Skip to main content

Quantifying the 2021 Texas Freeze with Satellite Econometrics

In February 2021, Winter Storm Uri descended on Texas, triggering one of the most catastrophic infrastructure failures in modern American history. Temperatures plunged, demand spiked, and the state’s isolated power grid—operated by ERCOT—collapsed. Millions lost power for days.

While utility companies provided their own outage estimates, we can use remote sensing to independently verify the scale of the economic damage. Using Python, Google Earth Engine, and the Difference-in-Differences (DiD) framework, we analyzed the blackout from space.

The Natural Experiment: A Tale of Two Grids

To measure the true impact of the grid failure, we cannot simply look at Houston in isolation. We need a control group—a city that faced the same freezing temperatures but did not lose power.

Enter El Paso. Unlike the rest of Texas, El Paso is not on the ERCOT grid; it is connected to the Western Interconnection. This creates a perfect "Natural Experiment":

  • Treatment Group: Harris County (Houston) – ERCOT Grid (Failed).
  • Control Group: El Paso County – Western Grid (Functional).

Visualizing the Shock

We utilized VIIRS Nighttime Lights data, which measures the radiance of light emissions from Earth. By comparing the monthly trends, we can see the "Parallel Trends" leading up to the disaster.

Figure 2: Spatial Analysis & Economic Impact. Left: El Paso (Control) shows stability. Right: Houston (Treatment) shows widespread power loss (Blue areas).

The Economic Toll

Our regression model confirmed a 20.37% drop in monthly radiance for Houston. But what does that mean in dollars?

Converting light loss to economic loss is a challenge, but we applied two standard econometric methods to estimate the damage:

Metric Methodology Est. Loss (Houston)
Business Interruption Light-to-GDP Elasticity (1:1) ~$7.6 Billion
Welfare Loss Value of Lost Load ($9,000/MWh) ~$15.8 Billion
Key Finding:
A 20% monthly drop in light implies a near-total blackout for the 5-day duration of the storm. While satellites capture the immediate grid failure cost, they likely underestimate the total damage (estimated at >$80B) because they cannot see internal physical destruction, such as burst pipes.

Appendix: Deriving the Damage

The estimates above are derived directly from the regression coefficients and Harris County economic data. Here is the step-by-step math:

1. The Radiance Drop

We first calculate the percentage of light lost by dividing the "Shock" (DiD Coefficient) by the "Baseline" (Intercept + Natural Difference).

Formula % Drop = | DiD Coefficient / (Intercept + Treat) |

Calculation % Drop = | -6.83 / (9.30 + 24.23) |
% Drop = 0.2037 (20.37%)

2. Business Interruption (GDP)

Using the Henderson Method, we assume a 1:1 elasticity between light loss and GDP loss for acute disasters.

Formula Loss = (Annual GDP / 12) × % Drop × Elasticity

Calculation Loss = ($450 Billion / 12) × 0.2037 × 1.0
Loss = $37.5 Billion × 0.2037
Loss = $7.64 Billion

3. Welfare Loss (VOLL)

Using the "Value of Lost Load," we calculate the social cost of the unserved energy at the price cap set by ERCOT ($9,000/MWh).

Formula Loss = (Avg Load × Hours in Feb) × % Drop × Price Cap

Calculation Total Energy = 12,000 MW × 672 Hours = 8,064,000 MWh
Lost Energy = 8,064,000 × 0.2037 = 1,642,636 MWh
Loss = 1,642,636 × $9,000
Loss = $14.78 Billion

Analysis generated using Google Earth Engine & Python.

Comments

Post a Comment

Constructive feedback is always welcome. Thank you

Popular posts from this blog

Mapping the Blast Radius: What Happens When the "World's Factory" Stops?

In global trade, "efficiency" often masks "fragility." We know that China is central to the global electronics supply chain, but how central? And if that node were to go dark, who would feel the shockwaves first? To answer this, I moved beyond standard trade statistics and built a network simulation using the OECD Inter-Country Input-Output (ICIO) Tables (2023 Edition) . This dataset maps the DNA of the global economy, tracking every dollar of input across 66 countries and 45 industries. 1. The Blast Radius: Tracing the Contagion I treated the global economy as a directed graph and simulated a total supply shock to Chinese Electronics (CHN_C26) . By tracing the flow of inputs across three tiers of buyers, I visualized the "Blast Radius" of this disruption. Fig 1: The Supply Chain Cascade. The shock originates in China (Red) and immediately hits "Tier 1" assembly hubs (Dark Blue) before cascading to global consumers...
Post a Comment
Read more

Turkey's Informality Tax

A general-equilibrium model of Turkey's dual labour market says the cost of taxing formal work doesn't show up where we usually look for it. The bottomline: Making Turkey's transfer to the unemployed a third more generous raises unemployment from 8.6% to 9.2% . That is the honest cost, and it is not large. How you pay for it matters more than whether you pay for it. Funded by payroll taxes, the poor end up 0.6% worse off than before the transfer was raised — the policy defeats itself. Funded by VAT, they are 0.5% better off . The reason is not unemployment, which is nearly identical under both. It is informality . A higher payroll tax pushes formal jobs into the unregistered sector, where the wage is 43% lower and nothing is taxed. Turkey is not on the wrong side of the payroll-tax Laffer curve — revenue peaks around 53%, well above today's ~37.5% wedge. But the marginal cost...
Post a Comment
Read more

The Labor Market Impacts of AI in Türkiye: Beyond Occupational Exposure Analysis

Beyond Exposure AI and Labor Markets · Türkiye Occupational exposure to AI identifies where technology may affect work, but not whether employment will contract, reorganize, or expand. Applying the Richmond (2026) AI Jobs Transition Framework to Türkiye's 2024 Household Labour Force Survey, the dominant high-exposure archetype is not automation. It is reorganization. Mpumelelo Nxumalo and Dhushyanth Raju · Summary of the working paper Beyond Exposure: AI and Labor Market Reorganization in Türkiye Most discussion of AI and jobs runs on a single number: how much of an occupation's task content a model can perform. That number identifies where technology reaches. It does not say what happens next. An exposed occupation may shed workers, or it may keep them while the work itself is rebuilt around the technology. Those are different outcomes, they call for different policy, and exposure alone cannot tell them apart. Applying the Richmond (202...
Post a Comment
Read more