ارزیابی چندبعدی کارایی واحدهای صنعتی پرورش مرغ گوشتی در ایران: رویکرد ترکیبی تحلیل پوششی داده‌ها و تحلیل مرزی تصادفی

Introduction
Chicken meat plays a strategic role in food security and animal-protein supply because of its relatively affordable price, short production cycle, and broad accessibility to consumers. In Iran, industrial broiler production constitutes the dominant source of chicken meat and depends heavily on continuously supplied variable inputs, particularly day-old chicks, feed, and labor. The economic performance of broiler farms therefore depends not only on production capacity but also on how efficiently these inputs are transformed into output, whether farms operate at an appropriate scale, and whether the input combination is consistent with relative input prices. A farm may operate close to its optimal scale while still using excessive quantities of inputs, or it may be technically efficient but incur unnecessarily high costs because its input mix deviates from the cost-minimizing combination. Consequently, evaluating farm performance solely through a single measure of technical efficiency may conceal economically important sources of inefficiency.
Existing evidence on broiler-farm efficiency, particularly in Iran, has largely concentrated on individual provinces, single time periods, or specific dimensions of performance. Technical efficiency has received considerably greater attention than scale, allocative, and cost efficiency, while reliance on a single frontier approach may also make efficiency estimates sensitive to the assumptions used to construct the production frontier. Against this background, the present study develops a multidimensional assessment of industrial broiler farms in Iran by decomposing the performance gap into technical, scale, allocative, and cost components. Data Envelopment Analysis (DEA) is employed as the main analytical framework, while Stochastic Frontier Analysis (SFA) is used as a complementary approach to examine the robustness of relative technical-efficiency patterns under an alternative frontier specification. The study further compares provincial efficiency structures across two independent cross-sectional periods to identify how the composition and geographical pattern of inefficiency differ across production environments.
Materials and Methods
The study is applied in purpose and quantitative, non-experimental, and comparative in design. Farm-level microdata from the 2015 and 2020 national surveys of industrial broiler farms were used. After data validation and exclusion of incomplete or invalid observations, 1,896 farms in 2015 and 2,412 farms in 2020, totaling 4,308 observations, were retained for DEA. Broiler meat production was specified as the output, while day-old chicks, feed, and labor were treated as the principal variable inputs. Their corresponding prices were incorporated into the cost-minimization model; therefore, the resulting allocative and cost-efficiency measures refer only to these three inputs.
Input-oriented DEA models were estimated separately for each year. Overall technical efficiency was obtained under constant returns to scale, pure technical efficiency under variable returns to scale, and scale efficiency from their decomposition. A variable-returns-to-scale cost-minimization model was then used to estimate allocative efficiency, cost efficiency, and potential cost saving.
SFA was employed as a complementary assessment of technical efficiency. Because this approach does not require input-price information, the SFA sample consisted of 2,072 farms in 2015 and 2,463 farms in 2020. Separate translog stochastic production frontiers were estimated for the two periods, incorporating flock scale, feed, labor and housing-area intensities, and climatic controls. Automation status and the centered number of production cycles were included in the inefficiency-variance specification. Provincial mean scores and rank correlations were finally used to assess concordance between DEA and SFA.
Results and Discussion
The DEA results indicated substantial scope for improving input use in Iranian industrial broiler farms. Mean overall technical efficiency increased from 0.527 in 2015 to 0.560 in 2020, while pure technical efficiency rose from 0.571 to 0.604. This implies potential proportional input reductions of about 42.9% and 39.6%, respectively, without reducing output. Scale efficiency remained relatively high and stable, decreasing only from 0.932 to 0.929, indicating that the main performance gap was associated more with inefficient input transformation than with inappropriate farm size.
Allocative efficiency improved from 0.862 to 0.887, whereas cost efficiency increased from 0.484 to 0.533. Accordingly, potential cost saving for the three principal variable inputs declined from 51.59% to 46.69% but remained substantial. The combination of relatively high scale and allocative efficiency with lower pure technical and cost efficiency suggests that capacity expansion alone is unlikely to resolve the main efficiency gap.
Considerable provincial heterogeneity was also observed. Mazandaran recorded the highest pure technical efficiency in 2015, while Golestan performed best in allocative and cost efficiency. In 2020, Golestan ranked highest in pure technical and cost efficiency, whereas Fars showed a comparatively balanced efficiency structure. Kerman, Fars, and Razavi Khorasan exhibited notable improvements across several dimensions.
SFA provided complementary evidence. Mean technical efficiency was 0.853 in 2015 and 0.834 in 2020. Although absolute efficiency levels and inter-period changes differed from DEA, provincial rankings showed stronger concordance in 2020. The Spearman correlation between SFA and DEA under variable returns to scale increased from 0.422 to 0.701, while the corresponding correlations under constant returns to scale were 0.591 and 0.765. Thus, provincial efficiency differences were relatively robust across frontier approaches, particularly in 2020, although the magnitude and temporal pattern of efficiency remained sensitive to model specification.
Conclusions
The findings indicate that inefficiency in Iranian industrial broiler production is multidimensional, but its dominant component is associated with the physical transformation of inputs into output rather than inappropriate production scale. Although allocative and cost efficiency improved between the two cross-sectional periods, substantial potential remained for reducing the cost of day-old chicks, feed, and labor without lowering production. Efficiency-enhancing policies should therefore move beyond uniform capacity expansion and place greater emphasis on improving feed and chick quality and supply stability, strengthening technical and extension services, promoting more efficient input combinations, and facilitating the diffusion of production practices observed among benchmark farms. Provincial differences further imply that interventions should be tailored to the specific source of inefficiency rather than applied uniformly across regions. Finally, because the cost and allocative measures cover only three major variable inputs and the two datasets represent independent cross-sectional samples, the estimated cost-saving potential should not be generalized to total production costs or interpreted as changes in the performance of the same farms over time.