Fixed or open-ended? Labor contract and productivity in the Colombian manufacturing sector

Castellani, Francesca; Lotti, Giulia; Obando, Nataly Article Fixed or open-ended? Labor contract and productivity in the Colombian manufacturing sector Journal of Applied Economics Provided in Cooperation with: University of CEMA, Buenos Aires Suggested Citation: Castellani, Francesca; Lotti, Giulia; Obando, Nataly (2020) : Fixed or openended? Labor contract and productivity in the Colombian manufacturing sector, Journal of Applied Economics, ISSN 1667-6726, Taylor & Francis, Abingdon, Vol. 23, Iss. 1, pp. 199-223, https://doi.org/10.1080/15140326.2020.1729582 This Version is available at: https://hdl.handle.net/10419/314088 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Journal of Applied Economics ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/recs20 Fixed or open-ended? Labor contract and productivity in the Colombian manufacturing sector Francesca Castellani, Giulia Lotti & Nataly Obando To cite this article: Francesca Castellani, Giulia Lotti & Nataly Obando (2020) Fixed or openended? Labor contract and productivity in the Colombian manufacturing sector, Journal of Applied Economics, 23:1, 199-223, DOI: 10.1080/15140326.2020.1729582 To link to this article: https://doi.org/10.1080/15140326.2020.1729582 © 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. Published online: 27 Feb 2020. Submit your article to this journal Article views: 1840 View related articles View Crossmark data Citing articles: 3 View citing articles Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=recs20 RESEARCH ARTICLE Fixed or open-ended? Labor contract and productivity in the Colombian manufacturing sector Francesca Castellani a , Giulia Lotti b and Nataly Obando c a Office of the Vice President for Countries, Inter-American Development Bank, Washington, DC, USA; b Office of Strategic Planning and Development Effectiveness, CAGE Warwick and Inter-American Development Bank, Washington, DC, USA; c Educational Psychology PhD program, University of Illinois at Chicago, Illinois, USA ABSTRACT Using the Colombian Annual Manufacturing Survey (AMS) between 2000 and 2014, this paper investigates the effect of labor contract modalities on firm productivity within the industrial sector through a structural model. We find that the elasticity of revenue with respect to temporary workers is often lower than that with respect to permanent workers, but not in small firms, where some forms of flexibility might lead to productivity gains. ARTICLE HISTORY Received 30 November 2018 Accepted 4 February 2020 KEYWORDS Labor contracts; temporary workers; skilled workers; firms’TFP 1. Introduction Over the past few years, as a result of the waning commodity boom, Latin America (LA) has been grappling with the puzzle of growth. Productivity has not played a remarkable role in this quest for stronger economic performance and has contributed instead to the gap with more advanced economies. On average, LA labor productivity is less than 30% of US. 1 The growth of productivity in the economy as a whole depends not only on how fast productivity is rising in individual industries, but also on whether resources are moved in or out of those industries in response. Labor allocation and labor markets play an important role in it. Given the surge in temporary (fixed-term) modalities of contracting in advanced and emerging economies since the 1980s (Cazes & Laiglesia, 2015), the literature has investigated the effect of contractual terms on productivity growth. This paper explores the relationship between labor contracts and firm productivity in the Colombian industrial sector, using the information of the Annual Manufacturing Survey (AMS). In Colombia (Figure 1, panel a), total employment in manufacturing passed from around 485,000 in 2000 to almost 584,000 in 2014. 2 Both workers with permanent (open-ended) contracts and workers with temporary (fixed term) contracts increased, but the share of temporary did so to a larger extent (Figure 1, panel b): in 2000, 25% of workers had temporary contracts visa vis 35% in 2014. When all economic sectors are considered, Colombia exhibits one of the highest shares of temporary workers CONTACT Giulia Lotti [email protected] CAGE Warwick, Inter-American Development Bank 1 Authors’calculations based on The Conference Board Total Economy Database, 2015. 2 These figures refer to total employment in firms with more than 10 employees only. JOURNAL OF APPLIED ECONOMICS 2020, VOL. 23, NO. 1, 199–223 https://doi.org/10.1080/15140326.2020.1729582 © 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. if compared to the rest of Latin America and the Caribbean 3 and OECD countries. 4 This surge in temporary employment, often considered a way to curb high labor costs due to the rigidities of labor market regulations, has taken place across all subsectors within manufacturing, except basic metals (Figure 2). While most firms in Colombia are small or medium, the large ones generate most of the employment and permanent positions. In 2014, 42% of permanent workers were hired by large plants, 38% by medium ones, 20% by small (Figure 3, panel a). 5 However, the larger the plant, the smaller is the share of its permanent workers: 54% of the labor force has a permanent contract in large firms, with respect to 61% in medium, and 70% in small ones (Figure 3, panel b). Although large firms contribute the most to permanent positions, they hold lower shares of permanent workers. The percentage of plants hiring temporary workers also increases by plant size, and export status. 6 Permanent workers are more likely to be skilled 7 and larger plants exhibit slightly higher shares of skilled workers. Figure 4 depicts the share of skilled (vs. unskilled) workers by contract choice (panel A), by plant size (panel B), and by industry subsector (panel C). The comparison between Figures 2 and 4shows that the subsectors with the highest share of skilled workers (petroleum products, chemicals, medical equipment) are also the ones with the highest proportion of permanent contracts, while the subsectors with the lowest shares (apparel, leather, textiles) are ones with the lowest proportion of skilled workers. This is in line with the hypothesis that, when technologies require a more skilled labor force, firms tend to hire on a longer-term basis as they value their workforce more (Eslava, Haltiwanger, Kugler, &Kugler,2014). (a) (b) Figure 1. Total manufacturing employment. (a) Type of labor contract (b) Share of temporary contracts. Source: Dane-AMS; authors’calculations 3 Source: national household surveys, 2015. 4 Source: OECD Data, 2015. 5 Firms are classified by size in small (between 10 and 49 employees), medium (between 50 and 200 employees), and large (more than 200 employees). 6 94% of large firms has at least one worker with a temporary contract vs. 75% of medium-size firms, and 49% of small. Exporting firms are also more likely to be hiring temporary workers (78% of exporters hire temporary workers vs. 59% of non-exporters in 2014). Moreover, they tend to be larger in size, and with higher shares of temporary workers (37% of workers are temporary in exporting firms vs. 34% in non-exporters in 2014). 7 44% of workers with permanent contracts is skilled vs. 28% of temporary workers. 200 F. CASTELLANI ET AL. The reduction in the share of permanent workers between 2000 and 2014 may bear consequences for firm growth potential. The type of labor contract can affect workers’ efforts through different channels. If a worker expects to stay in a workplace for a short period, her willingness to exert efforts might diminish, which in turn can negatively affect the firm’s productivity. Moreover, the high turnover encouraged by temporary contracts (a) (b) Figure 3. Permanent contracts by plant size. (a) Number of permanent workers (b) % Permanent contracts. Source: Dane-AMS; authors’calculations. Figure 2. Share of temporary contracts by subsector in 2000 and 2014. Source: Dane-AMS; authors’calculations. JOURNAL OF APPLIED ECONOMICS 201 may limit incentives for job trainings, further harming productivity. On the other hand, a temporary worker might exert more effort to be rewarded with a contract renewal or with a permanent contract. Temporary contracts might also have a positive effect on productivity as they allow for more flexibility, facilitating firms’allocative efficiency (hiring the right worker at the right time) and allowing for an easier workforce adjustment when needed. In a nutshell, the labor contract choice can affect productivity in either direction. As we can see from Table 1, temporary workers are more likely to be female, but significantly more so in larger firms only; they are also younger than permanent ones, irrespective of firm size, even though the gap in average age is greater in medium/large (a) (b) (c) Figure 4. Percentage of skilled and unskilled worker in 2014. (a) by contract choice (b) by plant size (c) by subsector. Source: Dane-AMS; authors’calculations. 202 F. CASTELLANI ET AL. Table 1. Characteristics of workers by contract type and firm size. Small Medium/Large Temporary Permanent Difference in Means S.E. Temporary Permanent Difference in Means S.E. Age 33.22 36.01 −2.797*** 0.861 32.42 36.66 −4.241*** 0.614 Female 40.50% 37.20% 3.2 4.280 40.40% 35.00% 5.4* 3.160 Education Years of Education 10.13 10.97 −0.84** 0.346 11.23 11.93 −0.69*** 0.198 No Education 0.45% 0.30% 0.2 0.350 0.02% 0.60% −0.6** 0.271 Primary Incomplete 4.11% 4.27% −0.2 1.720 1.28% 2.31% −1.0 0.780 Primary Complete 9.13% 8.07% 1.06 2.490 5.76% 4.59% 1.17 1.530 Secundary Incomplete 24.10% 17.90% 6.2* 3.530 11.00% 9.04% 2 1.890 Secundary Complete 35.40% 31.40% 3.9 3.910 44.10% 35.70% 8.3*** 3.180 Tertiary Incomplete 6.73% 7.33% −0.6 2.470 11.10% 7.65% 3.4* 1.890 Tertiary Complete 18.30% 29.60% −1.13*** 3.660 24.90% 38.70% −1.39*** 2.880 Occupation: Vocational/Technician 6.02% 12.00% −6.0** 2.370 12.40% 14.60% −2.2 2.090 Director or Senior Officer 4.45% 5.11% −0.7 2.170 1.40% 6.13% −4.7*** 1.180 Administrative or intermediate level 10.20% 13.50% −3.3 2.910 15.20% 15.00% 0.2 2.430 Retailers and Sellers 7.44% 6.17% 1.27 1.910 9.20% 12.50% −3.3* 1.810 In Services 4.48% 4.67% −0.2 1.740 8.53% 6.70% 1.82 1.930 Farm Workers 0.08% 0.17% −0.1 0.190 1.12% 2.31% −1.2 0.879 Blue-Collar (outside agriculture) 66.80% 58.30% 8.5** 4.170 52.00% 41.70% 10.3*** 3.210 Other 0.52% 0.00% 0.5* 0.303 0.13% 1.04% −0.9 0.673 Formal 54.40% 93.80% −3.94*** 3.310 89.60% 98.10% −8.4*** 1.840 Part time 1.52% 0.05% 1.5** 0.701 0.28% 0.29% −0.00 0.282 Wage 865,656 1,395,000 −529,523*** 176,055 1,017,000 1,735,000 −718,009*** 114,893 Source: Colombia Great Integrated Household Survey 2014; authors’calculations. Notes: *p < 0.1, **p < 0.05, ***p < 0.01. JOURNAL OF APPLIED ECONOMICS 203 firms. 8 Besides being younger and female, temporary workers are also characterized by fewer years of education. Most of the workers in manufacturing are blue-collar and most of the bluecollar workers are hired through temporary contracts, across all firm sizes. A striking difference is the proportion of formal jobs among temporary workers in small firms, where only half are affiliated to or pay social security contributions. In larger firms, they are approximately 90%. Moreover, temporary workers are remunerated less than permanent ones, and the gap is larger in medium and large firms. 9 In this paper, we investigate how hiring through temporary contracts affects plant productivity. By applying Addessi’s(2014)model,wefind that the elasticity of value added of temporary workers in the manufacturing sector is lower than that of permanent workers. This holds true in exporting and non-exporting plants, in medium and large plants, and in different geographical areas, and in most of the subsectors. However, in small plants and in a few subsectors, the elasticity of value added of temporary and permanent workers are not so different. A higher share of temporary workers in small size firmsisevenbeneficial to TFP withalag,whiletheoppositeistrueforlargerfirms. Our results are not driven by the difference in the skill composition of temporary and permanent workers, but by the contract modality itself. The paper develops in seven sections beyond this introduction. The second section offers a brief literature review. The third presents the institutional background. The fourth provides the data description. The fifth explains the empirical model, and the sixth describes the results. The seventh section concludes. The Appendix contains detailed information on the data cleaning process and definitions of variables. 2. Literature review Much of the literature on the relationship among labor markets and productivity relates to employment protection legislation (EPL) 10 and to OECD countries, 11 without being conclusive on the effect of contract modalities on productivity, nor providing much evidence beyond the OECD. Generally, research acknowledges that different labor contracts might imply different levels of productivity and finds that the use of, or the regulations related to the use of, temporary contracts can have negative or non-significant effects on firms’productivity (Addessi, 2014; 12 Bassanini, Nunziata, & Venn, 2009; Aguirregabiria & Alonso-Borrego, 8 Source: Colombian Great Integrated Household Survey (Gran Encuesta Integrada de Hogares). We limit the sample to wage workers in the manufacturing sector between 15 and 64 years old in 2014. Given the way the survey was implemented, we are unable to distinguish between medium and large firms, so we will group the two categories together. 9 The difference in pay could be a factor demotivating effort, hence the lower productivity level, but we cannot rule out the reverse causality either, i.e., that precisely because these workers are less productive, they are compensated with lower wages. 10 The indicators of employment protection legislation (EPL) outline conditions under which workers can be hired on both regular and temporary contracts. 11 The Employment Protection Legislation (EPL) methodology developed by the Organization for Economic Cooperation and Development (OECD) measures the degree of stringency of employment protection legislation of OECD and G201 countries, to determine, amongst others, its labor market impacts. The Inter-American Development Bank (IDB), in collaboration with the OECD, has applied this methodology to Latin American and Caribbean (LAC) countries. 12 Addessi (2014) develops a structural model that allows for the labor-contract choice to influence labor-augmenting factors and capture the effects on the evolution of the firm productivity process and finds the use of temporary contracts to be detrimental to TFP growth. 204 F. CASTELLANI ET AL. 2014; Damiani & Pompei, 2010; 13 Cappellari, Dell’Aringa, & Leonardi, 2012; 14 JonaLasinio & Vallanti, 2011; Nielen & Schiersch, 2016). 15 Dolado and Stucchi (2014)find that the effect of temporary contracts on TFP growth is negative when the conversion rate from temporary to permanent contracts is low, because the effort provided by temporary workers diminishes. Using a natural experiment Autor and Houseman (2010) show that temporary-help positions reduce subsequent earnings and employment outcomes in the United States. With an eye to European labor markets, Blanchard and Landier (2002) and Cahuc and Postel-Vinay (2002) model temporary contracts as churning policies that affect negatively wage setting and may generate higher unemployment and lower productivity. According to Boeri and Garibaldi (2007), if fixed-term contracts are used as a bufferstock to boost the number of hiring in a boom, employment may go up at least temporarily, but they find the use of fixed-term contracts in Italy to be associated with lower average productivity. The model by Cahuc, Cheron, and Mahlebert (2016) shows that the substitution of temporary for permanent positions induced by job protection does not change much total employment but reduces aggregate productivity. However, according to Ichino, Mealli, and Nannicini (2008) temporary contracts can be good screening devices and stepping-stones into permanent posts and therefore increase productivity. Giuliano, Kampelmann, Mahy, and Rycx (2017), while not finding any significant effect of hiring through fixed-term contracts on firm competitiveness in the Belgian manufacturing sector, observe a positive and significant effect on labor productivity in services, which they attribute to some service subsectors being more labor intensive, in need of less sophisticated technology and less able to create stocks. Finally, Eslava et al. (2014)find that in Colombia the use of temporary contracts is associated with higher firms’productivity, particularly in businesses with lower skill needs. In the latter, a more intensive use of fixed-term contracts plays as a shock absorber and leads to productivity gains. However, using fixed-term employment more intensively has a negative effect on TFP for the more high-skill intensity plants. 3. Institutional background Colombia is characterized by a dual labor market, where workers are entitled to different employment protection depending on the contract they hold. Fixed-term contracts must be written and can last for a maximum duration of three years. However, they can be renewed indefinitely if they are of the duration of one-year (Articles 45 and 46 of the Labor Code). Open-ended contracts instead have no time limit. 13 Damiani and Pompei (2010) analyze the effect of labor protection on Total Factor Productivity (TFP) growth in 18 European countries between 1995 and 2005. They also control for the effect of growth in temporary employment on TFP, ending a negative and significant relation. 14 Cappellari et al. (2012) use 13,000 firm level observations of all Italian sectors between 2004 and 2007 to study the effects of deregulation reforms of apprenticeship and fixed-term contract. They find a small negative, but only weakly significant, effect of fixed-term employment on labor productivity, driven by the reduced job tenure and increased use of external collaborators that hampered the optimal allocation of resources. 15 Bassanini et al. (2009) measure productivity as TFP and find no evidence of a productivity effect of facilitating the use of fixed-term contracts in the OECD; Damiani and Pompei (2010)find that fixed-term contracts can have detrimental effects on TFP growth; Jona-Lasinio and Vallanti (2011), Aguirregabiria and Alonso-Borrego (2014), and Nielen and Schiersch (2016) measure productivity as labor productivity. JOURNAL OF APPLIED ECONOMICS 205 Bench1, X is not instrumented since there is no reason to expect correlation between the composition of the labor contracts at time t-1 and the productivity shock at time t, while in Bench2 X is instrumented with its lagged value. Table 2 shows the results for the entire sample and by plant size. Table 3 divides plants between exporters and non-exporters, while Table 4 presents the results by geographical area. As we can see in column (1) of Panel A (Table 2), when estimating an OLS regression in the manufacturing sample, the elasticities of value added with respect to capital and labor services, βkand βl, are 0.308 and 0.831. Moving beyond the OLS procedure, we need to guess an initial value for the parameters to be estimated; our guess is 0.4 for βk, 0.6 for βl, 1 for s(i.e., temporary and permanent contracts are characterized by the same labor-augmenting factor) and 0 for γ (i.e., labor-contract composition has no effect on the dynamic generation process of TFP). When we split labor into temporary and permanent workers and run the 2-steps estimation (column 2), βkand βlare 0.085 (s.e. 0.005) and 0.986 (s.e. 0.009), while sis Table 5. Results on skilled vs unskilled in the aggregate and by plant size. Panel A. Aggregate Panel B. Small Size 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 (1) (2) (3) (1) (2) (3) β perm,sk 0.986*** 0.986*** 0.986*** 0.987*** 0.989*** 0.989*** (0.009) (0.009) (0.009) (0.010) (0.010) (0.010) s temp,sk 0.760*** 0.754*** 0.758*** 0.918*** 0.911*** 0.917*** (0.022) (0.021) (0.022) (0.028) (0.028) (0.028) s perm,unsk 0.914*** 0.907*** 0.917*** 0.942*** 0.936*** 0.947*** (0.016) (0.016) (0.016) (0.021) (0.020) (0.021) s temp,unsk 0.788*** 0.786*** 0.790*** 0.898*** 0.896*** 0.901*** (0.018) (0.018) (0.018) (0.027) (0.026) (0.026) γ skilled 1.601 0.032*** 1.251 0.031*** (2.086) (0.008) (2.181) (0.006) γ unskilled 1.19 0.004 1.211 −0.005 (2.086) (0.008) (2.180) (0.007) Observations 57,967 57,967 57,967 29,551 29,551 29,551 Panel C. Medium Size Panel D. Large Size 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 (1) (2) (3) (1) (2) (3) β perm,sk 1.009*** 1.010*** 1.009*** 0.926*** 0.926*** 0.926*** (0.012) (0.012) (0.012) (0.023) (0.023) (0.023) s temp,sk 0.669*** 0.666*** 0.668*** 0.486*** 0.483*** 0.485*** (0.033) (0.033) (0.032) (0.056) (0.056) (0.056) s perm,unsk 0.868*** 0.863*** 0.869*** 0.978*** 0.972*** 0.978*** (0.031) (0.030) (0.031) (0.088) (0.087) (0.089) s temp,unsk 0.720*** 0.719*** 0.721*** 0.499*** 0.498*** 0.499*** (0.029) (0.029) (0.029) (0.055) (0.054) (0.055) γ skilled 1.653 0.039*** 2.55 0.007 (2.556) (0.014) (5.453) (0.064) γ unskilled 1.654 0.024 2.532 0.002 (2.555) (0.015) (5.455) (0.060) Observations 21,235 21,235 21,235 7,181 7,181 7,181 Notes: Bootstrapped standard errors in parentheses, ***p < 0.01, **p < 0.05, *p < 0.1 212 F. CASTELLANI ET AL. 0.831 (s.e. 0.012), significantly different from 1, suggesting that labor contracts induce a static difference in labor productivity. Unlike in Addessi (2014), introducing the labor-contract composition in the TFP dynamics does not change the estimated coefficients in our setting. When we estimate the model through Bench1,wefind that sis 0.832 (s.e. 0.012), significant at 1% level (column 3). The finding is confirmed under Bench2 estimation (column 4). 26 This means that the elasticity of production to temporary workers is lower than that of permanent workers. The estimated γaccording to the Bench1 estimation is −0.007 (s.e. 0.003), significantly different from zero at 5% level, but according to Bench2 it is not significantly different from zero. This signals that the share of permanent workers in a plant has a negative effect on TFP with a lag, if any. An analysis of the heterogenous effects by size presented in Panels B-D of Table 2 sheds some light on these results. First, we find that the larger the plant, the lower the parameter s. While the elasticity of revenue with respect to temporary workers in small plants is just slightly smaller than that of permanent workers (sis just a bit smaller 1, but the difference is significant), in mediumsize plants it is lower than that of permanent workers, and even lower in large plants. According to Bench1 estimations (column 3, Panel B, Table 2), sis 0.945 (s.e. 0.016), 0.775 (s.e. 0.015) and 0.495 (s.e. 0.037) for small, medium, and large companies, respectively. The coefficient γindicates that the incidence of permanent contracts in total contracts has a negative effect on TFP with a lag in small size firms, but a positive effect in larger firms. We have shown in the previous section that plants hire sizable proportions of temporary workers, and this seems to affect negatively both production and TFP with a lag, except for small firms, where there seems to be a positive effect of flexibility. This result is in line with the idea that the flexibility of temporary contracts may lead to within-firm productivity gains in firms with low sophistication, which may well be proxied for by a small size. In the case of both exporters and non-exporters, the elasticity of revenue with respect to temporary workers is lower than that of permanent workers, with the gap being larger in exporting plants (Table 3): according to Bench1 model, the estimation of sis 0.861 (s.e. 0.013) for non-exporters and 0.715 (s.e. 0.023) for exporters. In TFP dynamics we do not observe an effect from a higher share of permanent contracts in exporters. As for nonexporters, an increase of one unit in the share of permanent contracts might have a negative effect on TFP with a lag, but the result is not very robust. We also examine whether temporary workers are characterized by a different level of productivity depending on the geographical area where plants operate. To do so, we divide the sample into two groups: plants operating in the most populous departments, where we assume greater possibilities of networking among plants (Bogota D.C, Antioquia and Valle del Cauca), and plants operating in the remaining ones. In both areas, the elasticity of value added with respect to temporary workers is lower than that of permanent workers: sis 0.934 (s.e. 0.021) in the most densely populated areas and 0.872 (s.e. 0.010) in the remaining areas (Table 4); γis not significantly different from zero for more densely populated areas (−0.001, s.e. 0.003) and negative for less densely populated ones (−0.004, s.e. 0.002). However, the estimated coefficient for less densely populated 26 According to Bench2 estimates, sis 0.831 (s.e. 0.012). JOURNAL OF APPLIED ECONOMICS 213 areas is only marginally significant and the finding is not very robust to using a different approach (i.e., Bench2). Overall, it seems that the elasticity of revenue with respect to temporary workers is lower (or in the best-case scenario not significantly different) compared to that of permanent workers. To assess whether this result is driven by the lower shares of skilled workers among the temporary, we estimate Equation (11). Results are presented in Table 5. In the aggregate manufacturing sector, we find that the elasticity of revenue of permanent unskilled workers is slightly lower than that of permanent skilled workers (sperm;unsk = 0.907, s.e. 0.016), but it is mainly the elasticity of skilled and unskilled temporary workers that is much lower than that of permanent workers (stemp;sk = 0.754, s.e. 0.021; stemp;unsk = 0.786, s.e. 0.018). Let us recall that we found that temporary workers are almost as productive as permanent workers in small plants. When we disentangle workers by skill levels too, we find that unskilled workers with temporary or permanent contracts are equally productive: sperm;unsk is 0.936 (s.e. 0.020), and stemp;unsk is 0.896 (s.e. 0.026), even though both of the estimated parameters are significantly different from 1. stemp;sk is 0.911 (s.e. 0.028), which is also significantly different from 1, 27 suggesting that within skilled workers in small plants temporary contracts can be detrimental. However, the majority of workers in small plants are unskilled, so the latter does not have much weight overall. In medium and large plants, we find that temporary workers increase value added to an even lower extent than permanent:stemp;sk = 0.666, s.e. 0.033; stemp;unsk = 0.719, s.e. 0.029 in medium firms and stemp;sk = 0.483, s.e. 0.056; stemp;unsk = 0.498, s.e. 0.054 in large firms. In sum, in medium and large firms the elasticity of value added of temporary workers, either skilled or unskilled, is always lower than that of permanent workers. In a nutshell, when the elasticity of value added of temporary workers is lower than that of permanent workers, this is not driven by their skill composition. 28 Finally, when we allow the production function to differ across subsectors (Tables A1– A3), estimations differ depending on the methodology adopted (LP/OP), which makes it hard to draw robust conclusions. 6.1. Robustness checks First, it is reassuring to note that Bench1 and Bench2 estimates are almost identical. Our estimates are robust to changing the degree of the polynomial used in the estimation of the first and third steps, to changing the propensity score, to calculating the first stage distinguishing by skills, to changing the number of replications used for bootstrapping errors, to using consumption of all intermediate good rather than electricity only, to varying the initial parameters’values assigned in non-linear squares estimations. The same applies for the estimate of the sand γparameters when we re-conduct all the analysis through OP (Tables A4–A6). Only the findings across subsectors are not robust, hence we do not rely on them. 27 The p-value of the difference between sts and 1 is 0.051. 28 γskilled is sometimes significantly different from zero, but the estimates are not very robust. 214 F. CASTELLANI ET AL. 7. Conclusion The rise in temporary contracts in Colombia raises questions on its consequences on firm productivity. To estimate the effects of hiring workers through temporary rather than open-ended contracts, we follow Addessi (2014) and apply a Levinsohn and Petrin (2003) structural model in three steps, including the Ackerberg et al. (2016) correction and endogenizing TFP. We also use the Olley and Pakes (1996) approach. We find that in general the elasticity of value added with respect to temporary workers in the manufacturing sector in 2000–14 is lower than that of permanent workers. However, when exploring the effects of the contract mix on TFP dynamics, we find that a higher share of permanent contracts can have a negative influence on TFP with a lag. Analyzing results by size sheds some light: while in larger firms the elasticity of value added is larger for permanent workers, who also seem to have a positive effect on TFP with a lag, this is not true for small firms, which represent the most numerous group of firms in our setting. One possibility might be that for small firm some flexible forms of hiring can be important. Results also present some heterogeneity by exporting status or geographical area. We analyze whether the lower productivity of fixed-term contracts is driven by a higher proportion of unskilled workers holding those contracts and do not find support for this thesis as results seem to be driven mostly by the contract type itself. Even though we are unable to measure the temporary-to-permanent conversion rate, the large share of temporary workers that prevails in Colombia might suggest that temporary contracts are a dead-end cost-reduction device rather than a steppingstone for workers to become permanent. This can have some negative effects on productivity, depending on firm characteristics. In the OECD, there have been proposals to replace dual labor markets with a single open-ended contract, 29 where termination costs that increase smoothly with job tenure allow the gains of both fixed-term and permanent contracts (Bentolila, Boeri, & Cahuc, 2012; Blanchard & Tirole, 2003; Boeri & Garibaldi, 2008; Cahuc, 2012; Dolado, 2012; IMF, 2014,2015,2017; OECD, 2014). Due to the low degree of employment protection in the initial period, this instrument can be used as a screening device, but as employment protection becomes more stringent with job tenure, it also serves as an incentive for plants to invest in training for all workers and search for the right matches. Given the similar EPL that characterizes Colombia, this single open-ended contract could be a tool to fight a segmented labor market that in some cases can slow productivity down. At the same time, fixed-term contracts should not be eliminated, which might have other unintended consequences, as flexible forms of contract can also allow less employable people to be employed more easily, and allow aflexible flow of resources from low productivity uses to high productivity ones. Acknowledgments We thank Marta Ruiz Arranz, Regional Economic Advisor at the Inter-American Development Bank for suggestions and for making the project possible. We are also indebted to William Addessi, Marcela Eslava, Alessandro Iaria, Carlos Ospino, Rodolfo Stucchi, Germán Coloma (the editor) and an anonymous referee for very useful comments and advises that greatly improved the paper. We also thank the National Department of Statistics of Colombia, DANE, that provided 29 In Spain, Italy, and France. JOURNAL OF APPLIED ECONOMICS 215 us with the access to the data. The content and findings of this paper reflect the opinions of the authors and not those of the Inter-American Development Bank. Corresponding author: Giulia Lotti (email: [email protected]). Disclosure statement No potential conflict of interest was reported by the authors. Notes on contributors Francesca Castellani is Principal Economist in the Office of the Vice President for Countries of the Inter-American Development Bank in Washington DC. Before joining the Vice President’soffice, she was a Lead Economist in the Andean Group Department in Washington DC, and Country Economist in the IDB Office in Bogotá. Francesca worked at the OECD in Paris, the World Bank in Washington DC and UNCTAD in Geneva, Switzerland. She is an economist from Bocconi University in Milan with a master's degree and a doctorate in economics from the University of Geneva, Switzerland. Giulia Lotti is an Economist at the Inter-American Development Bank (IDB) and Research Associate at the Centre for Competitive Advantage in the Global Economy (CAGE). She previously worked on labor markets of the Andean region for the IDB and as a consultant for the World Bank. Her research interests include applied microeconomics, labor economics and international finance. Giulia holds a PhD degree in Economics from the University of Warwick and a Master of Science in Economics from University College London (UCL). Nataly Obando is a second-year doctoral student in The University of Illinois at Chicago, in the Educational Psychology program. Before she worked as a consultant for the Colombian government, Inter-American Development Bank (IDB), and the Organization of Ibero-American States (OEI). She is an economist from The National University in Colombia with a master’s degree in economics from Our Lady of the Rosary University in Colombia, and a master’s degree in Education at The University of La Rioja in Spain. ORCID Giulia Lotti http://orcid.org/0000-0002-8675-502X References Ackerberg, D., Caves, D., & Frazer, G. (2006). Structural identification of production functions (MPRA Paper 38349). University Library of Munich. Addessi, W. (2014). The productivity effect of permanent and temporary labor contracts in the Italian manufacturing sector. Economic Modelling,36, 666–672. Aguirregabiria, V., & Alonso-Borrego, C. (2014). Labor contracts and flexibility: Evidence from a labor market reform in Spain. Economic Inquiry,522, 930–957. Autor, D. H., & Houseman, S. N. (2010). Do temporary-help jobs improve labor market outcomes for low-skilled workers? Evidence from”work first. American Economic Journal: Applied Economics,2(3), 96–128. Bassanini, A., Nunziata, L., & Venn, D. (2009). Job protection legislation and productivity growth in OECD countries. Economic Policy,24, 349–402. Bentolila, S., Cahuc, P., Dolado, J., & Le Barbanchon, T. (2012). “Two-tier labour markets in a deep recession: France vs. Spain.”The Economic Journal,122, 155–187. 216 F. CASTELLANI ET AL. Blanchard, O. J., & Landier, A. (2002). The perverse effects of partial labor market reform: Fixed duration contracts in France. The Economic Journal,112, 214–244. Blanchard, O. J., & Tirole, J. (2003). Contours of employment protection reform (MIT Department of Economics Working Paper, No. 03-35). Boeri, T., & Garibaldi, P. (2007). Two tier reforms of employment protection: A honeymoon effect? The Economic Journal,117, 357–385. Boeri, T., & Garibaldi, P. (2008). Un Nuovo Contratto per Tutti. Torino: Chiarelettere. Cahuc, P. (2012). For a unified contract. European Labour Law Journal,33, 190–205. Cahuc, P., Cheron, A., & Mahlebert, F. (2016). Explaining the spread of temporary jobs and its impact on labor turnover. International Economic Review,572, 533–572. Cahuc, P., & Postel-Vinay, F. (2002). Temporary jobs, employment protection and labor market performance. Labour Economics,9,63–91. Cappellari, L., Dell’Aringa, C., & Leonardi, M. (2012). Temporary employment, job flows and productivity. A tale of two reforms. The Economic Journal,122, 188–215. Cazes, S., & Laiglesia, J. R. (2015). Temporary contracts, labour market segmentation and wage inequality. In Labor markets, institutions and inequality (Chap. 6, pp. 147–183). Edward Elgar Publishing. Damiani, M., & Pompei, F. (2010). Labour protection and productivity in EU economies: 1995–2005. European Journal of Comparative Economics,7(2), 373–411. Dolado, J. (2012). The pros and cons of the latest labour market reform in Spain. Spanish Labour Law and Employment Relations Journal,1,22–30. Dolado, J. J., & Stucchi, R. (2014). Do temporary contracts affect TFP? Evidence from Spanish manufacturing firms (IZA Discussion Paper N. 3832). Eslava, M., Haltiwanger, J., Kugler, A., & Kugler, M. (2014). The effects of regulations and business cycles on temporary contracts, the organization of firms and productivity (Documentos de Trabajo del CEDLAS). Giuliano, R., Kampelmann, S., Mahy, B., & Rycx, F. (2017). Short notice, big difference? The effect of temporary employment on firm competitiveness across sectors (IZA Discussion Paper Series). Ichino, A., Mealli, F., & Nannicini, T. (2008). From temporary help jobs to permanent employment: What can we learn from matching estimators and their sensitivity? Journal of Applied Econometrics,233, 305–327. IMF. (2014). Italy; Staffreport for the 2014 Article IV consultation (IMF StaffCountry Reports 14/ 283). International Monetary Fund. IMF. (2015). France; Selected issues (IMF StaffCountry Reports 15/179). International Monetary Fund. IMF. (2017). Spain; 2016 Article IV consultation-press release; staffreport; informational annex; staffstatement; and statement by the executive director for Spain (IMF StaffCountry Reports 17/ 23). International Monetary Fund. Jona-Lasinio, C., & Vallanti, G. (2011). Reforms, labour market functioning and productivity dynamics: A sectoral analysis for Italy (Working Papers LuissLab, 1193). Levinsohn, J., & Petrin, A. (2003). Estimating production functions using inputs as controls for unobservables. Review of Economic Studies,70, 317–342. Nielen, S., & Schiersch, A. (2016). Productivity in German manufacturing firms: Does fixed-term employment matter? The International Labour Review,155,4. OECD. (2014). Employment outlook 2014. OECD Publishing. Olley, G. S., & Pakes, A. (1996). The dynamic of productivity in the telecommunications equipment industry. Econometrica,64, 1263–1297. Wasmer, E. (2006). General versus specific skills in labour markets with search frictions and firing Costs. American Economic Review,963, 811–831. JOURNAL OF APPLIED ECONOMICS 217 Appendix A. Data and definition of variables A. Sample Selection Initially, there are 12,434 plants (11,745 firms), corresponding to 97,312 observations with nonmissing industry values and with more than 10 employees in the period 2000–14. First, we drop observations in the top/bottom 1% of value added; and we are left with 12,359 plants (11,681 firms) and 95,388 observations. We also exclude observations that show more than 400% growth from 1 year to the other in the total labor variable, as this level of growth in employment does not seem plausible: we are left with 12,359 plants (corresponding to the same number of firms, and 94,746 observations). Following Eslava et al. (2014), we also drop plants that consistently have fewer than 15 employees: we are left with 10,761 plants (10,142 firms), corresponding to 89,859 observations. We remove firms from industries with too few observations (less than 70): tobacco (28 observations) and office & equipment (35 observations); we are left with 10,750 plants (10,132 firms) and 89,796 observations. Since in Bench2 estimations we need to include variables up to their second-lag and since we prefer to use the same sample for all estimations (OLS, 2-steps, Bench1, Bench2) to obtain comparable results, we need to drop observations that do not have up to two lags in the variables of interest. Hence, we lose observations in the first two years of the sample: our regression will relate to the period 2002–2014. We also lose observations for firms with missing values in the two previous lags in any of the variables entering the regressions. Finally, when applying OP we also lose all the observations in which investment is either missing or zero; while applying LP we lose observations in which consumption of electricity is either missing or zero. For OP we are left with 6,997 plants (6,591 firms), corresponding to 37,222 observations. The sample when running LP estimates is larger: 8,430 plants (7,980 firms), corresponding to 57,967 observations. B. Variable Definitions Value Added: it is the total revenue received from the use of inputs (land, capital, labor, business organization) involved in the production function. The DANE calculates the value added as the difference between gross output and intermediate consumption. Gross Output: the value of gross output is equal to the “Value of all the establishment’s manufactured goods”, plus the “Revenues from industrial subcontracting made for others”, plus the “Value of the electric energy sold”, plus the value of the revenue from CERT, 30 plus the “Value of stocks of the products in the manufacturing process at the start of the year”, minus the “Value of the products in the manufacturing process at the end of the year”, plus the value of other operating revenues, such as those generated by rent of goods produced and the installation, maintenance and repair of products manufactured by the establishment. Capital: book value of fixed assets; it corresponds to all the fixed assets of the establishment, defined as those relatively durable goods not to be sold, for the use of the establishment and the development of its industrial activity; it includes all the physical assets that are expected to have a productive life longer than 1 year, and all those extensions, additions or improvements and major reforms that extend the life or normal economic efficiency of the assets (land, buildings, machinery and equipment, ongoing construction, office equipment, etc.). Gross Investment: includes the sum of the investments in new and used fixed assets, manufactured for their own use or received in transfer from other establishments, less sales, withdrawals, and transfers of fixed assets owned by the plant. 30 Discounts caused during the year by exports of products y and subproducts prepared by the establishment. 218 F. CASTELLANI ET AL. Energy Consumption: value of electric power purchased (only the one attributable to the manufacturing process); it corresponds to the value in kwh. acquired by the establishment for its operation. Permanent Personnel: corresponds to the yearly average of payroll personnel or plant personnel with a direct labor link with the establishment through an indefinite term contract. Temporary Personnel: temporary staffhired directly by the establishment (corresponds to the yearly average staffhired through fixed-term contract to accomplish specific tasks under an agreed remuneration), and temporary staffhired by companies specialized in the supply of personnel (corresponds to yearly average staffwithout either a labor or contractual relationship with the plant, hired through agencies specialized in the supply of personnel to carry out specific tasks in the industrial establishment). Skilled personnel: professionals, technicians, and production technologists (national and foreign; it refers to persons directly linked to productive activities or tasks directly related with production; this category includes mechanical, chemical, industrial, electrical, mining, and oil engineers, etc., and technicians and technologists who work directly in the production process), and administration and sales staff(the people who lead the economic, financial, and administrative aspects of the establishment and responsible for the elaboration and conduction of the general policy of the company, as managers, sub-managers, and paid directors; this includes: heads of the administrative staff, supervisors, security guards, staffthat do not work in the area of production, vendors, distributors and/or delivery agents, if their remuneration is borne by the establishment; it does not include warehouse staff, administrative offices, management, warehouses and other auxiliary units not directly linked to the establishment or located in a place different from the productive plant). Unskilled personnel: production workers (it corresponds to workers dedicated to manufacturing, processing, assembling, maintenance, inspection, storage, packing, loading such as: internal delivery couriers, chauffeurs, cleaning machines personnel, supervisors, and foremen who work manually, drivers of vehicles for transport of raw materials, materials or products only within the establishment, workers dedicated to maintenance and repair (mechanical, electrical, etc.) of the machinery and industrial equipment; administrative supervisors, supervisors dedicated mainly to the supervision of the workers, surveillance personnel, guards, staffworking in production area). Deflators: variables are adjusted with the producer price index (PPI) for each two-digit ISIC (International Standard Industrial Classification of All Economic Activities) sector; December values are taken for each year; base period December 2014 = 100. As PPI is at the sector level (not plant level), what we measure is TFPR (“revenue”productivity). Exporters:firms that sell at least 10% of their total sales abroad. JOURNAL OF APPLIED ECONOMICS 219 Table A1. Results by industrial subsector (LP-ACF). Panel A. Food & Beverages Panel B. Wood Panel C. Textiles 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 (1) (2) (3) (1) (2) (3) (1) (2) (3) β k 0.134*** 0.134*** 0.134*** 0.057 0.055 0.054 0.066*** 0.064*** 0.065*** (0.005) (0.005) (0.005) (0.059) (0.061) (0.061) (0.013) (0.013) (0.013) β l 0.915*** 0.915*** 0.915*** 1.098*** 1.097*** 1.099*** 0.933*** 0.930*** 0.932*** (0.017) (0.017) (0.017) (0.066) (0.066) (0.066) (0.024) (0.024) (0.024) s 1.122*** 1.124*** 1.122*** 0.958*** 0.954*** 0.949*** 0.830*** 0.825*** 0.826*** (0.042) (0.042) (0.042) (0.158) (0.158) (0.155) (0.031) (0.033) (0.032) γ−0.009 0.002 −0.031 −0.041 −0.018 −0.011 (0.006) (0.006) (0.054) (0.034) (0.012) (0.010) Observations 11,740 11,740 11,740 866 866 866 2,607 2,607 2,607 Panel D. Apparel Panel E. Leather Panel F. Paper 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 (1) (2) (3) (1) (2) (3) (1) (2) (3) β k 0.090*** 0.089*** 0.089*** 0.053*** 0.049*** 0.050*** 0.137*** 0.140*** 0.139*** (0.010) (0.010) (0.010) (0.015) (0.015) (0.015) (0.029) (0.029) (0.029) β l 0.834*** 0.833*** 0.833*** 1.035*** 1.026*** 1.028*** 1.029*** 1.031*** 1.031*** (0.014) (0.013) (0.013) (0.019) (0.018) (0.018) (0.050) (0.048) (0.049) s 0.787*** 0.786*** 0.785*** 1.140*** 1.134*** 1.127*** 0.722*** 0.725*** 0.722*** (0.022) (0.023) (0.023) (0.050) (0.049) (0.049) (0.050) (0.050) (0.050) γ−0.014* −0.014** −0.059*** −0.050*** −0.034* −0.028 (0.009) (0.007) (0.012) (0.013) (0.020) (0.020) Observations 4,830 4,830 4,830 1,753 1,753 1,753 1,493 1,493 1,493 Panel G. Printing Panel H. Petroleum Panel I. Chemicals 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 (1) (2) (3) (1) (2) (3) (1) (2) (3) β k 0.137*** 0.137*** 0.137*** 0.253* 0.246 0.225 0.101*** 0.101*** 0.101*** (0.010) (0.010) (0.010) (0.148) (0.152) (0.175) (0.012) (0.012) (0.012) β l 1.042*** 1.041*** 1.042*** 0.357*** 0.339** 0.342** 0.913*** 0.912*** 0.913*** (0.019) (0.019) (0.019) (0.130) (0.132) (0.143) (0.027) (0.027) (0.027) s 0.761*** 0.758*** 0.762*** 0.082 0.101 0.072 0.664*** 0.667*** 0.664*** (0.029) (0.031) (0.031) (1.668) (1.857) (4.445) (0.033) (0.033) (0.033) γ−0.010 0.001 −0.251 −0.419** −0.017** −0.009 (0.013) (0.010) (0.242) (0.200) (0.008) (0.009) Observations 3,344 3,344 3,344 242 242 242 5,497 5,497 5,497 Notes: Bootstrapped standard errors in parentheses, ***p < 0.01, **p < 0.05, *p < 0.1 220 F. CASTELLANI ET AL. Table A1. Results by industrial subsector (LP-ACF –continuation). Panel J. Rubber & Plastic Panel K. Minerals Panel L. Basic Metals 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 (1) (2) (3) (1) (2) (3) (1) (2) (3) β k 0.078*** 0.077*** 0.078*** 0.094*** 0.094*** 0.094*** 0.079** 0.073** 0.074** (0.012) (0.012) (0.012) (0.014) (0.014) (0.014) (0.036) (0.036) (0.036) β l 0.942*** 0.942*** 0.942*** 1.123*** 1.122*** 1.123*** 1.125*** 1.123*** 1.122*** (0.024) (0.024) (0.024) (0.037) (0.036) (0.037) (0.036) (0.036) (0.036) s 0.782*** 0.782*** 0.783*** 0.770*** 0.769*** 0.770*** 0.757*** 0.748*** 0.743*** (0.026) (0.026) (0.026) (0.063) (0.064) (0.065) (0.050) (0.056) (0.055) γ−0.002 0.006 −0.019 −0.002 −0.034 −0.040 (0.008) (0.008) (0.025) (0.020) (0.034) (0.026) Observations 4,965 4,965 4,965 3,539 3,539 3,539 1,108 1,108 1,108 Panel M. Metal Panel N. Equipment Panel O. Electronic Machinery 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 (1) (2) (3) (1) (2) (3) (1) (2) (3) β k 0.140*** 0.137*** 0.139*** 0.017 0.014 0.015 0.067*** 0.068*** 0.067*** (0.014) (0.014) (0.014) (0.012) (0.012) (0.012) (0.013) (0.013) (0.013) β l 0.946*** 0.943*** 0.946*** 1.065*** 1.059*** 1.060*** 1.106*** 1.107*** 1.107*** (0.020) (0.020) (0.020) (0.017) (0.017) (0.017) (0.046) (0.046) (0.046) s 0.820*** 0.814*** 0.818*** 0.825*** 0.819*** 0.812*** 1.125*** 1.127*** 1.131*** (0.034) (0.033) (0.034) (0.032) (0.033) (0.033) (0.104) (0.106) (0.108) γ−0.039*** −0.009 −0.057*** −0.054*** 0.025 0.023 (0.013) (0.011) (0.020) (0.013) (0.033) (0.030) Observations 3,656 3,656 3,656 3,144 3,144 3,144 1,091 1,091 1,091 Panel P. Medic. Instrum. Panel Q. Vehicles Panel R. Otrher Transp. 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 2 steps Bench 1 Bench 2 (1) (2) (3) (1) (2) (3) (1) (2) (3) β k −0.027 −0.026 −0.027 0.072*** 0.067*** 0.071*** 0.075 0.075 0.067 (0.035) (0.037) (0.036) (0.018) (0.018) (0.018) (0.135) (0.135) (0.136) β l 1.074*** 1.075*** 1.074*** 1.059*** 1.055*** 1.058*** 1.279*** 1.274*** 1.277*** (0.049) (0.050) (0.049) (0.036) (0.037) (0.036) (0.099) (0.101) (0.100) s 0.739*** 0.739*** 0.739*** 1.050*** 1.042*** 1.047*** 0.721*** 0.723*** 0.704*** (0.103) (0.103) (0.102) (0.045) (0.048) (0.046) (0.080) (0.084) (0.082) γ0.012 −0.002 −0.055** −0.014 −0.059 −0.143 (0.054) (0.051) (0.024) (0.020) (0.093) (0.089) Observations 411 411 411 1,284 1,284 1,284 330 330 330 Notes: Bootstrapped standard errors in parentheses, ***p < 0.01, **p < 0.05, *p < 0.1 JOURNAL OF APPLIED ECONOMICS 221