نوع مقاله : پژوهشی اصیل
عنوان مقاله English
نویسندگان English
Abstract
This paper identifies and disentangles the determinants of persistent and transient inefficiency in Iranian thermal power plants. To obtain reliable efficiency estimates, a four-component Stochastic Frontier Analysis (SFA) model with unbalanced panel data is employed, enabling the separation of persistent (time-invariant or long-term) inefficiency from transient (time-varying or short-term) inefficiency. This distinction facilitates more targeted policy recommendations aimed at reducing inefficiency, enhancing productivity, and lowering electricity generation costs. Steam, gas, and combined-cycle power plants operating between 2012 and 2022 are examined. The results indicate that, except for publicly owned steam power plants, most thermal power plants exhibit persistent efficiency levels exceeding 80%. The most influential determinant of transient inefficiency is the plant’s operation during peak load periods. Efficiency improved steadily until 2020; however, inefficiency increased in 2021 and 2022.
Purpose/Aims:
Since its introduction by Aigner et al. (1977), SFA has become one of the principal methods for measuring efficiency. In stochastic frontier models, the efficiency of production (or cost or profit performance) is assessed relative to an estimated production frontier. However, approaches to frontier estimation and inefficiency specification differ across model formulations. A defining feature of SFA models is the decomposition of the error term into two components: a symmetric statistical noise component and a non-negative inefficiency component. The statistical noise term captures random shocks and measurement error, whereas the non-negative component reflects technical inefficiency, defined as the shortfall of observed output from potential (frontier) output.
Methodology & Framework:
This study distinguishes between persistent and transient inefficiency determinants in Iranian thermal power plants by employing a four-component SFA model with unbalanced panel data. Specifically, we apply a third-generation SFA framework introduced by Colombi et al. (2014) and estimate a translog production function using R software.
In this specification, the composite error term is divided into four components: plant-specific heterogeneity, random noise, persistent inefficiency, and transient inefficiency . Estimation is conducted using a single-stage maximum likelihood approach, which simultaneously estimates the production function parameters and the variance components associated with inefficiency determinants. This framework enables a clear separation between persistent (time-invariant or long-term) and transient (time-varying or short-term) inefficiency effects. Such differentiation allows for more precise policy recommendations aimed at improving productivity and reducing electricity generation costs.
The empirical analysis covers steam, gas, and combined-cycle power plants over the period 2012–2022.
Findings:
The results indicate that fuel is the most influential input in electricity production, with an estimated elasticity of 0.925; thus, a 1% increase in fuel input increases output by approximately 0.925%. Capital is identified as the second most important production factor, with an elasticity of 0.164, whereas labor is statistically insignificant. The insignificance of labor likely reflects the high degree of mechanization in thermal power generation and the relatively limited role of labor in the production process.
The estimated scale elasticity suggests slightly decreasing returns to scale (DRS) (–0.058). Output increases modestly over time (0.004), and although this growth accelerates slightly (0.002), the overall magnitude remains limited.
Regarding transient inefficiency, the most important determinant is plant operation during peak load periods. A 1% increase in the peak load ratio (PLR) reduces the variance of transient inefficiency by 4.38%. Pollution intensity has a positive and statistically significant coefficient, while plant age has a negative and marginally significant coefficient, suggesting that older plants may exhibit lower transient inefficiency variance. In contrast, the producer price index and ambient temperature are statistically insignificant.
In the persistent inefficiency component, transitioning toward gas (–1.611) and combined-cycle (–9.449) technologies, as well as private ownership (–1.306), significantly reduces inefficiency variance.
Discussion:
With the exception of publicly owned steam power plants, most thermal power plants demonstrate persistent efficiency levels exceeding 80%. Given this relatively high persistent efficiency, variations in total inefficiency are largely driven by differences in transient inefficiency. Plants exhibiting higher transient inefficiency consequently experience greater overall inefficiency.
Among technologies, gas power plants display the lowest transient efficiency. Because transient efficiency is primarily influenced by the ratio of peak load to nominal capacity, the relatively higher transient efficiency of steam power plants can be attributed to operational constraints. Steam units require longer ramp-up times to return to maximum capacity; therefore, system operators tend to keep them online. In contrast, gas plants are more frequently taken offline during periods of lower demand, contributing to lower transient efficiency.
Efficiency trends indicate marked improvement during most of the study period. Between 2012 and 2020, total efficiency increased from 50% to 66%, while transient efficiency rose from 67% to 77%, corresponding to growth rates of 32% and 14.9%, respectively. However, both indicators declined in 2021 and 2022. Although the overall trajectory remains positive, the average total efficiency remains relatively low.
With respect to scale properties, steam power plants operate under nearly constant returns to scale (CRS), suggesting proximity to their optimal production scale given current technology. By contrast, gas power plants predominantly exhibit increasing returns to scale (IRS), indicating unrealized economies of scale. This pattern is particularly pronounced among privately owned gas plants, where scale efficiency improves over time.
Conclusion & Implications:
Despite substantial efficiency gains during most of the study period, the Iranian thermal power sector continues to operate below its potential. With an average total efficiency of approximately 60%, electricity generation could theoretically increase by around 66% using the same level of inputs if all plants operated on the production frontier.
The analysis of scale properties using the Delta method highlights important structural differences across technologies. Overall, 13% of plants operate under DRS, with no gas plants in this category. Approximately 25% of steam and combined-cycle plants exhibit DRS. Combined-cycle plants show the highest prevalence of CRS, while gas plants display the strongest IRS. These findings underscore the importance of technology-specific policy design. In particular, expanding the operational scale of gas power plants may play a pivotal role in enhancing overall efficiency in the electricity sector.
کلیدواژهها English
