ECONOMETRIC METHODS FOR CAUSAL INFERENCE: THEORY AND APPLICATIONS

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: COMPLETE
• Funding Source: HATCH
• Reporting Frequency: Annual
• Accession No.: 0216160
• Project Start Date: Oct 1, 2008
• Project End Date: Sep 30, 2013
• Program Code: [(N/A)]- (N/A)

Project Director
Flores-Lagunes, A.

Recipient Organization
UNIVERSITY OF FLORIDA
G022 MCCARTY HALL
GAINESVILLE,FL 32611

Performing Department
Food and Resource Economics

Non Technical Summary
The estimation of cause-and-effect relationships are of central importance in applied research and policy making. Accurate estimation of causal effects allows the appropriate evaluation, design, and funding decisions of governmental policies. Examples of policy questions that require estimation of causal effects to answer them abound: is the U.S. "Energy Bill" responsible for the recent spike in oil prices Do the long-run benefits of the 4-H program outweigh its short-term costs in the state of Florida Is the use of tobacco during pregnancy responsible for the costly incidence on low birth weight, and, if so, what can policymakers do to prevent it In natural sciences such as biology or chemistry, the implementation of experimental trials--i.e. the assignment of units into treatment and control groups--constitutes the canonical method to determine causality. In social sciences, however, experimental trials are often difficult to implement and thus causal effects need to be estimated with non-experimental data, employing econometric methods. The basic-science component of this project develops new econometric methods for the estimation causal effects. Specifically, the new methods deal with the estimation of the causal mechanisms through which a "treatment" (or "program") causally works. For example, if the 4-H program is estimated to have a positive impact on childrens grades, how much of that positive effect is causally due to e.g. leadership skills programs as opposed to other types of 4-H programs Answering this type of questions allows a better understanding of the causal relationship under consideration, which is relevant for policy purposes. The applied component of this project relates to the application of econometric methods for causal inference to relevant policy questions. Three applications concerning the state of Florida will be considered. The first of them applies the new econometric methods described above. It employs Florida Natality Records to analyze the role of gestation time (a potential causal mechanism) in the transmission of the negative effects of smoking during pregnancy on birth weight. Low birth weight is a costly condition widely associated with a myriad of health, behavioral and socioeconomic problems in later stages of life. Therefore the interest in understanding its determinants with the purpose of conducting policies aimed at minimizing incidence. The second application is an economic evaluation of the 4-H program in Florida, a youth development program aimed at providing youth with life and leadership skills to help them reach their full potential. Given the amount of resources devoted to Floridas 4-H (over $21 million in 2007; IFAS Extension 2007 Annual Report), it is important to conduct economic evaluations that yield cost-benefit analyses of the program. The last application evaluates the federal job training program Job Corps, a residential program aimed at disadvantaged youth in which the federal government spends about one billion annually, and which operates 3 centers throughout the state of Florida.

Animal Health Component

(N/A)

Research Effort Categories

Basic

(N/A)

Applied

(N/A)

Developmental

(N/A)

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
610	6010	3010	50%
609	6010	3010	50%

Knowledge Area
609 - Economic Theory and Methods; 610 - Domestic Policy Analysis;

Subject Of Investigation
6010 - Individuals;

Field Of Science
3010 - Economics;

Keywords

causal inference

causal mechanisms

net causal effects

consequences of smoking

low birth weight

4-h program

economic program evaluation

job corps effectiveness in the southeast

Goals / Objectives
The overall objective of this project is the theoretical development and empirical application of econometric causal inference methods. The specific objectives of this project are to: (1) Develop new methods for causal inference, with special emphasis on the estimation of causal mechanisms of a "treatment" (or government intervention) under the mildest statistical assumptions possible and in settings relevant to economics. (2) Apply the theoretical developments in causal inference within this project (objective 1) to analyze the role of gestation time in the transmission of the negative effects of smoking during pregnancy on birth weight in Florida and other U.S. states. (3) Apply econometric causal inference methods to the estimation of the economic impacts of the 4-H program in Florida. (4) Apply recently developed methods for causal inference to the Job Corps training program, with emphasis on the estimation of program impacts in Florida and the southeastern United States.

Project Methods
The procedures to be employed consist of both theoretical techniques for statistical identification of causal effects and empirical strategies for the estimation of causal effects. The methods to be employed in each of the objectives of the project are as follows. (1) This objective is about the theoretical contribution of the project. An initial difficulty in the identification of mechanism effects is that their definition requires controlling for an intermediate variable (the mechanism) that is affected by the treatment. The procedures to be employed consist of the use of "principal stratification"--a concept first introduced by Frangakis and Rubin (2002) for the definition of causal effects controlling for intermediate variables--to clearly define the parameters of interest. This will be followed by the use of statistical parametric and non-parametric identification techniques that will allow the consistent estimation of the parameters of interest (causal mechanism effects). (2) The theoretical developments to be introduced in objective (1) of this project will be applied to Floridas Natality Records Data Sets from various years to analyze the role of gestation time (a potential causal mechanism) in the transmission of the negative effects of smoking during pregnancy on the birth weight of a baby. (3) This will be the first application of causal inference methods to the economic evaluation of Floridas 4-H program. The procedures consist on obtaining information through a survey from 4-H participants and non-participants on their outcomes impacted by the program (grades, standardized tests, etc.) and demographic characteristics. The data will be subsequently analyzed employing causal inference methods. (4) This objective continues previous applications of causal inference methods to the Job Corps training program by the principal investigator. The new aspects are (i) the use of recent methodologies that allow the estimation of causal effects from continuous treatments (as opposed to binary, treated/non-treated treatments), and (ii) an emphasis on the estimation of Job Corps impacts for the state of Florida and the southeastern United States. Relevant data from the National Job Corps Study, a randomized social experiment commissioned by the Department of Labor, is available to the principal investigator.

Progress 10/01/08 to 09/30/13

Outputs
Target Audience: I left the University of Florida in August 2011. This report is just to close the project. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? All goals were accomplished.

Publications

Progress 10/01/10 to 09/30/11

Outputs
OUTPUTS: The basic science results of the project (objective 1) have been disseminated to academic communities. Among the places where presentations were delivered in the reporting year are: the University of Alabama. In terms of objective 3 (the estimation of the effects of the 4-H program), the results have attracted attention from the National 4-H Council. PARTICIPANTS: Nothing significant to report during this reporting period. TARGET AUDIENCES: Nothing significant to report during this reporting period. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.

Impacts
In terms of the basic science portion of the project (objective 1), an extension of my bounding results has been extended to the bounding of the average treatment effect in the presence of invalid instrumental variables. In addition, extensions to bounding average treatment effects (as opposed to local average treatment effects) is under way. One Ph.D. student of this project's investigator is scheduled to complete her dissertation employing methods developed within this project

Publications

• No publications reported this period

Progress 10/01/09 to 09/30/10

Outputs
OUTPUTS: The basic science results of the project (objective 1) have been disseminated to academic communities. Among the places where presentations were delivered in the reporting year are: the New York Camp Econometrics in Blue Mountain Lake, NY and the Fourth Annual Meeting of the Impact Evaluation Network (IEN) at the University of Miami. IEN is a partnership organization of the World Bank and the Latin America and Caribbean Economics Association. This year has been especially productive in terms of the dissemination of results pertaining to objective 3 (the estimation of the effects of the 4-H program). Results have been delivered through invited seminars at Virginia Tech and The Ohio State University (OSU). In addition, a dissemination session with personnel from Ohio 4-H was held during the investigator's visit to OSU. Finally, the results were also presented at the Southern Economics Association Meeting in Atlanta, GA. PARTICIPANTS: Nothing significant to report during this reporting period. TARGET AUDIENCES: Efforts: (i) The basic science components (objective 1) have been delivered through academic manuscripts, seminars, and presentations. They have also been delivered through graduate courses taught by the PI and graduate student contact. (ii) The results for objective 3 have been delivered through two invited seminars and to a group of 4-H faculty at the Ohio State University. (iii) The results for objective 4 have been delivered through academic conference presentations. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.

Impacts
In terms of the basic science portion of the project (objective 1), the following outcomes were generated. In working on the main bounding results for the causal net and mechanism average treatment effects, the investigator and his coauthor realized that the same overall technical strategy could be employed to obtain results to bound the widely-used local average treatment effect (LATE) in the presence of invalid instrumental variables. This important extension of the main basic-science results of this CRIS project, although not part of the original project, was accomplished during this reporting period. In addition, an extension to the bounding of the average treatment effect in the presence of invalid instrumental variables is under way. In terms of applied objective 3, the following outcomes were generated: (i) results indicating that the rate of participation in 4-H programs is positively related to the performance on the math and reading portions of the Florida's Comprehensive Assessment Test (FCAT); (ii) results indicating that 4-H is more effective in urban areas relative to rural areas in the state of Florida; (iii) preliminary results indicating that the rate of participation in the 4-H program is positively related to the performance on the math, reading, and science portions of the Ohio Assessment Test; (iv) preliminary results indicating that the 4-H program seems to be more effective in the state of Florida than in the state of Ohio. In terms of applied objective 4, the following outcomes were generated: (i) results indicating that at most sixty percent of the positive effects of Job Corps on participant's employment and earnings work through the attainment of an academic or vocational degree within the program (this was accomplished through the application of the basic-science results in this project); (ii) results indicating that the attainment of an academic or vocational degree by disadvantaged youth in the U.S. results, on average, in an earnings gain of at most 30 percent and a related increase of the employment probability of at most 10 percentage points (these results also apply the basic-science results in this project). Finally, two Ph.D. students of this project's investigator have started research on applications of the basic-science results to the evaluation of different aspects of the Job Corps program.

Publications

• Flores-Lagunes, A., Gonzalez, A. and Neumann, T. "Learning but not Earning The Impact of Job Corps Training for Hispanic Youths" Economic Inquiry, vol. 48, number 3 (July 2010), 651-667.
• Flores, C.A., Flores-Lagunes, A., Gonzalez, A. and Neumann, T. (2010) "Estimating the Effects of Length of Exposure to Instruction in a Training Program: The Case of Job Corps", Forthcoming, The Review of Economics and Statistics.

Progress 10/01/08 to 09/30/09

Outputs
OUTPUTS: The basic science results of the project (objective 1) have been disseminated to academic communities. The vehicles of this dissemination are seminars at Universities and research centers, and presentations at scientific conferences. Among the places where seminars were delivered in the reporting year are: CEPS/INSTEAD (Luxembourg), University of Wisconsin, Georgia State University, Universidad Autonoma de Nuevo Leon (Mexico), and the Colegio de Economistas de Nuevo Leon (Mexico). The scientific conference presentations include: Southern Economics Association Annual Meeting (San Antonio, TX), Latin American Meetings of the Econometric Society (Buenos Aires, Argentina), European Meetings of the Econometric Society (Barcelona, Spain), New York Camp Econometrics (Lake Placid, NY), American Economics Association Annual Meeting (San Francisco, CA), Latin American Meeting of the Econometric Society (Rio de Janeiro, Brazil), Midwest Econometrics Group Meeting (University of Kansas). There have also been preliminary applied results of the project (objectives 3 and 4) that include the estimation of causal effects of (i) the Job Corps federal training program in the U.S. south east and Florida, and (ii) the 4-H program on standardized test outcomes in Florida. The dissemination of these results, given their preliminary nature, have taken place at informal groups at the University of Florida that include researchers interested in the topic (e.g., personnel from Florida 4-H). PARTICIPANTS: - A non-formal collaboration has been established with the Florida 4-H office whereby they have provided us with data for objective 3. TARGET AUDIENCES: Efforts: (i) The basic science components (objective 1) have been delivered through academic manuscripts, seminars, workshops, and presentations. They have also been delivered through graduate courses taught by the PI and graduate student contact. (ii) The preliminary results for objective 3 has been delivered through a seminar to selected 4-H faculty at the University of Florida. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.

Impacts
In terms of the basic science portion of the project (objective 1), the following outcomes were generated: (i) results about bounding the causal net average treatment effect on an outcome under a randomized treatment for the case where both treatment and mechanism are binary; (ii) an important by-product result whereby a test for the validity of an instrumental variable in just-identified models can be devised. In terms of the applied portion of the project (objectives 3 and 4), the following outcomes were generated: (i) preliminary results indicating that the rate of participation in 4-H programs is positively related to the performance on the math and reading portions of the Florida's Comprehensive Assessment Test (FCAT); (ii) preliminary results indicating that the federal Job Corps training program for disadvantaged youth has higher effectiveness in the Southeastern region of the U.S. relative to the nation, whereas in the state of Florida the program shows similar effectiveness as that of the nation.

Publications

• Flores-Lagunes, A., Gonzalez, A. and Neumann, T. (2009). "Learning but not Earning The Impact of Job Corps Training for Hispanic Youths", Economic Inquiry (in press).
• Flores, C.A. and Flores-Lagunes, A. (2009). "Identification and Estimation of Causal Mechanisms and Net Effects of a Treatment Under Unconfoundedness", IZA Discussion Paper Series No. 4237.
• Flores-Lagunes, A. and Timko, T. (2008). "Importance of the Economic Evaluation of Florida's 4-H", EDIS FE753, Department of Food and Resource Economics, Florida Cooperative Extension Service, Institute of Food and Agricultural Sciences, University of Florida.
• Mansoor, Abu. (2009). "Evaluating the Effectiveness of the Job Corps Program in the South: Does it Work Where it Matters the Most", Master of Science Thesis, University of Florida (Advisor: Dr. A. Flores-Lagunes).