About Me

I am a data scientist at Netflix. Previously I was an economist and machine learning engineer at Instacart. I also work on sports analytics and other entertainment products and applied research. I graduated with my PhD in economics from Stanford GSB in January 2022. My academic research is generally in applied microeconomics and econometrics where I utilize a mixture of causal inference, machine learning and optimization methods to answer unique research questions. In grad school I worked on topics in labor, child welfare and families, education, and inequality. My job market paper (my main big academic paper) focused on understanding which foster children benefit most from placement with foster families relative to institutions and how that informs optimal foster care policy.

Email: camerontphd@gmail.com

CV

Products and Applied Research

Sports Analytics

NBA Watchability

Building the Average Watchability Index (aWI) of daily NBA games to visualize and quantify the watchability of an NBA game and help people decide what games to watch with their limited time. Combining the expected competitiveness and quality of teams playing to predict the overall watchability of the basketball being played. Built into Streamlit app.

Field Vision Sports

Co-founder and Chief Scientist for Field Vision sports which focused on projecting outcomes for all skill position players and teams for NFL games integrating detailed play-by-play data and Vegas data. Acquired by Rafter Sports in 2025.

Academic Research

Accepted and Forthcoming Papers

Who Gets a Family? The Consequences of Family and Group Home Allocation for Child Outcomes (Conditionally Accepted at AEJ: Applied)

Abstract

Hundreds of thousands of children grow up in the US foster care system every year and are at high risk of experiencing negative outcomes such as incarceration and homelessness. This paper documents how the placement of foster children into families rather than group homes improves their outcomes using the exits of other children from families as an instrument for their placement setting. Policies that change which children are matched to families can achieve a large percentage of the gains from policies that add families to the foster care system due to heterogeneity in treatment effects.

Why Do Families Foster Children? A Beckerian Approach, (Review of Economics of the Household)

Final Manuscript, Published Version, Journal Page

Abstract

Less than half of the hundreds of thousands of abused and neglected children in foster care are able to find a foster family to take care of them while the rest are placed in restrictive group home settings. This paper proposes that households choose to foster children following a Becker-style model in which households maximize the human capital of the children they care for and can receive human capital flows from both foster children and biological children. The demand for foster children and the age of foster children depends on the number of biological children and the household wage. I test the main predictions of the model using twins as an instrument and a rich set of household observable characteristics. A parameterized version of the model suggests that the substitutability of foster children and biological children is a stronger lever affecting fostering than foster care subsidies, and the wage of a household is almost as important as the subsidy in determining fostering.

Working Papers

Information and Risky Behavior: Model and Policy Implications for COVID-19

Abstract

This paper studies a contagion model where individuals can take risky or safe actions to study the effects of testing and fines on disease spread and welfare. Testing gives agents knowledge to better assess the costs of exposing themselves to a disease. Whether testing increases or decreases disease spread depends on the private costs of the disease. If the private costs are small enough, then testing individuals increases infection. If the private costs are large enough, then testing individuals decreases infection. Punishing individuals for exposing themselves and others to the disease while also providing testing can also increase disease spread. Welfare in the economy is also examined in a simplified version of the model. Policy implications for public health responses to pandemics are discussed along with an application to crime.

Skill Ladders

Abstract

This paper presents a model of skills and derives properties of the optimal investment into educational skills. In the model students can acquire basic and advanced skills at a cost to a policymaker who is budget-constrained. The optimal policy is very sensitive to the structure of the returns to skill - even when advanced skills give unbounded marginal returns, it may be optimal to invest more in basic skills if skills represent a ``skill ladder''. These results offer new interpretations on the existing empirical evidence on education interventions. There is a single object that determines whether to invest more in basic or advanced skills and whether the skill ladder model applies. I develop a methodology to estimate the returns to skills and this object and apply it to mathematics (advanced skill) and self-esteem (basic skill) in the NLSY. The results show that the returns to skill reflect that the true state of the world is between the two stark viewpoints and that there is substantial racial heterogeneity in the returns to skills from the lens of the model, suggesting that there may be benefits to focusing more on basic skills in educational policy making and that optimal skill targeting may differ by race.

Information Goods

Abstract

The main goal of this paper is to understand how people will change their information acquisition strategies as information sources become more or less costly. To do this, I develop a model of information acquisition in the spirit of traditional consumer theory that treats information sources, which are distinct dimensions of the state space, as different consumption goods. A general form of the model shows that as information becomes more costly, people will demand less of it, and also characterizes when information sources are substitutes or complements. The models insights are extensively analyzed in two settings: determining the optimal firm recruiting strategy when considering technical and social skills, and determining the optimal way to evaluate students using testing and assessing creativity. Other insights into dating and media consumption are also discussed.

What Makes a Movie Great?

Details

I explore rich movie level data to understand the movie making "production function" and answer important questions in labor economics using synthetic controls and difference-in-differences methods.

Teaching, Classwork and Miscellaneous

Other Writing

NFL Deep Learning Project (poster)

Details

Project using NFL pre-play image and situational data with convolutional neural networks and transfer learning to predict play outcomes including yards gained and offensive play call.

Machine Learning Cheatsheet

Details

A machine learning cheatsheet to de-mystify some major machine learning methods for those with intermediate statistics and econometrics backgrounds. Also useful as a condensed reference for high-level overview of the methods.

Blog (MUCH OLDER AND NO LONGER UPDATED)

I wrote a few short articules about popular subject topics (mainly sports) using economics and statistics.

Teaching

Stanford GSB Research Fellow Econometrics Bootcamp 2019-2021 (1 week bootcamp on econometric methods)

Course Assistant MGTECON 603 Econometrics Methods I Fall 2019 (PhD course on statistical foundations of econometrics) Section Notes

Course Assistant HRMGT 302 Incentives and Productivity (MBA advanced economics)

Course Assistant OIT 274 Data and Decisions (MBA intro to econometrics and data science)

Course Assistant ALP 301 Data-Driven Impact (MBA / Masters / PhD applied machine learning class)