Wages, Competition, and the Decline of Union Membership in the US Manufacturing Industry Simona Aksman New York University, Department of Economics Undergraduate Honors Thesis Adviser : Katarina Borovickova Final Draft May 4, 2015 Abstract This paper examines the empirical application of the hypothesis that union membership has declined in the US manufacturing industry due to an increasingly competitive environment among firms, despite the fact that union wages have remained higher than nonunion wages. I construct an industry-level model for union membership as well as union wages that has competitive indicators, including inward foreign direct investment, the capital-to-labor ratio, terms of trade, and a ratio of export to import trade volume. I combine the theory of union-firm negotiations and the findings of my empirical study to make the argument that union wage increases drive union membership downwards through competition. As the manufacturing industry faces increasing competition from imports abroad as well as capital’s replacement of labor, firm profits diminish, and the ultimate result is that unions have declining membership. The key finding of my paper is that the underlying cause of the union membership decline is an increasing union wage, which triggers competitive forces to drive down union membership. 1 Introduction and Motivation For the past several decades, union membership in the United States has been steadily declining. According to a 2015 report by the Bureau of Labor Statistics (BLS), it declined from 35.5% to 11.1% between 1945 and 2014. For private industry, the 2015 BLS report stated that it declined even further, to 6.6% in 2014. Nevertheless, in 2014 union members had median usual weekly earnings of $970 while their nonunion counterparts had median earnings of $763, such that a positive union-nonunion wage differential exists. It is important to note that these wages are reflective of variations in individual demographic and industry characteristics and are therefore likely to be biased estimates. Still, according to the literature, this positive union-nonunion differential is a significant argument for the existence of unions. Given the assumption that this positive wage differential has persisted, I’d like to gain insight into the decline in union membership. 1 In particular, the manufacturing industry is of interest for study because it presents a case of a significant union membership decline in the past few decades. Historically, the manufacturing industry was highly unionized due to its concentration of production occupations (Griswold 2010).1 But then, between 1973 and 2006, the share of total manufacturing jobs to all private industry wage and salary jobs fell by 17.9% (Hirsch 2008). However, this drop occurred primarily in the unionized sector of manufacturing while the share of nonunion manufacturing jobs actually slightly increased between 1973 and 2006. The result is that 6 million fewer union members were employed in manufacturing in 2006 compared to in 1973. This trend is also reflected by the job growth of private industry in the US overall: while union sector jobs have declined, nonunion sector jobs have remained relatively flat in the past few decades. According to the literature, there exist three main categories of explanations for the decline in union membership: structural, competitive, and institutional (Hirsch 2008). The structural explanation captures significant movements in employment within occupations, industries, and regions where union density has been traditionally high. The competitive explanation captures the effects of unions on firms in a competitive environment, as well as an environment of globalization and technological change. The institutional explanation captures all other environmental factors that affect union organizing, including the legal environment, government regulations, worker preferences, and managerial opposition. I’d like to explore the competitive explanation for the manufacturing industry’s decline. The basis of this explanation is the assumption that as a market becomes more competitive, profits diminish for each firm. In addition, union rents can be thought of as taxes on company profits (Hirsch 2008). This taxation effect is not conducive to increasingly competitive industries that exhibit diminishing profits. Past studies and anecdotal evidence point to globalization as a primary driver of increasing competition (Slaughter 2007). In the past couple decades, US industries, and in particular the manufacturing industry, have become increasingly globalized: many mechanized processes that were once done on US soil have been moved abroad. In the short span of time between 1997 and 2002, the share of foreign-sourced goods in total manufactured inputs nearly doubled for the US manufacturing industry (Burke et al. 2004). My objective in conducting this research is to construct an empirical model for union membership’s decline, built with the intuitions noted above. In particular, I’m interested to see whether competitive changes in the manufacturing industry have had a negative effect on union membership over the long-run, which I’ve defined as the period between 1993 and 2012. In addition, I would like to test whether the union-nonunion wage differential is actually a significant determinant of union membership as the theory suggests. To test the competitive hypothesis, I use a collection of competition indicators that are supported by the literature and theory. In particular, foreign direct investment and international trade measures, such as terms of trade and export to import trade ratio, are suggested by Slaughter (2007; 2001). Capital-to-labor ratio is another measure I utilize, as suggested by Hirsch and Berger (1984). I give a more in-depth explanation of how these factors relate to union membership in section three. Additionally, due to 1 Production occupations include occupations such as precision production, craft, repair work; machine operation; transportation and materials moving; as well as equipment handling and construction (Caves and Krepps 1993). 2 the similar nature of these measures to one another, I select a combination of the best indicators for each of the manufacturing sub-industries using an algorithm that removes multicollinearity.2 I explain this further in section six where I describe results. In the following section, I give background on the relevant literature on union membership models and the effects of competition. In section three, I discuss a theoretical model for how unions and firms interact given profit and utility constraints. In section four, I discuss the methodology for my econometric analysis of the competitive determinants of union membership. In section five, I review the data I work with and how I used them within my econometric model. In section six, I give an overview of the results of my econometric analysis. The last section provides my conclusions. 2 Literature Review A key subset in the relevant literature examines the relationship between union membership and market competition, and particularly of interest, the effect of market competition on union membership. As I mentioned in the introduction, Hirsch (2008) argues that competition inhibits union rent-seeking activities, because union rent-seeking can be thought of as a tax on firm profits. An increase in competition eats away at these profits available to unions. In particular, he points to three main factors that explain competition in product markets: international trade, product market deregulation, and the entry of low-cost competitors. In a study of union membership and competition, Slaughter (2007) examines manufacturing industries, particularly those with a significant international presence, and finds evidence of union decline. He concludes that union membership declines as a result of competition arising from increased globalization, which he measures using inward foreign direct investment (FDI). Slaughter explains that FDI reflects an increase in capital mobility on the global scale, which has the effect of increasing labor demand elasticity. Specifically, unions are threatened by the relocation of production abroad that FDI indicates, and they are threatened by this effect even if domestic firms are not doing so. This theory that unions are adversely affected by indirect signals of globalization presents a case of the "demonstration effect", and has support from several other studies, such as Rodrik (1997) and Scheve and Slaughter (2004). Slaughter (2001; 2007) also explores the effects of international trade on union membership, and provides a similar theory to the link between FDI and union membership: increasing trade leads to increasing labor demand elasticity, resulting in lower union membership. However, though the theory provides a clear story of a negative impact of international trade on union membership, in studies, international trade indicators have been shown to have mixed effects on union membership, with sometimes an unexpected sign on the coefficient of the indicator. In his 2007 study, Slaughter finds that measures such as exports, imports, net exports, and transportation costs do not have a significant correlation with union membership or a consistently correct sign for the coefficient. In my study, I use some of the global competition indicators used by Slaughter in his 2001 and 2007 studies, such as FDI, exports, and imports. Griswold (2010) suggests an interesting theory of why international trade may actually 2 Multicollinearity is limited when the Variance Inflation Factor (VIF) for each indicator with respect to the other indicators is less than 5. 3 increase the union labor share instead of decreasing it. If workers are aware of international trade negatively affecting worker security within an industry, they may value union protection more, thereby actually increasing union membership. Griswold also criticizes the notion that much of the decline in union membership is the result of competition, and points to structural and institutional factors as important additional explanations. Baldwin (2003) finds that most of the decline in union membership is not the result of international trade, but rather, other factors such as employer opposition, unfavorable legislation, and declining worker demand for unionism. Additionally, Griswold discusses the literature and theories of the effects of competition on union membership. He points out that a common theoretical explanation for the decline in union membership is that competition makes changes in the union labor share more sensitive to changes in union wages. This concept is useful for understanding empirically how competition affects unions: the underlying trigger for a union membership change is a wage change. Competition has the effect of increasing the negative effects of wage increases on union membership, but competition alone is not responsible for these negative effects. This is an idea I explore further in the theoretical analysis section of this paper, and also has implications for the interpretation of my results. There have been several studies on union membership determination, though most have focused on individual rather than industry-level characteristics. In one of the earliest of these individual-level studies, Lee (1978) modeled the effects of a union-nonunion wage differential on union membership using microdata for the manufacturing industry. Lee concluded that a higher union-nonunion wage differential has a positive correlation with the union membership of an individual. The methodology proposed by Lee is the basis of my econometric model, as explained in the methodology section of this paper. This model was later adopted by Hirsch and Berger (1984), though the focus in this adopted model was instead a mix of individual and industry effects in the manufacturing industry using a cross-section of Current Population Survey (CPS) microdata for one year. Industry effects used by Hirsch and Berger included the capital-to-labor ratio and the Herfindahl-Hirschman Index (HHI), a measure of market concentration. The team concluded that the capital-to-labor ratio and HHI are statistically-significant determinants of union membership for the manufacturing industry. The union-nonunion wage differential, however, was found to have no significant effect on union membership. I use the capital-to-labor ratio and union-nonunion wage differential in my model of union membership determination as well. Farber (2001) points out that the union membership determination model by Lee (1978), which he calls the worker-choice model, ignores the issue of workers who are interested in union jobs but are not able to find them. This is because unions are costly to organize, and it is often the case that although union job seekers outnumber union jobs, job seekers are not willing to take on the cost of organizing a new union. This results in a queue for existing union jobs. This theory is the basis of a queuing model established by Farber (1983): the union status of a worker is not only based on decisions of the worker, but also the employers, who choose whether or not to unionize. There is also the issue of skill level, which biases whether or not a worker will be in a union. Unions set a wage based on the median worker along the skill distribution. Therefore, the workers that are the most skilled will not have an incentive to be in the union, and the those that are the least skilled will not be chosen by the unionized employer. Given that I have chosen 4 to work with the union membership determination model established by Lee (1978), I take the above criticisms into account when constructing my own model. The change I make to my model is to model union membership on the industry level rather than on the individual level, such that the dependent variable is union membership as a share of total employment for an industry rather than as a binary choice for an individual. 3 Theoretical Analysis Union contracts, and in particular, wages and employment, are determined within a union-firm bargaining process. It is assumed that the goal of the union is to maximize a utility function U (w, L), where w refers to wages and L refers to employment, subject to a constraint. There are two classes of models that have been used to solve this maximization problem: the labor demand curve model and the contract curve model (Farber 2001). The labor demand curve model assumes that the level of employment is fixed by the firm and the union therefore only aims to maximize its member’s wages. The contract curve model, on the other hand, assumes that both wages and employment are maximized by the union in the bargaining process, and the union’s utility is determined on a union-firm contract curve rather than the firm’s labor demand curve. This model has two cases: the weakly efficient case, where unions maximize wages and employment subject to the firm’s minimum profit constraint, and the strongly efficient case, where unions maximize wages and employment subject to the opportunity wage of workers. Based on the results of MaCurdy and Pencavel (1986), I assume that the weakly efficient case of the contract curve model best describes the maximization problem that unions and firms face. In this case, the union-firm bargaining process can be characterized from the union’s perspective as given below, max U (w, L) s.t. fi(w, L) Ø fi w,L (1) where U (w, L) is the utility of the union in terms of wage and labor and fi(w, L) Ø fi refers to the minimum profit constraint of the firm. This maximization objective can also be written in its dual form to express the firm’s perspective as: max fi(w, L) s.t. U (w, L) Ø U fi where fi(w, L) is the profit of the firm in terms of wage and labor, and U (w, L) Ø U refers to the minimum utility constraint of the union. Further, I assume that the union will set its utility exactly equal to the level at which the firm will make an agreement: U (w, L) = U (2) where U is the union’s minimum utility level. At this level, the firm and union make an agreement that satisfies the other party’s bargaining constraint. Given that the union’s utility will be set at this point, the following first order condition holds true: ˆU (·)/ˆL w ≠ M RL = ˆU (·)/ˆw L (3) where M RL is the marginal revenue product of labor, ˆU (·)/ˆL refers to the partial derivative of utility with respect to labor, and ˆU (·)/ˆw refers to the partial derivative of utility with respect to wages. Equation (3) shows that the marginal rate of substitution 5 between wage and employment will be equal for the firm and union. This equation can be re-written to solve for M RL , as given below: M RL = w(1 + ‘) (4) ˆU (·)/ˆL where ‘ = ˆU ú wL which can be simplified to ‘ = ˆw ú wL , the expression for the (·)/ˆw ˆL elasticity of labor demand. Elasticity of labor demand is defined as the sum of the substitution effect and the scale effect. According to the substitution effect, when wages rise, a firm will substitute capital for labor. The scale effect, on the other hand, refers to the overall cost of production: as wages rise, the overall cost of production rises, such that the quantity of output declines and the amount of labor required declines. Increased competition can raise elasticity of labor demand via both effects. FDI is capable of raising elasticity through both substitution and scale effects, and Slaughter (2007) gives an example of how FDI raises the elasticity via the substitution effect. If a vertically integrated multinational firm moves some of it’s manufacturing production stages abroad, it gains access to a greater pool of input resources. Then, given a wage rise, it is able to substitute capital for labor more easily, and ‘ rises. Since ‘ is assumed to be negative, as ‘ rises, M RL will diminish given the relationship between the variables established in Equation (4). As M RL diminishes, marginal profits will diminish since marginal profit can be decomposed into M fiL = M RL ≠ M CL (5) where M fiL is the marginal profit of labor and M CL is the marginal cost of labor. For a competitive market in the long-run, M fiL = 0, a condition equivalent to M RL = M CL . When this occurs, the minimum profit threshold fi will become a stricter boundary for unions, and union contract sets will reach lower levels of L, union membership, than would be the case if profits were not diminishing. An important note to make here is that w is increasing throughout, such that only L will decrease in the U(w,L) set. Given that the condition of U (w, L) = U from Equation (2) holds true, one can map out the union’s indifference curves, since U is an equilibrium outcome in union-firm negotiations. Figure 1 shows these equilibrium outcomes, and the data appears to affirm the theory that L diminishes as w increases with the various observed equilibrium outcomes for the years of my study, 1993 - 2012. I further explore the hypothesis that increasing union wages correspond to decreasing levels of union membership in the econometric study I conduct later in this paper. Other indicators, such as the capital-to-labor ratio, have a more direct link to the two effects of elasticity of labor demand. The capital-to-labor ratio provides a strong indicator of the relationship between the mechanization of the labor force and firm profitability. Given a wage rise, when firms substitute capital for labor, the substitution effect leads directly to an increase in the capital-to-labor ratio. Once the capital-to-labor ratio rises, ‘ rises and M fiL diminish, as shown previously in Equations (4) and (5). 6 7 Figure 1: Mean equilibrium outcomes of wage and labor for the manufacturing industry (9 sub-industries included). Figure (1a) and Figure (1b) show that there is a similar wage-labor trend overall with both the union wage and the union-nonunion wage differential. Source: Mean union wages are predicted using the 2SLS procedure explained in the methodology section of this paper, and mean union labor shares are based on data tables collected from unionstats.gsu.edu. Similarly, an increase in international trade is also theorized to increase ‘ via the substitution and scale effects, and thus diminish M fiL for the firm. Slaughter (2001) suggests that international trade can lead to a greater pool of resources, and thus more substitution possibilities, leading to a substitution effect. Given a wage increase, when import trade volumes increase, there are more resources with which the substitution and scale effects can occur. In terms of the export/import trade volume ratio (Exp/Imp), an increase in import trade volume corresponds to a decline in Exp/Imp. In addition, as a market becomes increasingly competitive, firms will become increasingly price sensitive and mark-ups shrink. This mark-up rule can be written as the following equation P ≠ M CL ≠1 = (6) P ‘ where LI is the Lerner Index, and P is price. Given the inverse relationship between mark-ups and ‘ in Equation (6), as firm-level ‘ rises, mark-ups shrink. When firm-level ‘ rises, the entire market will experience a rise in ‘. For example, one international indicator I use in this study is terms of trade (TOT), a ratio of export to import price indices. When production abroad becomes cheaper than domestic production, TOT will increase. An increase in TOT can therefore be reflective of an increasingly competitive trade environment and corresponds to increased price sensitivity, declining mark-ups, and an increase in firm-level ‘ as given in Equation (6). The union-firm bargaining process model given above reveals how unions and firms make their decisions based on profit constraints. When profits are diminishing, as is the case for an increasingly global and competitive market, fi becomes a scarcer quantity, and therefore, a stricter boundary in the union-firm bargaining process. The result is that unions see their equilibrium outcomes diminishing in labor opportunities over time. LI = 4 Methodology The model of union membership I estimate is based on a model originally presented by Lee (1978) and modified by Hirsch and Berger (1984). Based on the assumptions of my union-firm bargaining framework, workers determine whether or not to join a union based on wage preferences. For a given manufacturing worker i, Wui and Wni are his union and nonunion wages, respectively. Given that this worker has a reservation union-nonunion wage differential fli , he will choose to join a union given the following condition, Wui ≠ Wni > fli Wni (7) such that the percentage union-nonunion wage differential exceeds his percentage reservation wage differential, which then represents the worker’s preference for union membership, and can be positive or negative. fli can be described as a function of person’s individual characteristics and costs associated with union membership, given by the equation fli = –Xi + —Ci + Á1i (8) 8 where Xi is a vector of individual characteristics, Ci is the cost of union membership, 2 and Á1i is the error term, assumed to be distributed N(0, ‡1e ). Since the cost of union membership is not necessarily encompassed by an explicit cost, such as a yearly fee, Ci is split into two variables: the observable (explicit) and unobservable (implicit) cost, the latter of which can be assumed to be a residual and uncorrelated with Xi . The observable cost can be written as Ci = “1 + “2 Xi + “3 Zi + Á2i (9) where Xi is a vector of individual characteristics, Zi is a vector of industry characteristics, 2 and Á2i is the error term, assumed to be distributed N(0, ‡2e ). The individual worker will choose to join a union if the following condition is met Wui ≠ Wni > (– + —“2 )Xi + —“1 + —“3 Zi + Á1i + —Á2i Wui (10) i.e. their percentage union-nonunion wage differential exceeds the cost of joining a union. This can be rewritten to represent the union membership decision. The model of union membership takes a probit form, since union membership is a qualitative binary variable. Uiú > 0 if worker i is in a union and 0 otherwise. This can be written as Uiú = ”0 + ”1 3 4 Wui ≠ Wni + ”2 Xi + ”3 Zi ≠ Ái Wui (11) where Xi is a vector of individual characteristics, Zi is a vector of industry characteristics, and the error term is assumed to be distributed N(0, ‡i2 ). An individual’s union status is thus determined by his or her percentage union-nonunion wage differential, individual characteristics, as well as his or her industry’s characteristics. In order to estimate the union-nonunion wage differential, a simultaneous model as derived by Lee (1978) is introduced. Log forms of the union and nonunion wages are substituted in to simplify the functional form, since the percentage union-nonunion wage ni differential WuiW≠W can be approximately written as ln(Wui ) ≠ ln(Wni ). The equations ni are as given below, ln(Wui ) = ◊u0 + ◊u1 Xui + ◊u2 Zui + Áui (12) ln(Wni ) = ◊n0 + ◊n1 Xni + ◊n2 Zni + Áni (13) where ln(Wui ) and ln(Wni ) are the log union and nonunion wage rates for an individual i, Xui and Xni are vectors of individual characteristics for union and nonunion workers, and Zui and Zni are vectors of industry characteristics for union and nonunion workers. 2 The error terms for the union and nonunion industries are distributed N(0, ‡ui ) and N(0, 2 ‡ni ), respectively. Some consideration of union bias is required in order to properly estimate ln(Wui ) and ln(Wni ). The issue is that samples of union and nonunion workers are not randomly drawn from the population, and therefore, wages can be biased by union status. As a result, the error terms of the union and nonunion wages cannot be assumed to be independent of union status. This can be expressed as E(Áui |Ui = 1) ”= 0 and E(Áni |Ui = 0) ”= 0 (14) Given the above condition, it is necessary to include an instrumental variable as well as a modified least squares procedure when determining the union and nonunion wages in 9 Equations (11) and (12). This instrumental variable will be an expression for the means of the wage equations, E(Áui |Ui = 1) and E(Áni |Ui = 0), that adjust the error terms to remove the dependency on union status. The following expressions are instrumental variables for mean values that can be used to adjust the error terms: E(Áui |Uiú Ø 0) = (‡ui /‡i )[≠f (Uiú )/F (Uiú )] (15) E(Áni |Uiú < 0) = (‡ni /‡i )[f (Uiú )/1 ≠ F (Uiú )] (16) ln(Ŵui ) = ◊u0 + ◊u1 Xi + ◊u2 Zi + ◊u3 (‡ui /‡i )[≠f (Ûi )/F (Ûi )] + ÁÕui (17) ln(Ŵni ) = ◊n0 + ◊n1 Xi + ◊n2 Zi + ◊n3 (‡ni /‡i )[f (Ûi )/1 ≠ F (Ûi )] + ÁÕni (18) where ‡ui and ‡ni are the covariances between Áui , Áni and Ái , f (Uiú ) is the standard normal density function, and F (Uiú ) is the cumulative distribution function of the standard normal distribution. The sample selection bias is controlled by estimating the expressions for expected value, (‡ui /‡i )[≠f (Uiú )/F (Uiú )] and (‡ni /‡i )[f (Uiú )/1 ≠ F (Uiú )], and inserting these expressions into the union and nonunion wage equations to generate unbiased estimates: where E(ÁÕui |Ûi = 1) = 0 and E(ÁÕni |Ûi = 0) = 0. Equations (17) and (18) above express the union and nonunion wages conditional on individual and industry characteristics as well as the adjusted mean value terms, which control for wage bias from union status. These new estimated values of the union and nonunion wages are used to calculate the wages for each individual in the data set. The final modified least square procedure is a 2 stage least squares (2SLS) procedure: in stage 1, it produces an estimate of Ûi with a probit model, and then in stage 2, it produces an estimate of ln(Ŵui ) and ln(Ŵni ) with an OLS model that includes the instrumental mean values that correct for sample selection bias. Below are the individual-level features used for the determination of both union status and the union and nonunion wages: Xui1 = Xni1 = Xi1 = North-central regional dummy3 Xui2 = Xni2 = Xi2 = South regional dummy Xui3 = Xni3 = Xi3 = West regional dummy Xui4 = Xni4 = Xi4 = In metropolitan area dummy Xui5 = Xni5 = Xi5 = Is married dummy Xui6 = Xni6 = Xi6 = Is white dummy Xui7 = Xni7 = Xi7 = Is male dummy Xui8 = Xni8 = Xi8 = Is production worker dummy4 Xui9 = Xni9 = Xi9 = Years of schooling Xui10 = Xni10 = Xi10 = Experience 3 For all features marked as dummy variables, the value is 1 if the dummy variable condition is met and 0 otherwise. 4 As explained previously, production worker status encompasses blue-collar professions such as precision production, craft, repair work; machine operation; transportation and materials moving; as well as equipment handling and construction. Non-production worker status encompasses executive, administrative, and managerial work; professional specialty; technicians and support; sales; administrative support and clerical work; as well as service occupations (Caves and Krepps 1993). 10 Xui11 = Xni11 = Xi11 = Experience squared Xui12 = Xni12 = Xi12 = Number of children Xui13,14,15,..,32 = Xni13,14,15,...,32 = Xi13,14,15,...,32 = Year dummy (for each year 1994 through 2012) Each of the simultaneous equations is run on the sub-industry level, such that subindustry characteristics, Zui , Zni , and Zi are implied in the data. After the 2SLS procedure, all union-nonunion wage differential data is aggregated on the sub-industry level. Then, a linear regression model for union membership is created that is similar to Equation (10). Unlike that model, however, this model is a panel, with both industry and time dimensions. The dependent variable is union membership as a percentage share for each manufacturing sub-industry across time. This model is written below as: F DIjt K EXP ) + ⁄3 ( )jt + ⁄4 ( )jt + ⁄5 T OTjt + Ájt CP It L IM P (19) where Ujt is the share of unionized workers to total workers in a given sub-industry, ln(Ŵujt ) ≠ ln(Ŵnjt ) is the estimated mean union-nonunion wage differential, and the rest DIjt of the variables are industry characteristics: FCP is real inward foreign direct investment It K in terms of the CPI at time t, ( L )jt is the capital to labor ratio, ( EXP ) is the ratio of IM P jt the export volume to import volume, and T OTjt is the terms of trade, a ratio of export prices to import prices. All variables are across time periods t = 1993,...2012, and subindustries j = 1,...9. These sub-industries are specified in the following section. The error 2 term is assumed to be distributed N(0, ‡jt ). Sub-industry fixed effects are utilized when necessary, and are coded in Table 4. In addition, given the theoretical intuition established in section 3 of the union’s utility trade-off between labor and wages, I run an additional linear regression model where the union wage is now the dependent variable.5 Ujt = ⁄0 + ⁄1 [ln(Ŵujt ) ≠ ln(Ŵnjt )] + ⁄2 ( ln( Ŵujt K EXP ) = ⁄0 + ⁄1 Ujt + ⁄2 F DIjt + ⁄3 ( )jt + ⁄4 ( )jt + ⁄5 T OTjt + ÁjtÕ CP It L IM P (20) Ŵujt where ln( CP ) is the estimated real union wage in terms of the CPI at time t, and all It independent variables are the same as those given in Equation (19). The error term is 2 assumed to be distributed N(0, ‡jt Õ ). I use the estimated union wage instead of the unionnonunion wage differential as the dependent variable in Equation (20) given that I am interested in determining if competitive effects are tied to the union wage as suggested by the theory. As was the case for the previous model, sub-industry fixed effects are utilized when necessary. These fixed effects are coded in Table 5. By running regressions of both labor and wage, I consider both dimensions of the union’s equilibrium utility outcome U (w, L). This is a more comprehensive approach to modeling the effects of competition with respect to the union’s utility. 5 See Figure 1 for a visualization of the union’s trade-off between wages and labor. 11 5 Data In order to construct the union-nonunion wage differentials using the 2SLS procedure, I needed individual-level microdata to derive the fitted union membership values and individual characteristics that bias the wage differentials. I collected samples of this data from the University of Minnesota’s Integrated Public Use Microdata Series (IPUMS) from the Current Population Survey (CPS). The variables used from this data source were available for March of each year from 1993 though 2012. I used the March sample because it is the only month of the year that has a complete data set for both weekly earnings and all individual characteristics I wanted to include, which I based on a combination of what Lee (1978) as well as Hirsch and Berger (1984) included in their studies. Each year’s CPS data is a representative sample. The samples I worked with were restricted to wage-earning workers between the ages of 16 and 64, employed in both production and non-production manufacturing occupations. The sub-industries included in this study are: Food, Beverages & Tobacco, Paper, Chemicals, Plastics & Rubbers, Nonmetallic Minerals, Primary & Fabricated Metals, Machinery, and Transportation. The individual-level variables used are: year, region (northwest, north central, south, west), metropolitan area, marital status, production worker status (blue-collar), number of children, gender, race, years of schooling, experience, and experience squared. Categorical variables, such as year, marital status, number of children, race, and gender, were re-coded into dummy variables. Experience was computed using the equation: experience = age - years of schooling - 6. Experience squared was included given the intuition that an individual’s earnings over the course of their lives often exhibit a quadratic path rather than a linear one. To construct the union-nonunion wage differentials for each sub-industry, I computed the log of the predicted average weekly union wage minus the log of the predicted average weekly nonunion wage. The predicted wages were the output of the regressions given in Equation (17) and Equation (18) for each sub-industry in each year from 1993 though 2012. As for the industry level variables, I utilized a variety of data resources. For inward FDI, I collected data from the U.S Department of Commerce’s Bureau of Economic Analysis for the manufacturing sub-industry noted above for the period of 1993 - 2012. Given that the inward FDI data provided on the site was in nominal terms, I computed real FDI by dividing the nominal FDI by the average CPI for each year. Export and import trade volume data was collected from several resources. For the period of 1993 - 2005, NAICS-converted data was collected from the Yale Social Sciences Library webpage (Schott 2010; Brambilla et al. 2010). I used customs value-basis imports to determine import volume, and summed up both imports and exports for each year and across the relevant sub-industries. From 2005 - 2014, I used the US Census Bureau’s Industry Statistics Portal to collect values of export and import volume, which were already computed for each sub-industry based on NAICS values. The capital-to-labor ratio was collected from the BLS Productivity database for the period of 1993 - 2012.6 Terms of trade was also collected from the BLS Productivity database. I computed terms of trade by dividing the US price index for exports by the US price index for imports. Both capital-to-labor and terms of trade measures were 6 Note that in the BLS database, it is called the capital-to-hours ratio. 12 Union Labor Share (%), 1993 - 2012 Manufacturing Sub-Industry Mean Food 18.73 Beverages & Tobacco 20.09 Paper 25.56 Chemicals 9.02 Plastics & Rubbers 17.29 Nonmetallic Minerals 17.64 Primary Fabricated Metals 19.89 Machinery 13.38 Transportation 23.65 Nondurable Goods 18.14 Durable Goods 18.64 All 18.36 St. Deviation 3.82 5.60 3.82 5.53 2.93 4.63 4.77 5.25 4.36 6.95 5.86 6.48 Minimum 12.97 11.80 18.57 4.37 7.23 10.26 11.15 7.11 17.53 4.37 7.11 4.37 Maximum 26.88 31.60 37.33 14.48 24.03 28.08 28.29 21.74 29.82 37.33 29.82 37.33 Table 1: Source: Union membership share data collected directly from unionstats.gsu.edu. Predicted Wage Diff. (%), 1993 - 2012 Manufacturing Sub-Industry Mean St. Deviation Minimum Maximum Food 1.29 1.29 -1.00 4.09 Beverages & Tobacco 0.13 3.29 -6.13 5.29 Paper 2.16 1.21 0.22 5.77 Chemicals 1.03 1.56 -0.83 4.91 Plastics & Rubbers 1.76 1.61 -1.21 5.06 Nonmetallic Minerals 1.06 1.74 -2.09 5.17 Primary Fabricated Metals 1.04 1.03 -0.60 2.87 Machinery 0.76 1.49 -1.18 4.17 Transportation 0.90 1.59 -2.27 4.03 Nondurable Goods 1.27 2.03 -6.13 5.77 Durable Goods 0.94 1.47 -2.27 5.17 All 1.13 1.81 -6.13 5.77 Table 2: Source: These predicted union-nonunion wage differentials were constructed via the 2SLS procedure as explained in section 4 of this paper, with original data from the CPS. 13 Competitive Indicators, 1993 - 2012 FDI (Real $) Mean St. Deviation Nondurable Goods 502.20 656.51 Durable Goods 553.57 347.74 All 525.0 540.74 EXP TOT ( Index ) IndexIM P Nondurable Goods 0.99 0.07 Durable Goods 1.00 0.06 All 0.99 0.07 $EXP Exp/Imp ( $IM P ) Nondurable Goods 0.98 0.36 Durable Goods 0.76 0.24 All 0.88 0.33 K/L (Index Base = 100) Nondurable Goods 80.98 16.88 Durable Goods 73.34 15.14 All 77.58 16.53 Minimum Maximum 48.70 3148.90 67.62 1471.68 48.70 3148.9 0.77 0.86 0.77 1.18 1.19 1.19 0.26 0.39 0.26 1.85 1.33 1.85 36.93 45.02 36.93 109.00 102.48 109.00 Table 3: Source: FDI data from the U.S. Department of Commerce, TOT and K/L data from the BLS, and Exp/Imp data from the Yale Social Science Library and the US Census Bureau. available at the sub-industry level and were relatively consistent with the NAICS industry classifications I used, though averaging and imputation was sometimes required to construct some measures. Where imputation was required, I used an average value of the previous and next values as a proxy for the missing value. Finally, sub-industry level union shares came from a data set derived from CPS reports by Hirsch and Macpherson (Hirsch and Macpherson 2009).7 Table 1 presents the descriptive statistics for union membership over the period of this study. In addition, Figure 2 in the Appendix shows the trend over time for union membership. One can see here that union membership has had a consistent decline across the entire manufacturing industry. The most unionized sub-industry of manufacturing was Paper, which had a mean union share of 25.56% as well as the lowest standard deviation in union share. The Chemicals sub-industry, another Nondurable Goods industry, had the lowest mean union share at 9.02%, a little over one third of the Paper industry’s mean union share. These numbers suggest that while the union share across the manufacturing industries varies considerably, it is generally low, with the maximum of all industries at 37.33% and the mean at 18.36%. The decline in union membership is not a new trend, with the beginnings of it going back to the 1970s (Hirsch 2008). The union membership data in the time period of my study reflects a long-run decline in union membership that spans the whole manufacturing industry. Similarly, descriptive statistics for the predicted wage differentials have been given in Table 2. This table shows that, for most industries, the predicted union wage is 1-2% above the nonunion wage. Figure 2 in the Appendix portrays what’s happened over time, 7 Data set available online at unionstats.gsu.edu. 14 and it shows that the union-nonunion wage differential has increased across all industry segments, which corresponds with the wage trend visible in Figure 1. Finally, Table 3 presents descriptive statistics for all competitive effects. Figures 2 and 3 reveal that the competitive effects, with the exception of TOT, follow a clear trend over time. Figure 3 shows that TOT has two opposing trends for the Nondurable and Durable Goods segments over time, which results in an unclear trend overall. Perhaps this is reflective of vastly different export and import trade pricing for the Durable and Nondurable Goods industries. This lack of consistency in the TOT trend across industries is also reflected in its performance in my econometric study, where TOT has unexpected signs on its coefficients. Exp/Imp, on the other hand, has trends for the Nondurable Goods and Durable Goods segments that appear to move together. For manufacturing overall, it has a mean value of 0.88, meaning that import volume exceeds export volume. This suggests that the US manufacturing industry is at a comparative disadvantage in the long run, particularly among the Durable Goods industries, where Exp/Imp has a mean value of 0.76. Though the comparative disadvantage may be greater for the Durable Goods segment than the Nondurable Goods segment, where the mean Exp/Imp is close to 1, the standard deviation for the Nondurable Goods segment is higher than for the Durable Goods segment. Figure 3 also reflects this difference in variability of the two segments: while the Durable Goods segment has low, but relatively consistent Exp/Imp values of less than 1 for all twenty years of the study, the Nondurable Goods segment shows a rather steep descent in Exp/Imp over time. This is especially visible between 1995 and 2005, a period during which there was a 44% decline in Exp/Imp. For both segments, the trends seem to reverse between 2005 and 2009, and then again decline after 2009, this time perhaps as a result of the 2008 financial crisis. This decline around the financial crisis may also be visible in the trends of other competitive indicators as well: Figure 2 shows that FDI and K/L have clear upward trends until there is a decline in both around the period of the financial crisis. Though I do not take the financial crisis into account in my econometric study, it could potentially explain some of the residual. 6 Results Tables 2 and 3 present the results of the final regressions of union labor share (Ujt ) and Ŵujt wages (ln( CP )) for each sub-industry of manufacturing between 1993 and 2012. In the It tables, coefficients and standard errors for each explanatory variable are presented. In addition, given that for each sub-industry I ran separate regression models, each model has an R2 value and F-statistic to provide insight into how well each model performed overall. Variables that are marked with a dash in the tables were not included in the model for that specific sub-industry. This is the result of my modeling procedure, by which I removed variables with high multicollinearity using a stepwise algorithm that calculated Variance Inflation Factors for each combination of variables in a model.8 8 For more details on how this algorithm works, see https://beckmw.wordpress.com/2013/02/05/collinearity-and-stepwise-vif-selection/ 15 this webpage: Labor Share (Ujt ) Regression Manufacturing Sub-Industry Food Beverages & Tobacco Paper Chemicals Plastics & Rubbers Nonmetallic Minerals Primary Fabricated Metals Machinery 16 Transportation Nondurable Goods Durable Goods All Wage Diff. -40.375 (56.495) -87.301** (30.575) -132.757 (91.565) -61.472 (31.441) 1.497 (38.627) -11.419 (36.773) -83.853 58.414 -251.814*** (47.270) -91.229 (54.278) -80.988*** (18.798) -107.251*** (29.034) -56.703*** (15.897) FDI -0.002 (0.005) -0.002 (0.006) -0.067 (0.037) -0.013*** (0.003) - TOT -7.739 (8.106) 32.312** (14.387) -46.188** (18.771) -1.826 (21.044) -9.123 (12.953) 32.936 (17.437) 22.735*** (5.982) 18.546 (18.883) 3.873 (46.387) -24.358*** (5.675) 23.553*** (5.669) 7.674 (4.071) Exp/Imp 8.997** (3.134) -6.154 (11.408) 8.280** (3.253) 24.749*** (6.520) 26.574** (9.314) 8.194 (4.955) 2.939 (6.170) -2.317 (6.737) 8.014*** (1.451) 3.067** (1.206) K/L - D1 - D2 - D3 - D4 - D5 - D6 - D7 - D8 - R2 0.704 F-stat 8.923 - - - - - - - - - 0.656 10.16 - - - - - - - - - 0.611 5.885 - - - - - - - - - 0.732 14.57 - - - - - - - - - 0.707 12.89 - - - - - - - - - 0.872 25.5 -0.250*** (0.057) - - - - - - - - - 0.904 35.36 - - - - - - - - 0.654 10.1 - - - - - - - - 0.706 9.034 -0.413 (1.313) - 6.738*** (1.454) - 10.684*** (1.162) - -9.951*** (1.077) - - - - - 0.810 55.59 - 4.324*** (0.992) -15.609*** (1.085) -5.072*** (1.048) -3.977*** (0.905) -2.774*** (0.936) -9.451*** (0.859) -10.941*** (1.058) 51.33 -0.604 (1.018) -3.774*** (0.893) -4.288*** (0.963) 0.808 -3.315** (1.305) 0.814 60.93 -0.127 (0.067) -0.144*** (0.028) -0.174*** (0.018) Table 4: Significance codes: 0.01: ***, 0.05: **, Note: for each sub-industry’s attributes, the first row indicates the coefficient and the second row indicates the standard error. Nondurable Goods refers to Food (D1), Beverages & Tobacco (D2), Paper (D3), Chemicals (D4), and Plastics & Rubbers (D5). Durable Goods refers to Nonmetallic Minerals (D6), Primary Fabricated Metals (D7), Machinery (D8), and Transportation (No dummy). Codes D1 - D8 refers to the dummy variables that control for industry fixed effects included in the regressions for Nondurable Goods, Durable Goods, and All. Ŵujt Wage (ln( CP I )) Regression t Manufacturing Sub-Industry Food Union Share -0.005 (0.005) -0.025** (0.009) -0.006 (0.004) -0.010 (0.007) -0.002 (0.0140) -0.018** (0.007) - FDI 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000*** (0.000) - - Nondurable Goods -0.019*** (0.004) -0.010** (0.004) - Durable Goods - - All - - Beverages & Tobacco Paper Chemicals Plastics & Rubbers Nonmetallic Minerals Primary Fabricated Metals Machinery 17 Transportation - - TOT 0.481 ( 0.184) 0.150 (0.788) -0.143 (0.397) -0.147** (0.629) 0.533 (0.734) 1.358 (1.039) 0.077 (0.188) 1.272*** (0.011) -0.203 (0.810) -0.151 (0.230) 0.305 (0.157) -0.022 (0.132) Exp/Imp -0.184** (0.043) - K/L - D1 - D2 - D3 - D4 - D5 - D6 - D7 - D8 - R2 0.779 F-stat 13.23 - - - - - - - - - 0.428 3.988 0.217 (0.187) - - - - - - - - - - 0.338 1.916 - - - - - - - - - 0.665 10.6 -0.202 (0.493) 0.500 (0.627) -0.412*** (0.147) -0.018 (0.154) 0.055 (0.645) -0.140** (0.055) -0.132*** (0.038 - - - - - - - - - 0.112 0.674 - - - - - - - - - 0.357 2.967 0.005** (0.001) - - - - - - - - - 0.565 6.926 - - - - - - - - 0.744 15.51 - - - - - - - - 0.827 17.91 -0.287*** (0.049) - -0.114** (0.046) - -0.132*** (0.039) - 0.193*** (0.036) - - - - - 0.722 34.2 - -0.277*** (0.032) 0.064 (0.035) -0.124*** (0.034) -0.279*** (0.025) -0.268*** (0.030) -0.114*** (0.024) -0.062 (0.034) 53.01 -0.267*** (0.031) -0.278*** (0.025) -0.320*** (0.031) 0.782 -0.438*** (0.042) 0.741 43.74 0.003** (0.001) 0.004*** (0.001) 0.005*** (0.000) 0.004*** (0.000) Table 5: Significance codes: 0.01: ***, 0.05: **, Note: for each sub-industry’s attributes, the first row indicates the coefficient and the second row indicates the standard error. Nondurable Goods refers to Food (D1), Beverages & Tobacco (D2), Paper (D3), Chemicals (D4), and Plastics & Rubbers (D5). Durable Goods refers to Nonmetallic Minerals (D6), Primary Fabricated Metals (D7), Machinery (D8), and Transportation (No dummy). Codes D1 - D8 refers to the dummy variables that control for industry fixed effects included in the regressions for Nondurable Goods, Durable Goods, and All. Variance Inflation Factors (VIF) can be calculated using the following equation: V IFk = 1 1 ≠ Rk2 (21) where the VIF for each explanatory variable k is the reciprocal of the inverse of the R2 from the regression. The algorithm I used calculates the VIF for each k with respect to all other variables in the model and removes variables that present a VIF Ø 5, a condition equivalent to an R2 Ø 0.8. A variable that has an R2 that is 0.8 shows strong evidence of multicollinearity and should be removed in order to preserve the OLS assumption of linear independence. VIF thresholds are normally between 5 and 10; I chose to use a less strict threshold of 5 given the small number of observations in the data set for each regression. All variables left have a VIF < 5. Finally, the regressions for Nondurable Goods, Durable Goods, and the overall manufacturing industry include dummy variables that control for industry fixed effects. The notes beneath Tables 4 and 5 provide the dummy variable codes. The results in Table 4 and 5 show that union wages and labor share are inversely correlated: wage increases correspond to labor share decreases, while labor share decreases also correspond to wage increases, though not as often. The labor share regression results in Table 4 indicate that for the manufacturing industry overall, the union-nonunion wage differential was a statistically significant explanatory variable. Though these results present correlated variables and not necessarily causal relationships, the theory of how these variables interact discussed in section 3 supports the conclusion that increasing union wages have a negative effect on union labor share. A union wage increase due to a union labor share decline, on the other hand, does not have the same basis in the theory I’ve discussed. Additionally, for the overall manufacturing industry labor share and wage regressions, Exp/Imp and K/L are the only indicators that are both statistically significant and have the correct signs on the coefficients. The significance of K/L is supported by the findings of Hirsch and Berger (1986), while the significance of international trade indicators such as Exp/Imp is not supported by the literature, specifically Baldwin (2004) and Slaughter (2007). There are many reasons for why my results may differ from those of previous studies, but one possibility is that my study uses more data than these previous studies. Slaughter’s study used a panel for 10 years, 1983 to 1994, while Baldwin’s study used data from only three years: 1977, 1987, and 1997. Given that my study utilizes a panel for the entire period between 1993 and 2012, my results are perhaps more indicative of long-term international trade effects than these previous studies. Another reason why my results may differ is that the time period of my study is more recent, so perhaps my results reflect that these international trade trends have become more prominent in recent years as the US manufacturing industry has become more globalized. In any case, I consider Exp/Imp to be a strong indicator of international trade trends. Both the wage and labor share regression results show that Exp/Imp never has a wrong sign on the coefficient when it is statistically significant. Based on the theory as established in section 3, when there is a wage increase, an influx of imports, which corresponds to an Exp/Imp decline, leads to a greater pool of resources and more possibilities for capital to substitute labor. The result is that a substitution effect leads the elasticity of labor demand to increase. This brings down marginal profits and creates a stricter profit boundary, which in turns leads to a decline in the union labor 18 share. In general, the union wage results in Table 5 reveal an opposite trend to the labor share results in Table 4, which was expected. Given the hypothesis that union wage and labor share have an inverse relationship, for the wage regression I expected all competitive effects to also have signs on the coefficients opposite of what they had for the labor share regression. This is true for Exp/Imp and K/L, but not TOT, which has a negative sign for both the labor share and wage regressions, while I expected a negative sign for the labor share regression and a positive sign for the wage regression. It is important to note that the inconsistency of TOT’s coefficient sign does actually fall in line with the findings of Baldwin (2004) and Slaughter (2007) for international trade indicators. Despite the inconsistency of TOT’s coefficient sign, there is a case when it proves to be a useful indicator. Table 4 shows that the statistically significant negative TOT effect that appears for the Nondurable Goods segment is primarily driven by one sub-industry, Paper. I did additional research to find out if this effect was supported by empirical findings. According to the NGO World Growth, the Paper industry faced an import price crisis in the 2000’s that led to import sanctions in 2009 (Shapiro 2011). A World Growth report mentions a survey conducted by the International Trade Commission (ITC), which concluded that lower import prices have increasingly led US businesses to purchase paper imports, in particular from China and Indonesia. According to the ITC, price was the only factor for which US manufacturers were inferior to manufacturers abroad, but it was often the most determining factor when purchasing decisions were made. The theory is that low import prices decreased the profit margin of the Paper industry, allowing for the substitution and scale effects to increase the elasticity of labor demand. Given a wage increase, firms were more likely to substitute in capital and cheaper labor for the union labor, diminishing the elasticity of labor demand. As elasticity of labor demand diminished, profits in the Paper industry fell, and as a result, unions in the Paper industry brought down the union labor share to meet firm profit constraints. Another notable trend is that FDI and K/L are never present in the same regression because they are highly multicollinear. Unlike the results of Slaughter (2007), my union labor share regression results do not show FDI to be a statistically significant indicator of the decline in labor share for the manufacturing industry overall, though FDI is a significant indicator for some Nondurable Goods industries. Meanwhile, Table 5 shows that K/L is a statistically significant indicator for labor share and wage. Though my results do not support the results of Slaughter (2007), they support Slaughter’s underlying theory that declines in the union labor share are explained by resource substitution that occurs when wages rise. The results in Table 5 show that increases in the union wage are in part explained by increases in K/L, an indicator that can also be directly linked to the substitution effect. In particular, there is a 0.4% increase in the union wage for every 1-unit increase in K/L. However, the theory as outlined by Griswold (2010) and Slaughter (2007) suggests that the causal relationship goes in the opposite direction given a 0.4% increase in the union wage, there is 1-unit increase in K/L. This increase in K/L leads to an increase in the elasticity of labor demand via the substitution effect, as is the case for FDI. Additionally, the interchangeability of FDI and K/L suggests that capital replaces labor domestically and internationally in the same way for the manufacturing industry, since FDI is an indicator of international capital substitution, and K/L is an indicator 19 of domestic capital substitution, though possibly also international capital substitution. Given the existence of multinational corporations in many of these manufacturing subindustries, it is perhaps the case that K/L accounts for international capital substitution as well. 7 Conclusions The aim of this paper was to determine whether competitive changes have had a significant effect on the decline of union membership over time despite the persistence of a positive union-nonunion wage differential. My regression results show that an increasingly positive wage differential is a significant determinant of union membership’s decline. In addition, union utility models indicate that union wages and labor share are determined when unions and firms bargain subject to the firm’s profit constraints, and the union therefore faces a wage-labor trade-off that is subject to the conditions that affect firm profits. One of the forces that drives firm profits is competition. For the manufacturing industry overall, my results support the theory that competition, in the form of low cost and high volume imports as well as foreign and domestic capital substitution, has the effect of diminishing union membership. However, these competitive effects alone do not diminish union membership - instead, the theory suggests that the underlying cause of this competitive effect is a wage increase. Given a wage increase, competition causes an increase in the elasticity of labor, and thus, the sensitivity of union labor to replacement by less costly resources. For both the Durable Goods and Nondurable Goods industries, it is the combination of a wage rise, high import volume, and substitution of domestic (and possibly international) capital for union labor that diminishes union membership. In summary, the push for higher union wages often initiated by unions brings with it a trade-off in union membership. A wage increase in competitive conditions leads to an erosion of firm profits, and when this occurs, firms tighten profit constraints in union-firm negotiations, leaving less room for unions to maximize union membership. The end result is that the amount of union membership that unions and firms agree to diminishes. References [1] R. Baldwin, (2004). The Decline of U.S. Labor Unions and the Role of Trade, Journal of Economic Issues. 38(4), 1087-1091. [2] J. Brown, and O. Ashenfelter, (1986). Testing the Efficiency of Employment Contracts, Journal of Political Economy, 94, 40-87. [3] I. Brambilla, A. Khandelwal, and P. Schott, (2010). China’s Experience Under the Multifiber Arrangement (MFA) and the Agreement on Textile and Clothing (ATC), NBER Chapters, in: China’s Growing Role in World Trade, 345-387. [4] Bureau of Labor Statistics, (2015). Union Membership (Annual) News Release, USDL-15-0072. 20 [5] J. Burke, G. Epstein, and M. Choi, (2004). Rising Foreign Outsourcing and Employment Losses in U.S. Manufacturing, 1987-2002, Political Economy Research Institute, Working Paper Series Number 89. [6] R. Caves, and M. Krepps, (1993). Fat: The Displacement of Nonproduction Workers from U.S. Manufacturing Industries, Brookings Papers on Economic Activity, Microeconomics, 1993(2), 227-288. [7] H. Farber, (2001). Notes on the Economics of Labor Unions, Working Paper No. 452: Industrial Relations Section. Princeton University, [8] H. Farber, (1983). The Determination of the Union Status of Workers, Econometrica, 51(5), 1417-1437. [9] D. Griswold, (2010). Unions, Protectionism, and U.S. Competitiveness, Cato Journal, 30(1), 181-196. [10] B. Hirsch, (2008). Sluggish Institutions in a Dynamic World: Can Unions and Industrial Competition Coexist?, Journal of Economic Perspectives, 22(1), 153-176. [11] B. Hirsch, and M. Berger, (1984). Union Membership Determination and Industry Characteristics, Southern Economic Journal, 50, 665-679. [12] B. Hirsch, and D. Macpherson, (2009). Union Membership and Coverage Database from the Current Population Survey, Data Series IV: Industry: Union Membership, Coverage, Density, and Employment by Occupation, 1983–2014, http://unionstats.gsu.edu. [13] L. Lee, (1978). Unionism and Wage Rates: A Simultaneous Equations Model with Qualitative and Limited Dependent Variables, International Economic Review, 19, 415-443. [14] T. MaCurdy, and J. Pencavel, (1986). Testing between Competing Models of Wage and Employment Determination in Unionized Labor Markets, Journal of Political Economy, 94, 3-39. [15] D. Rodrik, (1997). Has Globalization Gone Too Far?, for International Economics. Washington, DC: Institute [16] K. Scheve, and M. Slaughter, (2004). Economic Insecurity and the Globalization of Production, American Journal of Political Science, 48(4), 662-674. [17] P. Schott, (2010). U.S. Manufacturing Exports and Imports by SIC or NAICS Category and Partner Country, 1972 to 2005, Yale School of Management & NBER, 1-4. [18] R. Shapiro, (2011). The Economic Impact of U.S. Trade Sanctions on Imports of Paper Products, Sonecon LLC, Commissioned by World Growth, 1-42. [19] M. Slaughter, (2007). Globalization and Declining Unionization in the United States, Industrial Relations, 46(2), 329-346. 21 [20] M. Slaughter, (2001). International Trade and Labor-Demand Elasticities, Journal of International Economics, 54(1), 27-56. 22 A Appendix 23 Figure 2: Trends in union membership, wage differentials, capital-to-labor ratio, and FDI from 1993 - 2012. 24 Figure 3: Trends in terms of trade and the export-import volume ratio from 1993 - 2012.
© Copyright 2026 Paperzz