no significant effect to 3.2% on high-priced property. The construction of Xing-Long social housing has positive spillover effect by 6.5% to 15.8%, however, the move in of the residents does not. Empirical results show that social housing newly built are able to enhance the property values nearby, which is contributed to the follow-up construction of the public housing.

The CEO and CFO are the two key executives of a firm. They work cohesively to ensure the growth of the firm. After the adoption of the Sarbanes Oxley Act (SOX) in 2002, the importance of CFOs has increased due to their personal legal obligation in certifying the accuracy of financial statements. Only a few papers such as Mian (2001), Fee and Hadlock (2004), and Geiger and North (2006) focus on CFOs in the pre-SOX era. However, a vacuum exists in research focusing exclusively on CFOs in the post-SOX era. The purpose of this dissertation is to delve into a comprehensive investigation of the CFOs. More specifically, I answer three questions: a) does the CEO change lead to the CFO change? b) does the CFO appointment type affect the firm’s debt-equity choice? and c) does the CFO appointment affect the firm’s information environment? I use Shumway’s (2001) dynamic hazard model in answering question ‘a’. For question ‘b’, I use instrumental variable (IV) regression under various estimation techniques to control for endogeneity. For part ‘c’, I use the cross sectional difference-in-difference (DND) methodology by pairing treatment firms with control firms chosen by the propensity scores matching (PSM). I find there is about a 70% probability of CFO replacement after the CEO replacement. Both of their replacements are affected by prior year’s poor performance. In addition, as a custodian of the firm’s financial reporting, the CFO is replaced proactively due to a probability of restatement of earnings. I find firms with internal CFO hires issue more equity in the year of appointment than firms with external hires. The promoted CFO significantly improves the firm’s overall governance which helps the firm obtain external financing from equity issue. However, I find that CFO turnover does not significantly affect the firm’s information environment. To ensure that my finding is not due to mixing up of samples of good and distressed firms together, I separated distressed firms and re-ran my models and my finding still holds. This dissertation fills the gap in the literature with regards to CFOs and their post SOX relationship with the firm.

This article investigates the decision of consumers at bottle refund machines to either reclaim their bottle deposit or to donate the refund to a non-profit organization. The study documents the unique pre-intervention data on donating behaviour and introduces a field experiment to increase donation levels. The design comprised the strategic framing of the situation by highlighting different cues about the normative, descriptive and local expectations of charitable giving as well as cues about the warm glow of donating money. The experiment took place in 20 supermarkets in Germany and lasted for 12 months. By varying the experimental design and using different modelling approaches, the study arrives at the conclusion that individuals largely act consistent with the assumption having self-regarding preferences that are stable and difficult to change. Hence, our pre-test and postintervention data stand in sharp contrast to results from lab experiments.

Dekonstruktion bezeichnet ein durch den poststrukturalistischen Philosophen Jacques Derrida entwickeltes Lektüreverfahren, das sich gegen den tradierten hermeneutischen Ansatz richtet. Es stellt eine letztgültige, im Text enthaltene Bedeutung in Abrede, legt Brüche und Verwerfungen frei und postuliert die Unabschließbarkeit des Interpretationsvorgangs. Statt selbst einen Sinn zu behaupten, erfüllt sich der Zweck der Dekonstruktion in der Prozessualität, das heißt im Akt des Fragens selbst.