Abstract

Data sources

We extracted the data for the logging industry (Code 1133, North American Industrial Classification System; Instituto Nacional de Estadística y Geografía (INEGI) of Mexico, Statistics Canada, and the United States Office of Management and Budget 2017) from quarterly workforce indicators (US Census Bureau 2021), timber product output (TPO) reports (USDA Forest Service 2021), wood supply chain analysis (Barynin et al. 2013), and TimberMart-South (Norris Foundation 2021b) to construct the time-series dataset used in this study (Table 1). Due to the limited availability of data, we work on a smaller dataset from 11 states (Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia) for 11 years, from 2007 to 2017. There were a few missing data in the TPO dataset in 2008, 2010, 2012, 2014, and 2016. As a result, we applied the average interpolation method for the missing data between two data points. The average interpolation method is one of the simplest and most applicable methods for imputing the missing economic data without changing the overall variable average (Saeipourdizaj et al. 2021). Table 1 shows the list of variables used in this study. To reduce the skewness of the raw data, the Emp, W, and Q were transformed into their respective logarithmic forms.

Table 1. List of variables.

View Fullscreen

Empirical approach

The following steps were used to investigate the causal relationships between employment, wage, mechanization, product prices, and production level in the logging industry. First, a multivariate time-series model, the vector autoregression (VAR) model, was applied. Next, a graphical modeling analysis, the directed acyclic graph (DAG) approach, was used to capture the dynamic relationships between variables and determine their contemporaneous causal relationships. Finally, a structural analysis, forecast error variance decomposition (VD), was conducted to identify how the shocks change over time.

Directed acyclic graph results

After testing the stationarity and cointegration of each panel series, we found that all variables are not stationary at their level but integrated under order one and have a long-run stable equilibrium relationship. Hence, we carried out a VAR model and got the residual correlation coefficient matrix of the VAR model. Following that, we applied the PC algorithm in Tetrad 6.8 to analyze the residual correlation coefficient matrix to obtain the DAG. The DAG, in turn, was used to disclose the contemporaneous causal structure.

In Figure 2, the complete undirected graph shows the undirected path connecting each variable with every other variable. It reveals that these five economic variables were interrelated with each other.

View Figure

• Download as Powerpoint
• Download Figure

Figure 3 presents the DAG on these variables at the 20 percent significance level, which allows us to more accurately identify a contemporaneous causal relationship in a small sample (Spirtes et al. 2000). It shows the causal relationship paths among employment, production level, wage, capital, and product price, indicating that these five factors are interconnected in contemporaneous time. There are seven paths in the graph: P → W, P → Emp, W → Emp, Emp → K, Emp → Q, W → K, and K → Q.

View Figure

• Download as Powerpoint
• Download Figure

The contemporaneous causal relationship between employment and its influencing factors can be refined into the following two paths: P → Emp and P → W → Emp.

• P → Emp: The results show that there is no causal link toward product price. In other words, the product price is not driven by any other factors, suggesting the price is exogenous in contemporaneous time. The product price can directly affect employment.

• P → W → Emp: The DAG algorithm shows that product price causes changes in wage, and wage directly impacts employment, indicating that the product price has an impact on employment directly and indirectly in the immediate short term.

Figure 3 also shows the contemporaneous relationship between employment, wage, capital, and production level, which is discussed here further:

• Emp → K: Employment affects capital in contemporaneous time.

• Emp → Q: Employment also has an influence on the production level.

• W → K: The wage has a direct effect on capital in the immediate short term. Both wage and employment are the contemporaneous causes of capital, and the price affects capital indirectly through employment and wage.

• K → Q: The production level is found to be affected by capital.

Forecast error variance decomposition results

After analyzing the contemporaneous causal relationship via DAG, we then investigated how much of the change in employment across time is caused by endogenous shocks by itself and how much is caused by exogenous shocks as well as other variables. The VD was used to investigate this dynamic relationship among variables (Table 2).

Table 2 reports the decomposition at horizons 1, 5, and 10 years. It shows that employment in the logging industry is most prominently explained by itself in the first year at 98.2 percent. The influence from employment itself is then reduced to 56.8 percent at the 5-year horizon and 9.0 percent at the 10-year horizon. Product prices and wages come in the second and third positions, respectively. The influence of the product price on employment gradually increases over time, and the percentage contribution rises from 0.5 percent at the 1-year horizon to 28.0 percent at the 5-year horizon and 64.2 percent at the 10-year horizon. Capital has an increasing influence on employment over time, and the percentage contribution increases from 0 percent at the 1-year horizon to 20.1 percent at the 10-year horizon. The wage and production level have a slight influence on employment over time.

The forecast error variance for the wage is most prominently explained by the endogenous shock of wage itself (68.7%), followed by the shock of the product price (31.3%) at the 1-year horizon. The shocks from employment, capital, and production are small compared to that of product price and wage itself, which are consistent with those of the DAG. In addition, the forecast error variance for capital is most prominently explained by itself and product price. Production level is most prominently explained by itself and product price and capital.

Directed acyclic graph analysis

P → Emp and P → W → Emp.—The product price is an important signal for the logging firms. As price takers, when the logging product price decreases, the logging firms have to reduce their costs. Faced with falling prices in a highly competitive and homogeneous industry like logging, the firms are more likely to reduce costs by reducing labor costs instead of nonlabor costs (Bertola et al. 2012). Therefore, laying off employees and reducing the wage would be the preferred adjustment strategy.

The logging product price is highly dependent on the business cycle (He et al. 2022). With the decrease in new housing units started during the economic downturn, as well as the decreasing demand for house renovation and furniture, the logging product price decreased (Xiang and Yin 2006, Drapala 2009). For example, the economic prosperity fostered more than two million housing starts in 2005 and then promoted the logging product price in 2005 (Keegan et al. 2011). However, the price of logging products decreased dramatically during the recession in 2008 because of the sharp decline in demand for wood-frame housing and a systematic reduction in the pulp and paper and furniture industries (Grushecky et al. 2006, Abbas et al. 2014), whose production fell by 75 percent between 2005 and 2009 (Hodges et al. 2011).

Thus, the logging firms lay off workers to minimize operating costs in response to falling prices. The reason why logging firms can lay off workers easily may be that logging workers do not have long-term employment contracts, or they are not the permanent employees. Seasonal operation is one of the important reasons for short-term contracts or temporary employment. For example, logging firms in the South are partially shut down in the third quarter while operating for the rest of the year (Bureau of Labor Statistics 2022). The North has always faced seasonal challenges because of spring thaw (Conrad et al. 2018). Paid logging workers in hourly, daily, or weekly and hiring undocumented foreign workers are also causes for the lack of long-term employment contracts (Bertola et al. 2012, He et al. 2021). Another reason why logging firms can lay off workers easily may be that the logging industry was not unionized (Neumann 2019). Thus, the logging workers have had no power to negotiate with the employers. Further, there was a closure wave among the mills and logging firms, even though some were not permanent closures, due to the severe market conditions (Keegan et al. 2011, Wilson 2017).

As a result, when logging product prices decreased during the recession in 2008, logging employment in 2010 fell by 22.3 percent from 2005 (US Census Bureau 2021). When the economy is prosperous, the increasing demand for logging production stimulates the price (Yin 2001). The logging firms would recruit more employees and in turn expand production.

Logging product price also has an impact on the wages of logging workers. According to classical economics, facing lower prices, firms would decrease the nominal wages, resulting in lower real wages, increasing or, at least, maintaining some profits (Bertola et al. 2012). For example, between 2007 and 2009, the real price of logging products among Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia fell by 16.9 percent, while the real wages dropped by 3.5 percent (Norris Foundation 2021a, US Census Bureau 2021), and the real profits per firm in some major forest states in 2009 were estimated to be 20.3 percent lower than those in 2007 (He et al. 2021). The reason why the wage decreased less than that of the price might be the wage rigidity. Logging firms are more likely to reduce employment than lowering wages because firms are constrained by the contracts signed at higher bargaining levels (Kahn 1997, Lebow et al. 2003). Thus, the logging firms would react to logging product price shocks by reducing employment first, especially temporary employment, then by cutting pay (Bertola et al. 2012).

In addition, wages can provide market signals for logging firms and workers. During economic recovery and expansion, the logging product price would increase, and the logging firms would raise wages to recruit more employees. The logging firms may offer more decent wages to provide incentives for the existing employees to work longer hours and to match the higher production level. Some firms may offer higher wages to hire trained employees to increase productivity. In the context of expected experience, the wage increase by the proficiency level of the workers is reasonable (Xu et al. 2014). The higher wages also signal opportunities and career prospects, attracting additional employees. When the American economy was recovering in 2010, the real price increased by 2.7 percent, while the real wage and employment increased by 4.9 and 1.2 percent, respectively (Norris Foundation 2021a, US Census Bureau 2021).

Emp → Q and Emp → K.—In response to price changes, logging firms are more likely to adjust production levels by adjusting employment. For example, during the recession in 2008, due to the severe market conditions, logging firms laid off workers to minimize the operating costs and then reduce production. In most major logging states, such as most southern states, logging employment has declined to various degrees (Fig. 1), thereby lowering the amount of timber harvested (USDA Forest Service 2021).

Employment also has a direct effect on capital. On the one hand, reduced employment can lead to an increase in capital. The US logging industry faces some severe situations, such as an aging workforce and declining recruitment of the upcoming generations (He et al. 2021). As more and more employees retire, recruiting new employees is challenging for low profit margins and is full of uncertainty, instability, and seasonal operations (Egan and Taggart 2009, Shivan et al. 2020). Logging is physically demanding and a hazardous civilian occupation in the United States because logging workers must spend all their time outdoors, sometimes in poor weather, and often in isolated areas (Scott et al. 2020, Bureau of Labor Statistics 2020b). However, logging firms cannot offer attractive salaries to attract new employees (He et al. 2021), especially younger workers to replace retirees (Baker and Greene 2008). Thus, the logging firms have to replace jobs by substituting mechanized harvesting systems for workers, which means that the decrease in employment has an impact on the increasing capital.

On the other hand, increased employment can result in a decrease in capital. The logging firms, especially small-scale firms, tend to employ more staff to minimize costs instead of investing in mechanized systems. Small-scale logging firms have been remarkably tenacious. They are developed with a long history and are widespread in the United States (Conrad et al. 2018). One of the main reasons for the presence of many small logging firms is parcelization (Yin et al. 1998, Milauskas and Wang 2006). The number of forestland owners has increased rapidly in the last decade, resulting in a decline in the average size of forest ownerships, most of which are nonindustrial private forest (NIPF) owners (Rickenbach and Steele 2006). However, the large logging firms with mechanized harvesting systems may not match this small-scale forestland as well as the low logging volume (Greene et al. 1998). In contrast, small logging firms have advantages in harvesting on small-scale forestland (Blinn et al. 2015). Unlike large firms, small logging firms do not own mechanized equipment and employ more staff to minimize costs (Shivan et al. 2020). Compared with large firms, small-scale firms are more inclined to operate seasonally and reduce capital expenses to maintain efficiency. They are also less likely to afford the cost of equipment repair and maintenance (LeBel and Stuart 1998). As a result, small logging firms are “hand crews,” hiring employees to harvest timber by chain saw instead of mechanized harvesting systems (Egan 2011).

W → K → Q.—Wages have a direct effect on capital, too. Obviously, high wages can provide incentives for logging firms to replace labor with machines. However, they also possibly have made the replacement harder and slower. It was estimated that initial investments in mechanized harvesting systems were around $0.4 to $1.5 million, and monthly finance payments were around $5,000 to $20,000 (Rickenbach and Steele 2005). Logging firms must remain profitable to remain in business and continue investing in their businesses, but the high wage reduces their profitability because the most considerable contribution to costs is wages (Jacobson et al. 2009). In the South, the wages account for an average of about 30 percent of the total costs (Xu et al. 2014). As a result, confronting high wages, logging firms, especially small logging firms, cannot afford the wage costs of large teams or investment in mechanized systems, instead, they would outsource some production processes instead of investing in mechanized systems, which can cope with the low production level (Stuart et al. 2010). Logging firms tend to contract labor-intensive activities because they can alleviate their workload and keep a small staff crew (Wang 1999), thus saving of salary costs. For example, trucking is a significant cost for logging firms (Yin and Caulfield 2002), and most small firms cannot afford substantial investments and expenses of trucking, so they would choose to contract out trucking, which can be much cheaper than operating their fleets, and focus on the harvesting business (Hamsley et al. 2007, Shivan et al. 2020).

In addition, logging firms can benefit from mechanization by using more equipment and technology to promote production levels (Mac Donagh et al. 2017). In the past two decades, logging firms have promoted the mechanization of harvesting systems. The proportion of loggers using the capital-intensive mechanized harvesting systems has increased over time, making logging a much more capital-intensive industry (Kollberg 2005).

The mechanization process enables logging firms to attain higher productivity. The productivity of logging firms in most states has increased since 1997 because of the widespread use of mechanized harvesting systems (He et al. 2021). The mechanized firms in Wisconsin produced an average of 0.73 million cubic feet yearly, while the nonmechanized firms produced only 0.23 million cubic feet yearly (Rickenbach and Steele 2005). Previous research shows that due to significant capital investments, production increased from 3.4 to 5.5 tons per person-hour between 1987 and 2012 in the South (Greene et al. 2013). In particular, the weekly production level per firm in Georgia has increased by 83 percent since 1987 because of mechanization (Baker and Greene 2008). Therefore, the mechanized harvesting systems contributed to increasing production level (Cubbage and Carter 1994).

Forecast error variance decomposition analysis

As shown in Table 2, the change in employment in the logging industry is most prominently explained by the endogenous shock of employment itself in the first five years. One reason for that might be the employment stickiness. The logging firms would consider the cost of employment adjustment and the indivisibility of employment when they tried to adjust the employment in the long run (Vermeulen 2006).

Unlike in the contemporaneous time, on one hand, the VD shows that the capital has an increasing influence on employment in the long run. The logging industry is more and more capital intense. In the sample states, the capital increased from $16.5/green short ton in 2007 to $21.3/green short ton in 2017. A drastic reduction in employment occurred between 2007 and 2009, from 30.3 thousand to 26.3 thousand, and employment was 26.4 thousand in 2017, not recovering to historical levels in 2007. Further, the production level was 8.1 million MCF in 2007, then dropped to 6.7 million MCF in 2009, and recovered to 8.2 million MCF in 2017. According to these data, the production level has recovered with the capital increase; however, the employment still kept at a low level. Thus, this indicates that capital substitutes for labor in the long run. The productivity of the mechanized harvesting systems is much higher than the manual systems, and thus the logging industry now has far more capacity than demand. The capacity utilization of logging firms with mechanized harvesting systems was only 70 to 84 percent on average (Conrad et al. 2017, 2018). Therefore, the demand for logging workers decreased because of the increased productivity caused by mechanization. Specifically, the mechanized harvesting systems are displacing the chainsaw fallers (Conrad et al. 2018). The employment of logging equipment operators had remained stable, from 22,130 in 2007 to 21,290 in 2017, while the employment of chainsaw fallers declined substantially, from 6,380 in 2007 to 3,880 in 2017 (Bureau of Labor Statistics 2020a), indicating that capital leads to a reduction in employment in the long run.

On the other hand, employment plays a decreasing role in driving capital with the extension of the forecast period. This result may be explained by the fact that the logging firms need to consider costs, markets, profitability, and ease of obtaining loans to decide whether to purchase machines in the long run or not. Therefore, the impact of employment on capital, in the long run, will be reduced.

As shown in Table 2, the wages have a small influence on capital in the long run. This result might be explained by the fact that wage increases have been long term and stable in promoting logging firms to choose to outsource some of their businesses instead of mechanization. In general, mechanized harvesting systems require tremendous investment.

It is acknowledged that employment is influenced by capital and the product price. However, the product price is determined merely by itself in the long run, which is consistent with results of the DAG. This feature may partly be explained by price rigidities, such as menu costs and other frictions in adjusting prices (Angeletos and La'O 2009) or imperfect information (Lucas 1972, Mankiw and Reis 2002).