Lecture 1 Introduction

Gang He

March 9, 2024

About us

Professor Gang He

Professor Samantha MacBride

Welcome and icebreaker

• Rose: “something that is working well or something positive”
• Thorn: “something that isn’t working or something negative”
• Bud: “an area of opportunity or idea yet to be explored”

Why PAF 9174?

Goal: sensible program evaluation at reasonable cost

1. Funder
2. Government/NGO/Public Institutions/Citizens
3. Program manager/staff
4. Executives
5. Evidence-based policy and management decisions

Organization

• Syllabus
• Blackboard
• Office hour
• Case studies
• Evaluation report

John Snow, Cholera, the Broad Street Pump

Cost saving of global supply chain

Performance analysis of the Queens Residential Curbside Organics Program

Program & Program Evaluation

Program:

A program is a set of resources and activities directed toward one or more common goals, typically under the direction of a single manager or management team.

Program Evaluation:

Program evaluation is the application of systematic methods to address questions about program operations and results. It may include ongoing monitoring of a program as well as one-shot studies of program processes or program impact. The approaches used are based on social science research methodologies and professional standards.

Capturing the wedege

Theory of change

flowchart LR
  A(IRA) --> B(Subsidies)
  B --> C(Investment)
  B --> D(Jobs)
  C --> E(Domestic Manufacturing)
  D --> E(Domestic Manufacturing)

Big data and AI

Perspective:

• Universe created:

  • 13.8 billion years ago

• Earth created:

  • 4.54 billion years ago

• Modern human:

  • 300,000 years ago

• Civilization:

  • 12,000 years ago

• Written record:

  • 5,000 years ago

• 1943: Neural networks
• 1957: Perceptron
• 1974-86: Backpropagation, RBM, RNN
• 1989-98: CNN, MNIST, LSTM, Bidirectional RNN
• 2006: “Deep Learning”, DBN
• 2009: ImageNet
• 2012: AlexNet, Dropout
• 2014: GANs
• 2014: DeepFace
• 2016: AlphaGo
• 2017: AlphaZero, Capsule Networks
• 2018: AlphaFold 1, BERT
• 2021: AlphaFold 2
• 2023: ChatGPT
• 2024: Sora, Gemini

NYC Open Data

Open NY

Federal Data

Data science for Social Good

Data science

• Collect
• Analyze
• Communicate

Causal Revolution

Evidence based medecine

Evidence based policy

Example: RAND’s Health Insurance Experiment (HIE)

Key findings:

• Cost sharing reduced spending for health care services
• Participants with cost sharing made fewer medical visits and were admitted to hospitals less frequently
• Reduced spending resulted entirely from less use of care; the costs of care were not affected
• Cost sharing had no detrimental effects on participants’ health, except for the sickest and poorest patients

Not all policies are based on evidence

• Value
• Politics
• Feasibility
• Quality of evidence and research

Program Evaluation

Data Science + Causal Inference

Case studies

• Food service resuse system for Governor’s Island
• Northwest Housing Associates
• Jobs Plus in New York City
• Advanced Research Projects Agency-Energy
• Benefits and Costs of the Clean Air Act 1990-2020

References

Helveston, John Paul, Gang He, and Michael R. Davidson. 2022. “Quantifying the Cost Savings of Global Solar Photovoltaic Supply Chains.” Nature 612 (7938): 83–87. https://doi.org/10.1038/s41586-022-05316-6.

Newcomer, Kathryn E., Harry P. Hatry, and Joseph S. Wholey. 2015. Handbook of Practical Program Evaluation. 4th edition. San Francisco: Jossey-Bass.