تجزیه و تحلیل مرزی تصادفی پیشرفت فنی، تغییرات بهره وری و رشد بهره وری در صنعت چوب و الوار سازی کشور های شمال غربی اقیانوس آرام

Stochastic frontier analysis was employed to investigate technical efficiency and productivity growth in the sawmilling industry of the U.S. Pacific Northwest over the period 1968–2002. The results of our analysis indicate that productivity growth was strong over the 30-year study period, due almost exclusively to technical progress. The model developed in this analysis was used to examine the cause of employment declines in the sawmilling industry between 1988 and 1994. We found that that 62% of the decline was due to changes in output and non-labor input factors and 38% was due to technical change alone.

Beginning in the late 1980s and continuing into the 21st century, the sawmill industry of Oregon and Washington (hereinafter Northwest) underwent substantial reductions in employment and is believed to have experienced significant technical change as well, resulting in productivity growth (Helvoigt et al., 2003). Much of the information regarding technical change and productivity growth has been anecdotal, however, and to the best of the authors' knowledge has not been empirically validated. In this paper, we employ a stochastic frontier production function (SFPF) to estimate technical change1, efficiency change2, and productivity growth3 for the Northwest sawmill industry. We also compute output elasticities, scale efficiency, and Morishima elasticities of substitution between input factors. Although its relative importance in the Oregon and Washington economies has slowly declined over the past decades, the forest product industry in the Northwest remains a major source of employment and economic output. Because of its historic and continued significance to the economy and culture of the Northwest and its special role in some communities, it is important to understand how the structure of the sawmilling industry has changed. The results of this analysis provide some insights. Our approach departs from past studies of the sawmill industry in the Northwest and other North American regions in several ways. First, the SFPF allows the direct estimation of technical efficiency4 at a point in time, as well as technical and efficiency change through time. This allows comparison of the aggregate position of lumber producers in one region to the best practices frontier of lumber producers in the entire Northwest. Second, most past studies have assumed lumber producers to be successful cost minimizers or profit maximizers and have employed cost or profit functions (see Stier and Bengston, 1992 for a review of studies). Implicit in these approaches is the assumption that any deviation from minimum cost or maximum profit is random noise. In contrast, the SFPF estimates both the frontier of the industry production function and measures the technical efficiency of each producer relative to the frontier. It does not require the assumption that producers are acting in an economically optimal fashion. Third, unlike past studies that have employed CES production functions (Greeber and White, 1982 and Stier, 1982), this study employs the flexible translog production function and gives explicit attention to regularity (curvature) conditions for the estimated function. SFPF estimates the industry's technical frontier based on the performance of the most efficient (i.e., most productive) production units (Kumbhakar and Lovell, 2000). Measures of the technical efficiency of each unit are then estimated based on the distance of the unit from the frontier. Estimates of productivity and/or technical efficiency derived from SFA are therefore relative, not absolute, measures.

Analysis of more than three decades of production data for the Pacific Northwest sawmilling industry reveals that lumber producers in Oregon and Washington experienced substantial productivity gains, averaging 2% per year over the 1968–2002 sample. Most of these gains were due to technical change (i.e., expansion of the industry's production frontier). The statistical evidence indicates that scale efficiency played almost no role in productivity growth, while efficiency change had a small, but negative effect. In the latter instance, even as the industry's production frontier expanded over time, the distance of the average unit from the industry's production frontier increased. Mills compete not only with other lumber producers within their same region but with mills throughout the Northwest as well. As technical improvement allowed the most innovative mills to push out the industry's production frontier, other Northwest mills were forced to adopt productivity-enhancing technology in order to keep up with the expanding frontier, or shut down. A portion of the productivity growth observed over the study period may be due to the closure of the least technically efficient firms, resulting in efficiency change (and productivity growth) through attrition (Stevens, 1991). In Washington alone, mill survey data indicate that there were more than 200 sawmills in 1968, but only 75 mills in 2002. Yet, over this same period, aggregate mill capacity in Washington actually increased. The net result is that average mill capacity tripled between 1968 and 2002. Estimates of technical efficiency are high for all regions for all time points (> 0.89). These high scores may be due in part to the small number of units in our study (hence most may be on the frontier) and also to rapid transfers of technology and innovation between sawmills in the Northwest, yielding little variation in technical efficiency between sawmilling regions. The results of this analysis indicate that over the 34-year study period, technical change was labor-saving, capital-using, and neutral with respect to sawlog use. Neutral technical change in sawlog use may seem counterintuitive given the perception of widespread adoption of sawing technologies that increase lumber yields per unit of sawlog input (Stier and Bengston, 1992). Neutral technical change does not, of course, mean a constant marginal product for sawlogs, only that the marginal product of sawlogs must vary over time in certain relations with the marginal products of other inputs. Results from earlier studies are mixed on this issue. Abt (1987) reports wood-using bias for Pacific Northwest sawmills over the period 1963–1978. Martinello (1987) found that technical change was neutral with respect to sawlog usage for coastal British Columbia (BC) but was wood-saving for interior BC. Analyzing data from 1957–1981 for the U.S. Pacific Northwest and BC, Constantino and Haley (1988) found that technical change was both labor- and wood-saving. Meil and Nautiyal (1988) found that with few exceptions, technical change was labor-saving and sawlog-using for the major lumber producing regions of Canada. The lack of consensus in the literature on the existence and direction of technical bias with respect to sawlog usage may be due to the spatial and temporal heterogeneity of sawlogs. That is, sawlog size and quality vary over time and region. Constantino and Haley (1988) use a quality index in their study to control for temporal variation. In the Northwest, average sawlog diameter decreased through the 1970s, 1980s, and 1990s, and it is likely that sawlog quality declined over this period as well (see, Størdal and Adams, 2005, for discussion). The trend toward smaller sawlogs has led to increased productivity because the milling of small logs is less labor intensive and more amenable to mechanization. Consistent with decreases in the size and quality of sawlogs over the past few decades has been a decrease in lumber quality (Størdal and Adams, 2005). This is not captured in the present analysis. The average quality of labor has likely also changed, as many unskilled workers have been displaced due to increased mechanization (Stevens, 1991). The SFPF model developed in this analysis is useful in examining the impact of technology on changes in input and output levels observed over the study period. Of particular interest is the period beginning in the late 1980s and extending through the mid-1990s when harvest levels on Northwest National Forests experienced substantial declines, many Northwest sawmills shutdown, and sawmill employment fell. Between 1988 and 1994, SIC 242 employment in Oregon and Washington declined from 39,944 to 30,538 and lumber production declined from 12.4 billion board feet (BBF) to 8.2 BBF. Over this same period, productivity grew by approximately 2.0% per year. Thus, even as lumber production declined, the Northwest sawmill industry was utilizing relatively fewer inputs to produce each unit of output. A portion of the employment decline over this period was due to the adjustment of output and non-labor inputs and a portion was due to technical improvement. We decompose the aggregate change by tracing the employment effects of three partial shifts on (and in) the production surface: i) the adjustment of output and log input from 1988 to 1994 levels under 1988 technology, ii) the change in capital stock and other inputs to 1994 levels under 1988 technology, and iii) the change in technology from 1988 to 1994 with all non-labor inputs at 1994 levels. Of the total sawmilling job loss over the period (9406 jobs), we find that 62% (5856 jobs) was associated with changes in output and non-labor factors and 38% (3550 jobs) to technical change alone. This study is the first to utilize a stochastic frontier production function to examine productivity growth, technical change, and other economic measures of the sawmilling industry. SFA methods allow for the relaxation of the assumptions that lumber producers are successful cost minimizers and/or profit maximizers. Instead of estimating the average production function and assuming that deviations from this function (either positive or negative) are random, SFA estimates the production frontier and provides estimates of each unit's inefficiency, relative to the estimated frontier. This study is also the first in more than a decade to examine the technical structure of the Oregon and Washington lumber producing industry, and to examine changes in that structure. Despite substantial declines in harvest on Northwest National Forests during the 1990s, Oregon is still the largest lumber producing state in the U.S., Washington produces more lumber today than at any other time in the post-WWII period, and some Northwest lumber mills are among the lowest cost producers in North America. Strong productivity growth over the last three decades has helped Northwest lumber producers remain competitive in an ever-increasing global marketplace.