Skip to main content

Measuring the Invisible: California Cap-and-Trade Analysis

How do we measure the impact of a policy? In a laboratory, you have a control group. In the real world, we only have one California. We cannot observe the "counterfactual"—a parallel universe where California never passed its landmark Cap-and-Trade bill (AB 32) in 2013.

However, using econometrics and Python, we can build that parallel universe.

Using the Synthetic Control Method (SCM), I analyzed real-world energy data to estimate the true impact of carbon pricing on California's emissions. Here is what the data says.

The Methodology

The Synthetic Control Method estimates the effect of an intervention (the policy) by comparing the treated unit (California) to a weighted combination of control units (other US states) that did not implement the policy.

To ensure a robust "apples-to-apples" comparison, I pulled data directly from the EIA Open Data API (v2), controlling for:

  • Covariates: Real GDP per Capita (economic health) and Population density.
  • Donor Pool: I specifically restricted the "Donor Pool" to states with "cleaner" energy grids and service-based economies (e.g., New York, Massachusetts, Washington) rather than coal-heavy industrial states (like Texas or Ohio), which would bias the pre-trend fit.
  • Smoothing: I applied a 3-year rolling average to the time series. This is critical for energy analysis, as it filters out high-frequency "weather noise" (heatwaves or wet winters) to reveal structural economic trends.

The Result: California vs. "Synthetic California"

The Python model constructed a "Synthetic California" composed of a weighted average of peer states. By comparing the two, we can visualize the policy's impact.

Figure 1: Actual California Emissions (Blue) vs. Synthetic Counterfactual (Red). The green arrow indicates policy success.

Analysis: Three Distinct Eras

The data tells a nuanced story in three parts:

1. The Validation (2009–2013)

Before the policy goes into effect, the "Actual" (Blue) and "Synthetic" (Red) lines move in parallel. This confirms the "Parallel Trends Assumption"—a prerequisite for SCM. It proves that our synthetic model effectively mimics California’s economy in the absence of the policy.

2. The "Drought Penalty" (2015–2018)

Unexpectedly, California appears to perform worse than the synthetic control in the mid-2010s. This is a classic example of omitted variable bias due to climate shock. Between 2011 and 2017, California suffered a historic drought. Hydroelectric generation plummeted, forcing the grid to rely heavily on natural gas. While the policy was working, its signal was temporarily drowned out by the climate reality.

3. The Divergence (2019–2021)

As the drought subsided and the emissions cap tightened, the lines diverged sharply. While the donor states continued a gradual decline, California’s emissions dropped significantly faster. By 2021, the gap suggests that Cap-and-Trade is actively driving emissions down beyond the national baseline.

Comparing to the Literature

This analysis aligns with authoritative studies in environmental economics, which suggest that carbon pricing works, though its effects are often gradual and compounding.

  • Hernandez-Cortes & Meng (2023): In their paper "Do Environmental Markets Cause Local Pollution Inequities?", the authors used similar rigorous causal inference methods to find that California's Cap-and-Trade program reduced GHG emissions and local air pollution, particularly narrowing the gap between disadvantaged and wealthy communities.
  • Bayer & Aklin (2020): Applying SCM to the European Union Emissions Trading System (EU ETS), they found that carbon pricing reduced emissions by approximately 3.8% initially, with effects accumulating over time—mirroring the "delayed divergence" we see in my California plot.
  • Severin Borenstein (UC Berkeley): Borenstein has noted that while the "Cap" is effective, complementary policies (like renewable portfolio standards) often do the heavy lifting, making it difficult to isolate the price signal from the regulatory signal.

Conclusion

Data science allows us to see the invisible. While a simple line chart of California's emissions shows a decline, it doesn't tell us why. By building a Synthetic Control, we can confidently argue that despite climate headwinds, California's carbon pricing mechanism is delivering measurable abatement results that exceed the national average.

Comments

Post a Comment

Constructive feedback is always welcome. Thank you

Popular posts from this blog

Unemployment by state in the USA

Below is a visualization of unemployment rates by county using a powerful Python library called Bokeh . The two maps are for the states of Texas and the Commonwealth of Massachusetts. As the second largest economy in the United States (10th largest in the world), Texas shows interesting county variation in macroeconomic factors such as unemployment. According to Wikipedia , in 2015, Texas was home to six of the top 50 companies on the Fortune 500 list and 51 overall (third most after New York and California). The northern counties were least affected by the financial crisis of 2008/09. Dallas–Fort Worth–Arlington area encompasses 13 counties which constitute the economic and cultural hub of the region commonly called North Texas or North Central Texas. Bokeh Plot The least affected counties in Massachusetts were the southernmost tourist areas of Nantuckett and Dukes County. The ...
Post a Comment
Read more

Modeling Core PCE inflation: A dual approach

Today's release of the August 2025 Personal Consumption Expenditures (PCE) inflation data drew widespread media attention, with coverage highlighting both the persistence of inflation and its implications for Federal Reserve policy. Across outlets, analysts pointed to resilient consumer spending and income growth as signs of underlying economic strength, even as inflation remains above the Fed's 2% target. The consensus among media reports is that while inflation is not worsening, its stubbornness continues to challenge policymakers navigating a softening labor market and evolving rate expectations. To provide deeper insights into inflation's trajectory, I've developed a forecasting framework that combines two econometric approaches — ARIMA time series modeling and Phillips Curve analysis—to predict Core PCE inflation. This analysis presents a unique opportunity to validate my forecasting methodology against eight months of 2025 data. ...
Post a Comment
Read more

Do Minimum Wage Increases Really Kill Jobs? Evidence from the "Fight for $15" Era

The debate over minimum wage policy has raged for decades, with economists, policymakers, and business leaders offering sharply different predictions about its effects on employment. Critics warn that raising the minimum wage will force employers to cut jobs, while supporters argue that higher wages boost worker productivity and spending power. But what does the actual data tell us. Using a comprehensive difference-in-differences analysis and Federal Reserve Economic Data covering 43 U.S. states from 2012-2020 of the "Fight for $15" movement between 2012 and 2020, I provide some evidence about how minimum wage increases actually affect employment in the real world. The Perfect Natural Experiment The period from 2012 to 2020 provided economists with an ideal "natural experiment" to study minimum wage effects. Here's why this timeframe was perfect for analysis: Federal Stability : The federal minimum wage remained frozen at $7.25 per hour since 2009, creating ...
Post a Comment
Read more