I Am
Eric
Penner
Current Position
Postdoctoral Scholar, University of California, Berkeley
Education
Ph.D., Economics, University of Colorado at Boulder
Field: Econometrics, Industrial Organization, Cognitive Economics
Thesis: Essays in Semiparametric Structural Estimation
Advisor: Professor Carlos Martins-Filho
M.A., Economics, University of Colorado at Boulder
B.S., Mechanical Engineering, University of Wyoming
Selection of Heterogenous Instruments in Fixed Effect Panel Model
A multistage estimator dealing with endogeneity in a
panel data context when the appropriate instruments for each cross
section is an unknown subset of a large collection. Lasso is utilized
to perform model selection.
Working Paper
Monte Carlo Study
Presentation
Oracle Efficiency and Stability in Additive Partially Linear Triangular Systems
A multi stage estimator which deals with endogeneity in
additive partially linear regression and examines the stability of the
estimator relative to small changes in sample data.
Working Paper
Presentation
Representation and Judgement: Memory and Sequencing in Performance Evaluation
I develop and estimator which explicitly accounts for the
role that memory and cognitive heuristics have on evaluative tasks. In
particular, incorporate the "Evaluation by Moments" hypothesis which
supposes that a judge assigns a judgment to a sequence of actions relative
to a mental representation that the judge has constructed.
Working Paper
Queuing and Capital Investment
I simulate a network of suppliers and distributors over time
where distributors must resupply periodically under stochastic inelastic
demand conditions with terminals who also have to resupply under stochastic
supply shortages. Distributors have a choice of investing in either
permanent capital (storage) or flexible capital (trucks for resupply).
Coming Soon
Information Visualization Display Using Associative Clustered Tiling and Tesselation
The amount and specificity of the information that each of us have access to is unprecedented. Inquiries once requiring a lengthy trip to the library or a long wait for the next edition of the local newspaper are now resolved in seconds and from practically anywhere our curiosity is aroused. Unfortunately, the typical manner in which collections of information are electronically presented fails to provide us with a sense of the interconnection between those pieces of information. Here we develop a method by which information can be electronically presented users that provides this sense of connection and interrelatedness allowing a viewer to easily navigate their way to a fuller understanding of a topic.
Professional Membership
Econometric Society, American Economic Association
Software and Programming Language Pro ciency
Software: Git, Jupyter Notebook/Lab/Hub, Rstudio, Matlab, STATA, Microsoft Office
Language: python, R, matlab, HTML, CSS, Javascript
Packages: Numpy, Pandas, Scikit-Learn, D3
RBN Energy, Consultant
- Algorithm Development
- Patent Writing
- Energy Industry Analysis
Developing software to find optimal data presentation
/ clustering configurations using combinatorial optimization
procedures such as random search and genetic algorithms, with
potential for the future incorporation of a position auction. I
have written energy industry articles discussing the economics of
shale oil production and decline curves in both domestic and
international markets, natural gas storage and transportation, and
the petrochemical industry. Additionally, I have prepared
presentation materials, and assisted on consulting projectsself.
Blog Posts
Jacobs Engineering Group, Mechanical Engineer
- Lead Mechanical and Project Engineer
- Energy Industry
- Food and Beverage Industry
Completed design, procurement, and construction phases of several large capital projects in the oil, natural gas, food, and beverage industries. Effectively used my knowledge of the design and code requirements for pressure vessels, pumps, compressors, heat exchangers, fired heaters, distillation columns, dehydration units, pipelines, process piping, water treatment systems, and amine systems toward successful project completion. Lead mechanical engineer for multi-million dollar projects for BP America and MillerCoors LLC. Successfully completed projects for Questar, Frontier Refining, BP, Exxon Mobil, MillerCoors, Integrys, and Enbridge Inc. Most recently held the position of Project Engineer on the Husky Diluent Metering Project.
US Army, Sergeant
- Squad Leader
- NATO Peacekeeper in Bosnia and Herzegovina
- Post 9/11 Veteran
Reconnaissance Specialist and Squad Leader spending significant time in dynamic and challenging environments such as South Korea and Bosnia and Herzegovina. Seven months service NATO peacekeeping forces in Bosnia and Herzegovina specializing in public relations. Selected for promotion ahead of peers throughout decorated service record due to personal integrity, dedication to hard work, attention to detail, enforcement of safety, and peer mentorship.
Machine Learning to Exploit the Unique Opportunities Provided by Big Data
- Econometrics
- High Dimensional Data
- Unobserved Heterogeneity
Model selection, analyzing unstructured data, balancing exploitation and exploration in assign- ing treatments, and proxy variable generation are but a few of the issues that Economists, and Data Scientists once dreamed of having. Thanks to the rapid development of the field of machine learning, tools now exist to accomplish tasks once dismissed as infeasible. My research here is focused on developing econometric tools to help researchers leverage machine learning techniques in order to construct causal inference and credible counterfactuals. My working paper Selection of Heterogenous Instruments in Fixed Effect Panel Model is an estimator which incorporates both a lasso shrinkage estimator to provide a data driven means of selecting valid heterogenous instruments, and a sample splitting routine in manner suggested by Chernozhukov et al. (2017). The original inspiration for this estimator was to improve the estimation of log technological change in Augmented Solow-Hall type regressions like the one found in Basu et al. (2006).
Developing Estimators That Account for Human Cognitive Processes
- Cognitive Economics
- Memory and Judgement
- Econometrics
Human decision making is notoriously difficult to analyze. We have inherent biases, we dont pay as much attention as we should, our memories are neither fixed nor as good as was previously thought, and we use heuristics far more than we would like to admit. In order to accurately analyze the decisions that consumers, producers, performers, and judges make one must depart from the simplifying assumption of total rationality, and model decision making as it more realistically happens. New theoretical frameworks for understanding these departures are numerous, but how can we estimate that parameters of a corresponding structural model when attributes like opinion, attention, and motivation are considered unobservable using with traditional methods. We can inform the estimation of these parameters with techniques like sentiment analysis, computer vision, and word embeddings. My research in this area is focused on utilizing machine learning tools to inform the estimation of structural models of departures to purely rational decision making.
Capital Investment Under Uncertainty, Product Switching, and Congestion Constraints
- Computational Economics
- Industrial Organization
- Rare and Extremely Costly Events
How do companies determine the correct mixture of capital investment types in order to prevent customer product switching in the face of uncertain market conditions? For firms, there is always uncertainty about what elements of their current business will ultimately cost them customers and how to invest capital to prevent this. For example developers of an online multi- player game (PUBG) may have to choose between fixing a software bug and renting server space while accounting for uncertain network congestion and competitors in the market (Fortnite). Often, in order to incorporate the full complexity of market conditions these questions must be answered computationally. I simulate the operation of such markets in order to determine a decision rule for the best mixture of capital types for a given set of market conditions. These decision rules are of independent interest to academic economists but what value can a small firm derive from them given that they can observe only a small subset of the market conditions needed for an optimal solution. My research here is focused on methods for adaptively reducing the dimension of the decision rule space to account for the information available to market firms.
University of California, Berkeley
W203 Statistics for Data Science
The goal of this course is to provide students with a
foundational understanding of classical statistics and how it fits within
the broader context of data science. Students will learn to apply the most
common statistical procedures correctly, checking assumptions and responding
appropriately when they appear violated. They will also learn to evaluate
the design of a study and how the variables being measured relate to research
questions. The course begins with a focus on exploratory analysis and
descriptive statistics. From there, we learn how statistical models are
built using the structure of probability theory. Next, we use the simple
example of the mean to demonstrate the use of estimators and hypothesis
tests. We then turn to classical linear regression, taking several weeks to
build a strong understanding of this central topic. Our treatment stresses
causal inference and includes a discussion of omitted variables. At the end,
we describe some of the concerns that arise in the process of specifying
linear models. Throughout the course, students will practice analyzing
real-world data using the open-source language, R.
Sample Lecture
Sample Assignment ,
University of Colorado, Boulder
ECON 3818 Introduction to Statistics with Computer Applications
This is a one-semester course in statistics, and is required for Economics majors. This is a math class and we will study probability, random variables, probability distributions especially the normal distribution, and descriptive and inferential statistics including estimation and hypothesis testing. Emphasis is on both theory and application.
ECON 1078 Math Tools for Economists (I)
The purpose of this course (and the following course) is to equip students with the necessary mathematical knowledge to be successful in the Economics courses given at the University of Colorado at Boulder
University of Colorado, South Denver
ECON 2010 Principles of Microeconomics
Economists study the process by which individuals and firms make decisions in a rational manner given their set of circumstances. Economists are concerned with the consequences for individuals, firms, and the economy of decisions made this way. Thus, in studying economics, reasoning skills are much more important than memorization skills. Having to think and reason (that is, make decisions) rather than just memorize makes economics a challenging subject for many students. On the other hand, studying economics also develops reasoning and analytical skills.