Le résumé en français n'a pas été communiqué par l'auteur. / Le résumé en anglais n'a pas été communiqué par l'auteur.

This dissertation has as main interest how households make important decisions over time. In the first part of the thesis I am interested in some theoretical aspects of this broad research question, first: what are some behavioral aspects of individuals' choice over time? Second: how can preferences be aggregated within a household? The first question is discussed in Chapter1 where I study a behavioral model of choice over time, in which a consumer derives utility not only from consumption, which I refer to as `event utility', but already from anticipating future consumption events. Furthermore, a consumer also obtains a disutility from recall of previous consumption events, due to the feature that the latter then act as a form of an anchor point with which new consumption events are compared. This generates a form of  reference-dependency inside the preference structure of a consumer. In Chapter 2, I address a new property, the non-intersecting Pareto curves, which allows an ordering over Pareto curves defined in a certain environment, e.g. the marriage market or the couple's utility possibility frontiers conditional on the choice of a public good. It is shown that the important property of transferable utility, which allows simple aggregation of preferences among individuals is not equivalent to the non-intersecting property of Pareto frontiers, whereas the latter also allows for (relatively) simple aggregation of preferences. In the last two chapters of this dissertation, I consider a couple of very important policy-related research questions. In Chapter 3, I study the incentives of individuals to invest in their own human capital through on-the-job training programs. I find empirical evidence which suggests that being married can impede on training incidence. This, not necessarily intuitive result, can be explained through a theoretical framework where, within a couple household members share resources through some (cooperative) bargaining process. In this case, possible negative externalities on the spouse of a worker who participates in training can negatively affect the benefits of the latter, reducing the willingness to pay for such training programs. Moreover, the relative say inside the household also affects the way the costs of such training programs are shared and I show that the returns to participating in training are decreasing in such variables which imply higher share of total costs to be paid by the worker who is considering to follow training. This empirical framework then generates an important policy insight: namely that if policy makers want to stimulate workers to invest in their own training, they can do this by improving the relative bargaining power of the targeted group of workers within their households, possibly through less intrusive policies (w.r.t. the labor market) than more macro-orientated policies.The final chapter of this dissertation then brings several insights together and addresses a central research question in the field of intertemporal family economics and household behavior: how good are couples able to insure themselves against risk and shocks over the lifecycle? The literature has suggested that one important means by which couples have more scope for self-insurance compared to single individuals, is the presence of a secondary earner, who can increase his/her labor supply and thereby stabilize household income (the added worker effect.) The extent to which this channel is operative depends on many things, e.g. the particular financial position of the household (how much debts does the household have, what is its borrowing capacity etc.) and how income resources within the household are allocated among the household decision makers (intra-household bargaining). In Chapter 4, I therefore combine a 2-earner lifecycle model (including labor supply decisions of household members) with two features: first, households are faced with endogenous borrowing constraints, where the endogeneity stems from the fact that the amount of borrowing depends on the presence of a collateral assets, which I assume to be housing in my model. Furthermore, and consistent with the state-of-the-art in models in family economics, I allow for bargaining over resources among household decision makers, where the latter can't commit to fixed bargaining weights. The latter then implies that bargaining weights might change over the lifecycle due to shocks affecting the household (e.g. job loss of the primary earner, productivity shocks etc.) I then calibrate the structural model and do some interesting counterfactual exercises pertaining to a tightening or loosening of the credit market, affecting the ability of households to obtain collateral assets (and, therefore, affect their borrowing constraints.) / Doctorat en Sciences économiques et de gestion / info:eu-repo/semantics/nonPublished

This dissertation investigates the role of reference-dependent preferences in different areas of application, both from an empirical/experimental and a theoretical perspective. Despite their common focus, all chapters are self-contained and can be read independently. In the first chapter, entitled "Does Loss Aversion Beat Procrastination? A Behavioral Health Intervention at the Gym", I analyze the implications of reference-dependent preferences in the domains of self-control and optimal incentive design. Financial incentives are a common tool to encourage overcoming self-control problems and developing beneficial habits. There are different means by which such incentives can be provided, yet, up to date there is little empirical evidence on the relative effectiveness of different incentive designs. I present the results of a field experiment that explores whether and how incentives that are economically equivalent but framed differently affect the likelihood of exercising at a gym. I find that framing incentives in terms of losses, meaning individuals lose cash incentives by not exercising, encourages more frequent visits to the gym than framing incentives in terms of financial gains. After removing these incentives, I observe habit formation in gym exercise only if incentives were framed as losses rather than gains. The findings are consistent with the concept of reference-dependent preferences and loss aversion and suggest that cost reductions and performance improvements can be achieved if opting to frame incentives in terms of losses. The second chapter, entitled "Salience-adjusted Expectation-based Reference Points: Theory and Experiment", studies the consequences of reference-dependent preferences in the domain of decision making under uncertainty. Recent theories of expectation-based reference-dependent preferences offer a structured approach of the formation of reference points, yet do not incorporate important context-specific characteristics. One implicit assumption is that individuals rationally form their reference point as expectations, by correctly predicting the probabilistic environment they are facing. A second assumption is that in subsequent unanticipated decisionmaking problems, individuals consider previously formed lagged expectations as their reference point. In an experimental setup, I demonstrate that specific contextual factors affect the composition of expectation-based reference points. First, while expectations are formed, outcomes that attract the moment of first focus receive a higher weight. Second, in subsequent unanticipated decision making under uncertainty, the outcomes of the choice set affect to which extent lagged expectations are considered as a reference point, depending on the associated intensity of gains and losses. Finally, apart from providing empirical evidence on the limitations of current theories of expectation-based reference-dependence, I present a theoretical extension that can overcome some of these limitations by allowing reference points to be contingent on salient contextual effects. In the third chapter, entitled "Competitive Persuasive Advertising under Consumer Loss Aversion", I examine the role of reference-dependent preferences in the domain of consumer choice. In particular, I analyze the effects of expectation-based loss aversion in imperfect competition when consumers’ gain-loss utility is susceptible to salience effects. I present a theoretical model in which consumers’ gain-loss utility associated with the expectation to buy the most salient products within their contextual environment is inflated upwards, whereas the gain-loss utility associated with the expectation to buy the least salient products is deflated downwards. Firms can strategically manage consumers’ gain-loss utility by investing in salience-enhancing activities, such as persuasive advertising. If consumers are initially aware of prices but uncertain about their individual match value from the purchase, persuasive advertising has strictly anticompetitive consequences. This is because it allows firms to mitigate consumers’ experienced losses from higher prices, which reduces competitivepressure. / Doctorat en Sciences économiques et de gestion / info:eu-repo/semantics/nonPublished

Titre de l'écran-titre (visionné le 15 mars 2024) / Cette thèse porte sur l'analyse microéconométrique des comportements sociaux dans le monde contemporain, plus spécifiquement, elle explore la conformité aux normes et aux comportements socialement désirés. Elle est articulée autour de trois chapitres doublement interreliés du point de vue de l'objet (les comportements individuels qui affectent le bien-être collectif) et de l'approche d'analyse (analyse des liens de causalité selon une approche microéconométrique). Les comportements individuels antisociaux peuvent être source d'externalités négatives. Ces externalités peuvent être réciproques ou non et donc affecter le bien-être individuel et collectif. Une meilleure compréhension des causes et des effets de ces comportements permet d'élaborer des politiques appropriées visant leur internalisation. Le premier chapitre a investigué sur la relation de causalité entre la perte de capital social et certains comportements antisociaux à savoir la fraude dans les programmes sociaux du gouvernement, la fraude dans les transports publics, la fraude fiscale et les pots de vin dans les services publics. D'une part, ces comportements antisociaux ont été utilisés comme un proxy de la violation des normes sociales. D'autre part, la confiance sociale a été utilisée comme proxy du capital social parce qu'elle en est la composante principale ; la confiance sociale étant mesurée par la confiance dans le gouvernement. Nous avons soutenu que l'approche d'analyse en données de comptage était la stratégie d'analyse la plus optimale comparativement à la stratégie traditionnelle d'analyse de variable discrète qui utilise les modèles probit ou logit. Nous avons en outre montré que la distribution binomiale négative s'ajuste mieux aux données de *Word Value Survey* pour le Canada. Ainsi, l'approche de variable instrumentale binomiale négative a été utilisée pour l'analyse des liens de causalité entre la confiance sociale et la conformité aux normes sociales. Nos résultats ne démontrent pas l'évidence d'un lien de causalité entre la confiance sociale et les comportements d'intérêt, toutefois, ils mettent en exergue une association entre la volonté de violer les normes sociales et le profil socio-économique individuel des personnes qui s'adonnent à ses pratiques déviantes. les résultats suggèrent par ailleurs un biais très important de l'effet de la confiance sociale sur les variables d'intérêt. Le deuxième chapitre analyse les comportements d'un groupe particulier de personnes (les personnes vivant avec des conditions médicales sous-jacentes) dans le contexte d'une pandémie. De façon spécifique, ce chapitre a exploré les comportements des personnes vivant avec des conditions médicales sous-jacentes en réponse aux directives du gouvernement dans le cadre du *COVID-19*. Nous avons disposé de deux vagues de sondage sur les comportements des individus aux plus forts moments de la pandémie de *COVID-19* au Québec. Nous avons mis en exergue les problèmes techniques inhérents aux données de panel court notamment l'attrition des données. Ces problèmes ne permettent pas de mener efficacement une analyse en données de panel. Nous avons donc utilisé la technique de reéchantillonnage suggérée par Hirano et al. (2001). Les résultats nous permettent de soutenir que les personnes souffrant de problèmes de santé sous-jacents se conforment mieux aux directives de santé publique, tout en percevant qu'elles risquent davantage de contracter le virus à l'avenir. Les résultats ne permettent pas de soutenir l'évidence que les comportements des individus avec des conditions médicales sont gouvernés par un effet contextuel ou par une défection de leur état psychologique. Nous soutenons que les croyances de ces personnes contrastent avec les croyances hypocondriaques. Le troisième chapitre a investigué sur une nouvelle façon de mesurer la littératie financière. L'analyse économique des effets de la littératie financière se heurte à deux préoccupations d'ordre méthodologiques : la mesure de la littératie financière et les biais découlant des erreurs de mesure. Nous avons proposé dans ce chapitre, une procédure articulée en trois étapes comprenant un technique de sélection des variables. la sélection des variables est fondée sur des propriététs mathématiques bien définies et permet d'exprimer de façon unique la littératie financière avec un indice. A partir des données du sondage 2019 de l'agence de consommation en matière financière du Canada, nous avons montré que notre indice permet de mieux prédire l'effet de la littératie financière sur la planification de la retraite que le variant de l'indice de double pondération existant. Notre indice est par ailleurs plus efficace. En effet, contrairement aux indices alternatifs, notre indice donne toujours des résultats convergents avec la technique de double pondération impliquant une combinaison du *principal component analysis (pca)* et du *relative to identified distribution (ridit)*. En fin, l'analyse en variable instrumentale met en évidence un effet causal entre la littératie financière et la planification de la retraite ou le revenu dans certains sous-groupes socio-économiques. / This thesis focuses on the analysis of compliance to social norms and socially desired behaviors using microeconometric methods. It is divided into three interconnected chapters that examine individual behaviors affecting collective well-being, employing a micro-econometric approach to analyze causal links. Individual antisocial behaviors can give rise to negative externalities, which may or may not be reciprocal, thereby impacting individual and collective well-being. By gaining a better understanding of the causes and effects of these behaviors, it becomes possible to develop appropriate policies to internalize them. The first chapter investigates the causal link between the loss of social trust and specific antisocial behaviors, such as fraud in government social programs, fraud in public transport, tax evasion, and bribery in public services. We argue that the count data analysis approach is more effective compared to the traditional discrete variable analysis strategy utilizing probit or logit models. Additionally, our analysis reveals that the negative binomial distribution better fits the data from the *Word Value Survey* for Canada. Consequently, we employ the negative binomial instrumental variable approach to examine causal links. Our findings do not provide conclusive evidence of a causal relationship between social trust and the targeted behaviors. However, they do highlight an association between the propensity to violate social norms and an individual's socio-economic profile. Overall, these findings contribute to the understanding of social behavior and provide insights for the development of targeted policies aimed at addressing and mitigating antisocial behaviors. The second chapter focuses on analyzing the behaviors of a specific group of individuals, namely those living with medical conditions, within the context of a pandemic. Specifically, this chapter examines the responses of individuals with medical conditions to government directives during the COVID-19 pandemic. We conducted two waves of surveys to gather data on individuals' behaviors at the height of the pandemic in Quebec. Throughout our analysis, we encountered technical challenges inherent in short panel data, such as attrition, which hindered effective panel data analysis. To overcome these challenges, we utilized the refreshment samples technique proposed by Hirano et al. (2001). Our findings indicate that individuals with underlying health conditions demonstrate greater compliance with public health guidelines, likely driven by their perception of being at a higher risk of contracting the virus in the future. The results do not support the notion that the behaviors of individuals with medical conditions are in uenced by contextual factors or a deviation from their psychological state. We argue that the beliefs held by these individuals contrast with hypochondriac beliefs. As for the third chapter, it investigates a new way of measuring financial literacy. The economic analysis of the effects of financial literacy encounters two methodological concerns: estimating financial literacy and addressing biases resulting from measurement errors. In this chapter, we propose a procedure that enables to express financial literacy through a composite index, providing a comprehensive measure. This is a procedure for uniquely measuring financial literacy from a database. Using data from the Financial Consumer Agency of Canada's 2019 survey, we demonstrate that our financial literacy index better predicts the effect of financial literacy on retirement planning compared to a variant of the existing double weight index using a combination of *relative to identified distribution (ridit)* and *principal component analysis (pca)*. Additionally, through instrumental variable analysis, we identify a causal effect between financial literacy and retirement planning or income in certain socio-economic subgroups. These findings shed light on the importance of financial literacy and its implications for individuals in the Canadian context. It is crucial to consider these results in the formulation of policies and interventions aimed at improving financial literacy and its associated outcomes.

Cette thèse en microéconomie appliquée se compose de trois chapitres, chacun abordant une question différente. Le premier chapitre, « La distribution séquentielle en présence de fraude », montre comment les entreprises peuvent exploiter le moment de la diffusion des contenus afin d’atténuer les effets de la fraude numérique, dans un monde où un certain piratage est inévitable. À travers un modèle, cet article fournit une explication claire de la façon dont une firme peut faire obstacle à la fraude à travers un changement de la distribution séquentielle du produit. En l’absence de fraude, les profits des entreprises sont indépendants de la manière dont le contenu est mis en vente. Cependant, lorsque la fraude représente une menace réelle, l’entreprise peut en atténuer les effets en choisissant de manière stratégique la part du produit qui est offert à chaque période. Cela permet de changer la valeur du produit pour les consommateurs et de rendre la fraude moins attractive de ce point de vue. L’entreprise en situation de monopole bénéficie de la libération de contenu en deux périodes différentes d’une manière asymétrique qui trouvent des analogies dans des exemples réels tels que le marché des outils logiciels spécialisés ou des émissions de télévision. Le deuxième chapitre, « L’offre complice dans les enchères à valeur commune », présent les effets d’un type d’enchère frauduleuse appelée « enchère complice ». Cette fraude consiste à placer des enchères anonymes sur ordre du vendeur en vue de faire monter artificiellement le prix de l’objet vendu. Nous concevons un modèle simple pour comprendre les incitations du vendeur à enchérir de manière complice dans une enchère anglaise à valeur commune, dans laquelle l’information privée des participants suit une distribution discrète. Nous montrons comment le caractère discret de la distribution affecte le profit espéré du fait d’enchérir de manière complice pour le vendeur ex-ante. Nous montrons aussi comment le vendeur révise ses enchères complices à partir de l’information qu’il reçoit à mesure que l’enchère se déroule. Nous trouvons que si le nombre de signaux est faible, le vendeur peut gagner à ne pas participer, même lorsque les autres participants sont myopes. Par ailleurs, et quel que soit le nombre de signaux dans l’enchère, si le nombre de participants est assez élevé, le fait d’avoir la possibilité de mettre en œuvre une enchère complice a toujours pour effet de détériorer le profit espéré du vendeur. Le troisième chapitre, « Le confort des médecins et le choix des césariennes », est co-écrit avec Shagun Khare et Alan Acosta. Cet article analyse les causes qui pourraient expliquer le taux élevé de césarienne à Buenos Aires, en Argentine, qui excède largement la recommandation de l’Organisation mondiale de la Santé (OMS). L’hypothèse d’une demande induite par le fournisseur, qui prévoit plus de césariennes que nécessaire du point de vue médical, pourrait expliquer cet écart. Dans cet article co-écrit avec Shagun Khare et Alan Acosta, nous étudions un aspect des incitations des médecins à stimuler la demande : le confort. A partir d’une enquête menée auprès de femmes enceintes à Buenos Aires, nous regardons si les chances qu’une femme fasse l’objet d’une césarienne dépend du moment de l’accouchement, en particulier s’il s’agit d’un jour travaillé ou non. En laissant de côté les césariennes planifiées, nous trouvons que le confort compte, mais seulement dans les hôpitaux privés. Nous trouvons aussi qu’une femme qui déclare préférer une césarienne à une naissance naturelle a de plus grandes chances de recourir à une césarienne dans les hôpitaux privés. Nos résultats montrent aussi que l’environnement institutionnel joue un rôle déterminant sur la manière dont le confort des médecins et les préférences des mères influencent ces choix. / This thesis in applied microeconomics is composed of three chapters, each one addressing a different question. The first chapter, “Sequential distribution in the presence of Piracy”, shows how firms can exploit the timing of the release of digital content as a way to mitigate the effects of piracy, in a world where some piracy is unavoidable. We develop an analytical model where a monopolist produces a particular good, and it can choose the time at which its product is available to consumers. On top of deciding on prices, the monopolist also chooses the share of the product it releases at each period. In the absence of piracy, firm’s profits are independent of the way in which content is released. However, when piracy is a real threat, the firm can soften its effect by strategically selecting the share of the product offered in each period, changing consumers’ valuation and making piracy less attractive from their perspective. The monopolist benefits from releasing content in two different periods in an asymmetric way which find analogies in real life examples such as the market of specialised software tools or of TV shows. The second chapter, “Shill bidding in common value auctions”, presents the effects of a particular cheating environment in common value auctions. Shill bidding consists of placing anonymous bids on the seller’s behalf to artificially drive up the prices of the auctioned item. We build a simple model to understand the incentives a seller has to shill bid in an English common value auction where the bidders’ private information is drawn from a discrete distribution. We show how the discreteness affects the seller’s ex-ante expected gain of shill bidding, and we also show how the seller updates his shill bid based on the new information he receives as the auction goes on. We find that if the number of signals is low, the seller might be better off refraining from participating even when bidders are fully myopic. Moreover, for any number of signals in the auction, if the number of participants is sufficiently high, the shill bidding strategy always deteriorates the seller’s expected profits. The third chapter, “Physician convenience and cesarean section delivery”, is co-authored with Shagun Khare and Alan Acosta. This paper analyses the causes that might explain the high rate of cesarean section in Buenos Aires, Argentina, that far exceeds the World Health Organization recommendation. The supplier-induced demand hypothesis, which predicts more c-section deliveries than otherwise medically needed, might be the reason for this disparity. In this paper, using a survey of pregnant women in Buenos Aires, we study one aspect of the physician’s incentives to induce demand: convenience. We look at whether a woman’s chance of getting a c-section depends on the period of delivery, i.e. whether it is a working day or not. Setting aside scheduled c-sections, we find that convenience matters, but only in private hospitals. We also find that women who state that they prefer c-sections over natural births have a higher chance of having a c-section in private hospitals. While physicians’ convenience and mothers’ preferences do matter, our research finds that the institutional environment plays a defining role in how much these matters.

En microéconomie, le sujet d'étude est habituellement l'agent économique et tout ce qui s'y rattache. Quand il maximise son utilité, l'agent prend en considération un grand nombre de facteurs: certains économiques, d'autres purement psychologiques. Le but de ce travail est de voir l'influence du remords sur la prise de décision de ce même agent. Parce que la littérature économique sur le sujet est de plus en plus abondante et aussi, parce qu'on ne peut se pencher sur des aspects autant psychologiques que philosophiques, une imposante revue de la littérature sur trois principaux sujets (remords, normes sociales, révision des croyances) a été faite. Pour formaliser l'histoire, nous avons eu recours à la théorie des jeux et nous avions comme hypothèse de départ que les remords devraient avoir une influence plus ou moins importante quant à la prise de décision. Et, en effet, suite à une série de raffinements, les équilibres ayant survécus incorporaient tous le remords dans leurs conditions d'existence. Dans la discussion, des explications sont données sur ce dernier point et nous avons fait une extrapolation très limité des possibilités de répétition du jeu, une ou plusieurs fois, dépendamment du résultat obtenu la première fois. ______________________________________________________________________________ MOTS-CLÉS DE L’AUTEUR : Remords, Révision des croyances, Normes sociales, Théorie des jeux, Microéconomie.

This thesis consists of three independent essays on the design of matching markets, with a primary goal to understand how information interacts with matching mechanisms especially in the applications to school choice and college admissions. The first chapter compares theoretically the non-strategyproof Boston mechanism and the strategy-proof deferred acceptance mechanism when taking into account that students may face uncertainty about their own priorities when submitting preferences, one important variation from the complete information assumption. The second chapter evaluates the effectiveness of a strategy-proof mechanism when students have to submit preferences before knowing their priorities using both theory and data. The third chapter turns attention to a new mechanism that is sequentially implemented and can encourage truth-telling. Nevertheless, such implementation often faces time constraint. This chapter therefore offers an inquiry of the pros and cons of the time-constrained sequential mechanism. / Doctorat en Sciences économiques et de gestion / info:eu-repo/semantics/nonPublished

This thesis consists of three independent essays which contribute to the literatures on organizational and regulatory economics.In the first part of the dissertation, I address questions related to the optimal incentive provision in hierarchies. In particular, I investigate how the choice of the optimal compensation policy of an organization is affected by the workers' psychological preferences for reciprocity. This essay relates to a recent strand of theoretical and empirical research that studies how the presence of reciprocity concerns impacts on the optimal organizational design (See e.g. Dohmen et al. 2009, Englmaier and Leider, 2012, Englmaier et al. 2015).The second part of the dissertation concerns the optimal design and regulation of a hierarchical organization in the presence of capture concerns. In many organizations the task of evaluating an agent's performance is delegated to a third party, a supervisor, who can opportunistically misreport information. The question of how the provision of incentives in hierarchies is affected by the supervisor's opportunism has long been studied in economics. Addressing this research question is of great importance since it can improve our understanding of the internal organization of firms and can have broad applications to regulatory design. / Doctorat en Sciences économiques et de gestion / info:eu-repo/semantics/nonPublished

My dissertation focuses on economics of higher education. Specifically, I study how scientists’ social network gives indications on their later career (Chapter 1), the universities’ research performance (Chapter2), and the overall production of research outputs (Chapter3). Building on the current surge of social network analysis, all the papers are built upon networks of co-authors. This thesis contributes to the study of social networks. It documents the prevalence of research collaborations and how they impact the production of science, making a case for taking this phenomenon into account when designing funding mechanisms. Chapter 1 looks at how a researcher’s professional network influences her career path, and I specifically consider the career of young economists on the American academic market. I exploit an original dataset building from the researchers’ individual vitae and their publication records. I investigate the impact of social network on career path by looking at the correlation between early career network metrics and the quality of the institutional affiliation of the researcher. I find that the number of social ties a researcher has as well as her relative position in the research network matters for explaining career mobility and success, even when controlling for publications. Having more co-authors boost the early career, while a higher quality of publications matters on the long run. In Chapter 2, I look at the impact of inter-university partnerships on the production of research outputs.Using an original data set of scientific publications and universities’ budgets, I analyze the network of research in Spain based on the network of Spanish co- authors. I show how the growth in research productivity of Spanish institutions before the crisis was linked to the increase in universities’ budgets and in inter- university collaborations. The results show that the size of the university is the key factor to understand universities productivity. The network multiplier is significant and positive, indicating that collaboration has a positive effect. Finally, in the context of the current crisis, I am able to identify the universities that are the least productive, taking into account their own characteristics and the indirect effects of the collaborations. This analysis has clear policy implications,as the least productive universities could be targeted to minimize the impact of further budgets cuts.Finally, Chapter 3 focuses on the link between the composition of the scientists’ workforce and the amount of research produced. Using Chapter 2’s dataset enriched by a list of the applicants to the two most prestigious postdoctoral grants in Spain, I am able to identify the young researchers in the co-authorship network. I study the link between the number of young researchers and the total research output. All three chapters show how important collaborations are in the production of science. The first chapter shows how some network metrics correlate with career outcomes, giving indication on how much to engage in collaborative work. The second paper shows how network analysis can be used to produce performance rankings of universities taking into account the partnerships. Finally, the third chapter makes a case for the importance of policies targeting young scientists. Further research can be done to understand the link between competition for students and resources and the co-authorship network, or the endogenous process of career changes and changes in the network. / Doctorat en Sciences économiques et de gestion / info:eu-repo/semantics/nonPublished

Cette thèse présente trois chapitres qui utilisent et développent des méthodes microéconométriques pour l’analyse de microdonnées en économique. Le premier chapitre étudie comment les interactions sociales entre entrepreneurs affectent la prise de décisions en face de risque. Pour ce faire, nous menons deux expériences permettant de mesurer le niveau d’aversion au risque avec de jeunes entrepreneurs ougandais. Entre les deux expériences, les entrepreneurs participent à une activité sociale dans laquelle ils peuvent partager leur connaissance et discuter entre eux. Nous recueillons des données sur la formation du réseau de pairs résultant de cette activité et sur les choix des participants avant et après l’activité. Nous trouvons que les participants ont tendance à faire des choix plus (moins) risqués dans la seconde expérience si les pairs avec qui ils ont discuté font en moyenne des choix plus (moins) risqués dans la première expérience. Ceci suggère que même les interactions sociales à court terme peuvent affecter la prise de décisions en face de risque. Nous constatons également que les participants qui font des choix (in)cohérents dans les expériences ont tendance à développer des relations avec des individus qui font des choix (in)cohérents, même en conditionnant sur des variables observables comme l’éducation et le genre, suggérant que les réseaux de pairs sont formés en fonction de caractéristiques difficilement observables liées à la capacité cognitive. Le deuxième chapitre étudie si les politiques de comptes d’épargne à avantages fiscaux au Canada conviennent à tous les individus étant donné l’évolution de leur revenu et les différences dans la fiscalité entre les provinces. Les deux principales formes de comptes d’épargne à avantages fiscaux, les TEE et les EET, imposent l’épargne à l’année de cotisation et de retrait respectivement. Ainsi, les rendements relatifs des deux véhicules d’épargne dépendent des taux d’imposition marginaux effectifs au cours de ces deux années, qui dépendent à leur tour de la dynamique des revenus. J’estime un modèle de dynamique des revenus à l’aide d’une base de données administrative longitudinale canadienne contenant des millions d’individus, ce qui permet une hétérogénéité substantielle dans l’évolution des revenus entre différents groupes. Le modèle est ensuite utilisé, conjointement avec un calculateur d’impôt et de transferts gouvernementaux, pour prédire comment les rendements des EET et des TEE varient entre ces groupes. Les résultats suggèrent que les comptes de type TEE génèrent en général des rendements plus élevés, en particulier pour les groupes à faible revenu. La comparaison des choix d’épargne optimaux prédits par le modèle avec les choix d’épargne observés dans les données suggère que les EET sont en général trop favorisées dans la population, surtout au Québec. Ces résultats ont d’importantes implications sur les politiques de « nudge »qui sont actuellement mises en oeuvre au Québec, obligeant les employeurs à inscrire automatiquement leurs employés dans des comptes d’épargne de type EET. Ceux-ci pourraient produire des rendements très faibles pour les personnes à faible revenu, qui sont connues pour être les plus sensibles au « nudge ». Enfin, le troisième chapitre étudie les problèmes méthodologiques qui surviennent fréquemment dans les modèles de régression par discontinuité (RD). Il considère plus précisément le problème des erreurs d’arrondissement dans la variable déterminant le traitement, ce qui rend souvent la variable de traitement inobservable pour certaines observations autour du seuil. Alors que les chercheurs rejettent généralement ces observations, je montre qu’ils contiennent des informations importantes, car la distribution des résultats se divise en deux en fonction de l’effet du traitement. L’intégration de cette information dans des critères standard de sélection de modèles améliore la performance et permet d’éviter les biais de spécification. Cette méthode est prometteuse, en particulier pour améliorer les estimations des effets causaux dans les très grandes bases de données, où le nombre d’observations rejetées peut être très important, comme le LAD utilisé au chapitre 2. / This thesis presents three chapters that use and develop microeconometric methods for microdata analysis in economics. The first chapter studies how social interactions influence entrepreneurs’ risk-taking decisions. We conduct two risk-taking experiments with young Ugandan entrepreneurs. Between the two experiments, the entrepreneurs participate in a networking activity where they build relationships and discuss with each other. We collect data on peer network formation and on participants’ choices before and after the networking activity. We find that participants tend to make more (less) risky choices in the second experiment if the peers they discuss with make on average more (less) risky choices in the first experiment. This suggests that even short term social interactions may affect risk-taking decisions. We also find that participants who make (in)consistent choices in the experiments tend to develop relationships with individuals who also make (in)consistent choices, even when controlling for observable variables such as education and gender, suggesting that peer networks are formed according to unobservable characteristics linked to cognitive ability. The second chapter studies whether tax-preferred saving accounts policies in Canada are suited to all individuals given they different income path and given differences in tax codes across provinces. The two main forms of tax-preferred saving accounts – TEE and EET – tax savings at the contribution and withdrawal years respectively. Thus the relative returns of the two saving vehicles depend on the effective marginal tax rates in these two years, which in turn depend on earning dynamics. This chapter estimates a model of earning dynamics on a Canadian longitudinal administrative database containing millions of individuals, allowing for substantial heterogeneity in the evolution of income across income groups. The model is then used, together with a tax and credit calculator, to predict how the returns of EET and TEE vary across these groups. The results suggest that TEE accounts yield in general higher returns, especially for low-income groups. Comparing optimal saving choices predicted by the model with observed saving choices in the data suggests that EET are over-chosen, especially in the province of Quebec. These results have important implications for “nudging” policies that are currently being implemented in Quebec, forcing employers to automatically enrol their employees in savings accounts similar to EET. These could yield very low returns for low-income individuals, which are known to be the most sensitive to nudging. Finally, the third chapter is concerned with methodological problems often arising in regression discontinuity designs (RDD). It considers the problem of rounding errors in the running variable of RDD, which often make the treatment variable unobservable for some observations around the threshold. While researchers usually discard these observations, I show that they contain valuable information because the outcome’s distribution splits in two as a function of the treatment effect. Integrating this information in standard data driven criteria helps in choosing the best model specification and avoid specification biases. This method is promising, especially for improving estimates of causal effects in very large database (where the number of observations discarded can be very large), such as the LAD used in Chapter 2.