📊 statistics

Mixed Frequency Stochastic Frontier Model: with application to the linkage of weather extremes and firm efficiency

This paper proposes a mixed-frequency stochastic frontier model that utilizes nonparametric functions and hybrid backfitting estimation to resolve skewness issues and analyze the impact of high-frequency extreme weather events on the efficiency of electric cooperatives in the Philippines.

Original authors: Erniel B. Barrios, Nur Syazwani Mazlan, Lim Foo Weng, Paolo Victor T. Redondo, Gian Karlo M. Torreno, Lee How Chinh

Published 2026-08-03

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Erniel B. Barrios, Nur Syazwani Mazlan, Lim Foo Weng, Paolo Victor T. Redondo, Gian Karlo M. Torreno, Lee How Chinh

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to grade a class of students on how well they solve math problems. You have their final test scores (the output), and you know how much time they studied and how many textbooks they owned (the inputs). But here's the catch: you also know that some students got sick, some had a noisy house, and some had a sudden thunderstorm that knocked out the power right before the exam. These "bad luck" factors aren't part of their natural ability, but they drag down their scores. In the world of economics, this is called a Stochastic Frontier Model. It's a fancy way of trying to draw a line representing the "perfect" performance and seeing how far each person or company falls short of that line. The tricky part is figuring out why they fell short. Was it because they were underperforming (inefficiency), or just because the storm knocked out the power (random noise)?

Now, imagine a new problem: the test scores are recorded once a year, but the weather data (the storms) is recorded every single day. If you just lump all the daily storms into one big "annual storm" number, you lose the scary details. You might miss the fact that one massive hurricane hit in July, even if the rest of the year was sunny. This is the "mixed frequency" puzzle. This paper tackles that puzzle by building a new kind of math model that can handle daily weather data while still looking at yearly business results. The authors are curious about how extreme weather, like typhoons and heavy rain, actually messes up the efficiency of electric companies in the Philippines. They want to know: when the wind howls and the rain pours, do these companies just get a little slower, or do they completely lose their way?

The Story of the Storm and the Spark

The authors of this paper, a team of researchers from universities in Malaysia, the Philippines, and Saudi Arabia, decided to investigate the electric cooperatives in the Philippines. These are the local groups that keep the lights on in rural areas. They noticed that while these companies have annual financial reports, the weather that threatens them—typhoons with winds over 300 KPH and massive rainfall—happens on a monthly or even daily scale.

The team realized that the old way of doing things was like trying to describe a hurricane by averaging the wind speed for the whole year. If a typhoon hits in July but the rest of the year is calm, the "average" wind speed looks weak, hiding the fact that the power lines were almost torn apart. The old models also had a math problem: they often got stuck or gave weird answers (like saying a company was 100% efficient even when it clearly wasn't) because they forced the "bad luck" numbers to fit a specific, rigid shape.

To fix this, the researchers built a Mixed Frequency Stochastic Frontier Model. Think of it as a super-smart detective that doesn't just look at the yearly report card. Instead, it peeks at the daily weather logs to see exactly when the storms hit. They used a special mathematical tool called a "logistic function" (imagine a smooth S-shaped curve) to make sure the "inefficiency" score could never be negative, which solved the math headaches that usually trip up these models. They also used a "backfitting algorithm," which is like a game of "hot and cold" where the computer keeps adjusting its guess about the weather's impact until it finds the perfect fit, rather than trying to solve everything in one giant, impossible leap.

They tested this new model on data from 114 electric cooperatives in the Philippines from 2010 to 2022. They compared their new "weather-aware" model against the old, standard models that just averaged the weather data out.

What They Found

The results were pretty clear. The old models were struggling. In many regions, the standard math simply refused to work (it "failed to converge"), meaning the computer couldn't find an answer at all. When it did find an answer, it often claimed the electric companies were nearly perfect, with efficiency scores hovering near 100%, even though everyone knows typhoons cause real damage.

The new model, however, told a different, more realistic story. It showed that when you actually look at the specific months with the worst winds and rain, the electric companies' efficiency drops significantly.

The Weather Matters: The study found that extreme weather events, like typhoons with maximum sustained winds and heavy rainfall, directly cause operational inefficiency. For example, in 2019, a year with strong typhoons, the efficiency of these cooperatives dropped to as low as 36% to 40% in some models, whereas the old models still thought they were doing great (around 73% to 97%).
The "Wrong" Skew: The paper explicitly argues against the idea that we can just ignore the timing of the storms. By averaging the data, the old models were "masking" the true impact of the disasters. The new model proved that you cannot simply lump high-frequency weather data into low-frequency business reports without losing the truth.
Simulation Proof: To be sure their new math wasn't just a fluke, the researchers ran 200 computer simulations. In these tests, their new model consistently estimated the true efficiency much closer to the actual numbers than the old model did. The old model's errors were often three times larger than the new model's errors.

The paper concludes that while electric cooperatives can't stop the typhoons, they can't pretend the storms don't happen. The new model provides a better way to measure how much these storms hurt their performance. It suggests that regulators and the companies themselves need to look at the specific timing of extreme weather to understand their true operational struggles. By using this new approach, they can better plan for resilience, perhaps by investing in stronger poles or better technology to handle the specific months when the wind blows hardest, rather than just hoping for the best based on a yearly average. The paper doesn't claim to have "solved" climate change, but it does offer a sharper, more accurate tool for understanding how the weather is currently affecting the lights in the Philippines.

Technical Summary: Mixed Frequency Stochastic Frontier Model with Application to the Linkage of Weather Extremes and Firm Efficiency

Problem Statement
This paper addresses two persistent methodological challenges in Stochastic Frontier Analysis (SFA) when applied to panel data involving environmental shocks:

The "Wrong Skewness" and Convergence Problem: Conventional SFA models often assume a half-normal or truncated normal distribution for the inefficiency term. When the data exhibits positive skewness (a common occurrence in regulated industries or when inefficient firms are barred from entry), the maximum likelihood estimation (MLE) frequently fails to converge or produces singular information matrices, leading to biased efficiency estimates (often converging to 1).
The Mixed-Frequency Data Problem: Firm-level production indicators (inputs and outputs) are typically observed at low frequencies (e.g., annually), while determinants of inefficiency, such as extreme weather events (typhoons, rainfall), exhibit high-frequency variability (monthly or daily). Standard approaches aggregate high-frequency data to match the low-frequency production data, which the authors argue leads to significant information loss. Aggregating extreme weather events (e.g., averaging monthly wind speeds) neutralizes the impact of rare but destructive events, masking their true effect on operational inefficiency.

Methodology
The authors propose a novel Mixed-Frequency Stochastic Frontier Model (VF-SFM) that integrates high-frequency covariates directly into the inefficiency equation without temporal aggregation.

Model Specification:
- Production Function: A standard Cobb-Douglas production function is specified in logarithmic form, linking output (Gross Operating Revenue) to labor and capital inputs. The model includes an autoregressive term to capture temporal dependencies in production.
- Inefficiency Equation: Instead of assuming a specific distribution for the inefficiency term ( $u_{it}$ ), the authors specify inefficiency through a logistic function. This ensures non-negativity of the inefficiency term without restrictive distributional assumptions.
- Nonparametric Integration: High-frequency weather variables (maximum monthly rainfall and maximum sustained wind speed) are incorporated into the logistic function via flexible nonparametric additive functions (using splines). This allows the model to capture heterogeneous and time-varying influences of extreme weather on inefficiency, accommodating data sparsity (e.g., zero values during dry months).
- Equation Structure:
  $\ln y_{it} = \alpha_0 + \alpha_1 \ln L_{it} + \sum \beta_j \ln C_{jit} + v_{it} - u_{it}$
  $u_{it} = \frac{1}{1 + \exp[-(\gamma Z_{it} + \sum f_j(R_{tj}) + \sum g_j(W_{tj}))]} + \epsilon_{it}$
  Where $Z$ represents firm-specific determinants (collection efficiency, system loss, interest expenses), and $R$ and $W$ represent monthly rainfall and wind speed, respectively.
Estimation Algorithm:
To avoid the convergence issues associated with MLE, the paper employs a hybrid backfitting algorithm. This iterative procedure combines:
1. Generalized Least Squares (GLS) for the production function parameters.
2. Ordinary Least Squares (OLS) for the parametric coefficients in the inefficiency equation.
3. Penalized Least Squares (PLS) for estimating the nonparametric functions (splines).
  The algorithm iterates between estimating the production function and the inefficiency equation until the Mean Absolute Percentage Error (MAPE) converges.

Empirical Application and Results
The methodology is applied to a panel of 114 electric cooperatives (ECs) in the Philippines from 2010 to 2022. The study links firm efficiency to extreme weather events (typhoons and rainfall) while controlling for financial and operational indicators.

Convergence and Predictive Ability:
- The proposed VF-SFM converged successfully for all regions.
- In contrast, the benchmark model (Battese & Coelli, 1995) using aggregated annual weather data and MLE failed to converge in most regions. Where it did converge, the VF-SFM demonstrated superior predictive accuracy, with significantly lower MAPE values compared to the benchmark.
Efficiency Estimates:
- The benchmark model produced inflated Technical Efficiency (TE) estimates (often near 100%), a symptom of the "wrong skewness" problem and model misspecification.
- The VF-SFM yielded more realistic TE estimates (ranging roughly from 30% to 80%), revealing that firms are not operating at the frontier.
- Weather Impact: The nonparametric functions successfully captured the impact of extreme weather. Regions with low rainfall exposure showed higher efficiency. The model identified that years with severe typhoons (e.g., 2018, 2019) correlated with increased system losses and reduced efficiency, effects that were obscured in the aggregated benchmark model.
Determinants: System loss and collection efficiency were found to be significant determinants of inefficiency, alongside the nonparametric weather terms.

Key Contributions

Methodological Innovation: The paper introduces a framework that resolves the "wrong skewness" issue by replacing distributional assumptions with a logistic function for inefficiency, thereby ensuring model convergence and robustness.
Mixed-Frequency Handling: It demonstrates a practical solution for integrating high-frequency environmental data into low-frequency production models using nonparametric additive functions, preserving the variability of extreme events that aggregation would otherwise erase.
Estimation Robustness: The hybrid backfitting algorithm provides a stable alternative to MLE, effectively handling the complexity of mixed-frequency data and nonparametric components without the convergence failures typical of conventional SFA.

Significance and Claims
The paper claims that its proposed framework broadens the applicability of stochastic frontier analysis to contexts characterized by mixed-frequency data and complex environmental determinants. By successfully linking climate-related shocks to operational inefficiency at the firm level, the study provides a practical tool for evaluating micro-level sustainability and resilience challenges.

The authors emphasize that while extreme weather cannot be controlled, the model allows regulators and firm managers to identify the specific operational vulnerabilities (e.g., system losses) exacerbated by these events. This supports the development of adaptive strategies, such as investing in resilient infrastructure and improving operational technologies, to mitigate efficiency losses in climate-vulnerable sectors like electricity distribution. The study concludes that the VF-SFM offers a more accurate and reliable approach for policy analysis in sectors where environmental shocks are frequent and severe.

The Story of the Storm and the Spark

What They Found

More like this