Causal Effects Model 的估計

==Causal Effects 估計範例==

in Angrist and Pischke (2017, JEP, p130-132),主要提出觀念的是 Dale and Kruger (2002, QJE)


Casusal Effect: 在申請入學時,同時接到私大和州大入學許可,但最後後選擇唸私大、或州大的學生為樣本:

Yi: 全部的樣本的畢業生所得,可觀察的,其中又分為兩類

Y1i : 第 i 個樣本的「受私校教育後」所得

Y0i : 第 i 個樣本的「受州校教育後」所得


Pi: =1 if 第 i 個樣本唸私立,=0 otherwise 唸州立大學


Y10 : 第 i 個樣本的原來能力

因為好學生集中去唸名私校,所以畢業後收入高,不見得是私校的努力。這個稱為 selection bias 樣本選擇偏誤。

私校的教學效果 (用大概畢業20年後的 earning 來衡量) 之差異為:

Y1i – Y0i

若教學有效的話,然後差異的平均是 β

H0 : E(Y1i – Y0i) = β>0

假設E(Y0i) = α, 即

Y0i =α + ηi

α 為學生原來的潛力, ηi是誤差,或個別差異,這個個別差異會和選私校有關係,例如家庭背景、爸媽是否畢業於私立..。
Caussal-Effect model

Yi = α+βPi+ηi

Pi 和 ηi 是(統計上)不獨立的,也就是無法滿足迴歸上原來的獨立性要求。

這個 causal-effect model 的想法創新就在此,他們提出比較不嚴格的「條件獨立性假設」(conditional independence assumption),

E(ηi|Pi,Xi) = E(ηi|Xi)

所以要找其它的可能影響畢業後所得能力的變數 X (例如 SAT 的分數…),來加入估計,觀念上是

E(ηi|Pi,Xi) = E(ηi|Xi) = E(ηi|Xi)

所以, causal-effect model 最後就變成


此法可建構出 unbias 和 consistent 的 β 估計,而且它有明確的意義:唸私校和唸公校的「效果差異」平均值。

Angrist, Joshua D., and Jörn-Steffen Pischke. “Undergraduate econometrics instruction: through our classes, darkly." Journal of Economic Perspectives, 31.2 (2017): 125-44.

Dale, Stacy Berg, and Alan B. Krueger. “Estimating the payoff to attending a more selective college: An application of selection on observables and unobservables." The Quarterly Journal of Economics, 117.4 (2002): 1491-1527.


Altruistic and risk preference of individuals and groups

Date: 2016-10
By: Yoshio Kamijo (School of Economics and Management, Kochi University of Technology)
Teruyuki Tamura (School of Economics and Management, Kochi University of Technology)
This study examines whether attitudes toward risk and altruism are affected by being in a group or being alone. Subjects in our experiment were requested only to show their faces to other members without any further communication, differing from previous studies. In experiments of both anonymous investments and donations, we found that subjects who made decisions in a group offered significantly lower amounts than individuals who made decisions alone, even controlling for individuals’ risk and altruistic preferences. Our results indicate that people are more risk averse and self-interested when they are in a group.
Keywords: Group decision, Altruism, Decision under risk
JEL: C91 C92 D81

Centrality measures in networks based on nodes attributes, long-range interactions and group influence

Date: 2016-10
By: F. Aleskerov
N. Meshcheryakova
S. Shvydun
We propose a new method for assessing agents influence in network structures, which takes into consideration nodes attributes, individual and group influences of nodes, and the intensity of interactions. This approach helps us to identify both explicit and hidden central elements which cannot be detected by classical centrality measures or other indices.

Collective Commitment

Date: 2016-07
By: Christian Roessler
Sandro Shelegia
Bruno Strulovici
We consider collective decisions made by agents whose preferences and power depend on past events and decisions. Faced with an inecient equilibrium and an opportunity to commit to a policy, can the agents reach an agreement on such a policy? We provide a consistency condition linking power structures in the dynamic setting and at the commitment stage. When the condition holds, commitment has no value: any agreement that may be reached at the outset coincides with the equilibrium without commitment. When the condition fails, as in the case of time-inconsistent preferences, commitment can improve outcomes. We discuss several applications.
JEL: D70 H41 C70

Individual and Group Preferences Over Risk: An Experiment

Date: 2016-07
By: Morone, Andrea
Temerario, Tiziana
The recent literature on individual and group choices over risk has led to different results. In some studies under unanimity, groups were found to be less risk averse than individuals, while those under majority did not highlight significant differences. However, both the types of studies impose the decision rule to the group. In the present work we elicited groups’ preference under risk using a consensus rule, i.e. groups are free to solve disagreement endogenously, just as in the real life. Results from our pairwise choices experiment shows that when group members are free to use any rule they want in order to reach unanimity, there is no statistical difference between individuals’ and groups’ risk aversion.
Keywords: Risk; uncertainty; decision-making; group decision; lottery; experimental economics; experiment;
JEL: C92 D81

Who are the voluntary leaders? Experimental evidence from a sequential contribution game

Date: 2016
By: Raphaële Préget (LAMETA – Laboratoire Montpelliérain d’Économie Théorique et Appliquée – Montpellier SupAgro – Centre international d’études supérieures en sciences agronomiques – UM3 – Université Paul-Valéry – Montpellier 3 – INRA Montpellier – Institut national de la recherche agronomique [Montpellier] – UM – Université de Montpellier – CNRS – Centre National de la Recherche Scientifique)
Phu Nguyen Van (UMR Beta – CNRS – Centre National de la Recherche Scientifique)
Marc Willinger (LAMETA – Laboratoire Montpelliérain d’Économie Théorique et Appliquée – Montpellier SupAgro – Centre international d’études supérieures en sciences agronomiques – UM3 – Université Paul-Valéry – Montpellier 3 – INRA Montpellier – Institut national de la recherche agronomique [Montpellier] – UM – Université de Montpellier – CNRS – Centre National de la Recherche Scientifique, Université de Montpellier)
We rely on the methodology of Fischbacher et al. (2001) in order to identify subjects’ behavioral types. We then link the likelihood to act as a leader in a repeated public goods game to the elicited behavioral types. The leader in a group is defined as the subject who voluntarily decides in the first place about his contribution. The leader’s contribution is then reported publicly to the remaining group members who take their contribution decisions simultaneously. Our main findings are that leaders emerge in almost all rounds and that subjects who are identified as conditional cooperators are more likely to act as leaders than other types, e.g. free-riders or triangle-contributors. We also find that voluntary leaders, irrespective of their behavioral type, contribute always more than followers. However the presence of leadership does not prevent the decay that is commonly observed in linear public goods experiments.
Keywords: Voluntary Contribution Mechanism,Leadership,Public Goods,Experimental Economics

Communication and voting in heterogeneous committees: An experimental study

Date: 2016-03
By: Mark T. Le Quement (University of Bonn)
Isabel Marcin (Max Planck Institute for Research on Collective Goods)
We study experimentally the effectiveness of communication in common value committees exhibiting publicly known heterogeneous biases. We test models assuming respectively self-interested and strategic-, joint payoff-maximizing- and cognitively heterogeneous agents. These predict varying degrees of strategic communication. We use a 2 x 2 design varying the information protocol (communication vs exogenous public signals) and the group composition (heterogeneous vs homogeneous). Results are only consistent with the third model. Roughly 80% of (heuristic) subjects truth-tell and vote with the majority of announced signals. Remaining (sophisticated) agents lie strategically and approximately apply their optimal decision rule.
Keywords: Committees, Voting, Information Aggregation, Cheap Talk, Experiment
JEL: C92 D72 D82 D83

Tell Me How to Rule: Leadership, Delegation, and Voice in Cooperation

Date: 2016
By: Marco Faillo
Federico Fornasari
Luigi Mittone
Following some recent studies, we experimentally test the effect of intra-group leadership in a public good experiment. Specifically, individuals taking part in our experiment are randomly assigned either the role of leader or the role of follower. Leaders take part in a public good game, aware of the fact that every decision they make directly affects their followers. In this sense, our experimental setting combines the dimension of leadership in cooperation with the one of delegated agents. In our experiment, we find that leadership produces two main effects: subjects contribute more, and tend to punish more frequently. In spite of the presence of higher contributions, we observe lower payoffs; these are caused by an aggressive behavior that push leaders to mane an undue use of punishment. Allowing one-sided communication between followers and leaders provide a different effect: communication reduces decision makers’ aggressiveness, leading to lower contributions and punishment, but better results in terms of final payoffs. The same welfare can be reached when leadership is not implemented at all; this suggests that the presence of a dictatorial leader in public goods with punishment can be beneficial only when there is communication.
Keywords: Voluntary contribution experiment, Leadership, Punishment
JEL: C72 C92 H41 O12

Where Do Social Preferences Come From?

Date: 2015-08
By: Chaning Jang (Department of Psychology, Princeton University)
John Lynham (Department of Economics & UHERO, University of Hawaii at Manoa; Center for Ocean Solutions, Stanford University)
Where do preferences for fairness come from? We use a unique field setting to test for a spillover of sharing norms from the workplace to a laboratory experiment. Fishermen working in teams receive random income shocks (catching fish) that they must regularly divide among themselves. We demonstrate a clear correlation between sharing norms in the field and sharing norms in the lab. Furthermore, the spillover effect is stronger for fishermen who have been exposed to a sharing norm for longer, suggesting that our findings are not driven by selection effects. Our results are consistent with the hypothesis that work environments shape social preferences.
Keywords: ultimatum game; social preferences; fairness; workplace spillovers
JEL: Q2 C9 C7 B4 D1

Can a single theory explain coordination? An experiment on alternative modes of reasoning and the conditions under which they are used

Date: 2016-01-18
By: Marco Faillo (University of Trento)
Alessandra Smerilli (PFSE-Auxilium)
Robert Sugden (University of East Anglia)
We investigate experimentally the conditions under which bounded best response and collective optimality reasoning are used in coordination games. Using level-k and team reasoning theories as exemplars, we study games with three pure-strategy equilibria, two of which are mutually isomorphic. The third is always team-optimal, but whether it is predicted by level-k theory differs across games. We find that collective optimality reasoning is facilitated if the collectively optimal equilibrium gives more equal payoffs than the others, and is inhibited if that equlibrium is Pareto-dominated by the others, considered separately. We suggest that coordination cannot be explained by a single theory.
Keywords: team reasoning, level-k theory, coordination games
JEL: C7 C9