Understanding firms demand for temporary labour in developing countries

Transcription

1 Understanding firms demand for temporary labour in developing countries Mariya Aleksynska * ILO Janine Berg ** ILO Preliminary draft. Please do not distribute or quote. Abstract We use data from the World Bank Enterprise Survey of private sector firms in developing countries to investigate the determinants of firms use of temporary labour. Our main findings suggest that nearly all of the micro-level hypotheses for firm use of temporary labour identified by previous literature as relevant for developed countries flexibility needs, cost saving strategies, and technological adjustments are relevant in developing countries. The relevance of any particular micro-level determinant mainly varies across sectors of firm s activity, with significant differences between manufacturing and services; a country s level of development is not relevant. On the macro level, it is the national regulations governing the use of fixed-term contracts, and notably whether fixed-term contracts are authorized for permanent tasks, that play a key role in determining the firm-use of temporary labour. While we do not find evidence that the overall level of employment protection for regular workers affects firms decisions to use temporary labour, the degree of protection afforded by substantive grounds for dismissals is relevant. Employment protection is most relevant in upper-middle and high-income countries, but not in lower- or lower-income countries, irrespective of the sector of firm s activity. JEL classification: Keywords: temporary work, temporary employee, fixed-term contracts, firm-level demand for labour, developing and emerging economies, flexibility * International Labour Office, 4 route de Morillons, CH-1211 Geneva, Switzerland. aleksynska@ilo.org ** International Labour Office, 4 route de Morillons, CH-1211 Geneva, Switzerland. berg@ilo.org We would like to thank Friederike Eberlein for excellent research assistance. The usual disclaimers apply. 1

2 1. Introduction Temporary employment has always existed in labour markets and serves specific purposes. Over the past several decades, however, there has been an increase in the use of temporary work in many parts of both the industrialized and the developing world, often for new motives and in sectors not previously characterized by these arrangements (ILO, 2015a). The shift toward temporary labour and other atypical forms of employment has prompted concerns and policy debates about the future of work and the status of the standard employment relationship and has led to calls for the need to ensure that all workers, regardless of their contractual arrangement, enjoy decent working conditions (ILO, 2015b). In many developing countries, although wage employment continues to be limited, there nonetheless appears to have been an evolution in contractual arrangements among waged workers. Such evolution has differed across countries, possibly as a result of different legislative frameworks and specific economic contexts. While in some countries, the growth in temporary employment could represent a transition from self-employment to wage employment as part of economic development, in others, it could also represent employers preference for fixed-term employment, potentially leading to long-lasting practices. There is a rich theoretical and empirical literature explaining why firms sometimes prefer temporary labour, yet it has been mainly focused on developed countries; little is known about firms preferences for using temporary labour in developing economies. This paper brings together several broad stands of literature: management, industrial relations, sociology, and labour economics, in order to test the relevance of three hypotheses most commonly advanced to explain the use of temporary labour flexibility, cost advantages, and technological change. A better understanding of why firms use temporary labour can help guide policy responses for the benefit of both workers and enterprises. Our study is based on the World Bank Enterprise Survey, which is the largest and most recent firm-level survey in the world. We use a sub-sample of over 72,000 private firms covering 118 developing countries during the period from 2006 to The survey contains detailed questions on firm s current and past activities, thus allowing to test the relevance, at the firm level, of a variety of reasons for the firm use of temporary labour for different types of firms and sectors. The vast country coverage also allows us also to examine the role of labour market institutions and of the macroeconomic environment in affecting firms demand for temporary labour, providing a new evidence for non-oecd, non-european countries, and also providing new evidence to developments during a period of global economic crisis. As 2

3 such, the study contributes to a better understanding of both firm-level and country-level determinants of temporary labour use in developing countries. Our main findings suggest that nearly all of the micro-level hypotheses for firm use of temporary labour identified by previous literature as relevant for developed countries flexibility needs, cost saving strategies, and technological adjustments are relevant in developing countries, too. The relevance of any particular micro level determinant mainly varies across sectors of firm s activity, but not across country s level of development; exhibiting significant differences across manufacturing and services. On the macro level, it is the regulations governing the use of fixed-term contracts (FTCs), and notably whether FTCs are authorized for permanent tasks, that play a key role in determining the firm-use of temporary labour. While we do not find evidence that the overall level of employment protection for regular workers affects firms decisions to use temporary labour, some of its aspects, and notably the degree of protection afforded by substantive grounds for dismissals, are relevant. We also find that employment protection is most relevant in upper-middle and high-income countries, but not in lower- or lower-income countries, irrespective of the sector of firm s activity. Our paper is close in spirit to studies examining the determinants of temporary employment in a multi-country setting, and also addressing the role of institutions (Kahn, 2007; Polaieja, 2006; Baranowska and Gebel, 2010; Hevenstone, 2010; Gebel and Giesecke, 2011; see also Hipp et al., 2015 for a recent literature review). However, those studies focused either on the aggregate level of temporary employment, or on micro worker-level incidence of temporary contracts; and all of them were limited to developed countries. In contrast, our contribution is to focus on the firm-level decisions to employ temporary labour in developing countries. Our study is also very close to Pierre and Scarpetta (2013), who use the same survey, but cover a substantially smaller sub-sample, and focus mainly on the role of institutions. Some of the differences in our findings, explained below, reflect these different sample and focus. The paper begins with a review of the literature and the main hypotheses that have been advanced to explain firms use of temporary labour. In Section III, we provide a brief description of the data, followed by descriptive statistics in Section IV. Section V provides the results of our analyses, divided between those factors that are more internal to the firm and those that are external, such as the macroeconomic situation and the labour legislation. The last section gives a summary of our findings and discusses the implications for policy debates as well as future research. 3

4 2. Literature review and hypothesis setting What leads firms to use temporary labour? Why do some firms use temporary labour more than others? There are several stands of literature management studies, industrial relations, sociology, and labour economics, that study the determinants of firm s decision to use temporary labour. The reasons most commonly put forward are flexibility, cost advantages, and technological change. These three reasons, however, are not necessarily independent of one another, as organizations may use temporary workers for any one, or a combination, of these reasons; notably, the cost advantage reason is often embedded into flexibility and technological change arguments. Nonetheless, the relevance of each of these reasons can be studied at the micro and at the macro level. 2.1 The flexibility argument Firms in competitive labour markets are subject to fluctuations in demand for their goods and services, either because of seasonality, changes in the business cycle, competition from other firms for market share, or external shocks. As labour is a variable cost in production (or at least a quasi-fixed cost (Oi, 1962)), firms have an incentive to ensure numerical flexibility in their labour force, so that they do not employ more staff than necessary when demand falls. Temporary labour is a convenient source of numerical flexibility. On the other hand, firms also need to ensure that they have sufficient, knowledgeable staff to carry out the core operations of the firm and ensure firm longevity. Thus firms often seek the right balance between stability and flexibility in their workforce. Economists have long recognized that firms operate with this consideration in mind. In their seminal study of internal labour markets, Doeringer and Piore (1971) explained how within a firm there are essentially two labour markets, a primary, or internal, market consisting of jobs that are wellpaid, stable and with advancement opportunities and a secondary, or external, market, which is lower-paid, lower-skilled and with fewer opportunities for training and advancement. The authors explained how many firms in diverse industries organized their workforce as internal labour markets, but also relied on a secondary group of workers whose skills were general and where the recruitment, screening and training costs were markedly lower. Building on the insights of Doeringer and Piore (1971) and the efficiency-wage model of Bulow and Summers (1986) and Shapiro and Stiglitz (1984), Saint-Paul (1996) shows mathematically how dualism along the permanent-temporary workers divide can arise endogenously within a firm as a response to demand fluctuations. Because searching and 4

5 recruiting workers with necessary skills and monitoring their work is costly, firms pay efficiency wages (wages above market clearing) to both motivate and retain workers. Since adjusting such labour to demand fluctuations is also costly, firms will have an incentive to split their workforce into a higher-paid, primary or core workforce and a secondary, or peripheral, workforce, for whom the adjustment costs are substantially lower. These adjustment costs may be related to the direct firing costs, but stem primarily because secondary employees are paid less than the efficiency wages that primary employees receive. Saint-Paul (1996) shows that such dualism arises naturally and can be optimal for a firm, even in absence of labour market institutions and regulations, such as employment protection legislation. In countries where firms have limited possibility of choosing which workers can be granted employment security because of the existing employment protection legislation, the dualism within a firm can also arise when legal provisions explicitly allow for the use of temporary labour, and especially if legal provisions stipulate different level of protection for permanent and for temporary workers, thereby providing an additional incentive to use temporary labour. Empirical evidence both in economics and in management literature shows that temporary workers can indeed be used to help firms attain numerical flexibility (Bentolila and Dolado, 1994), allowing them to survive in the adverse macroeconomic conditions (Holmlund and Storrie, 2002; Benito and Hernando, 2008), and deal with seasonal demand or with fluctuations in labour supply (Ko, 2003). In addition, organizations in industries with highly volatile demand are more likely to recur to temporary labour (Cappelli and Keller, 2013); but also organizations of smaller size, which are less likely to have employees available to meet temporary adjustment needs (Davis-Blake and Uzzi, 1993). Abraham and Taylor (1996) also show that temporary employment indeed represents a buffer stock to adjust to fluctuations in demand, but warn that while the demand for such work increases in less stable economic environments, it may also be mitigated by the firms ability to reschedule the delivery of some of its products and services to off-peak periods. In particular, in firms providing out-of-site services and facing unstable demand, less labour is contractedout, because employers do attach a certain value to maintaining stable relationships with their regular employees who may perform other in-house tasks during slow periods (see also Ton, 2014). Moreover, Hunt (2000) also warns that fixed-term contracts may offer firms a lessthan desired flexibility due to restrictions that typically apply to the renewals of these contracts. Caggese and Cugnat (2008) suggest that the flexibility provided by temporary workers is particularly useful for firms facing financing constraints. 5

6 Some further nuances regarding the use of temporary workers as buffers is provided by the literature examining the role of unions and collective bargaining in firm s decision to hire temporary workers. On the one hand, unions may contribute to growing recourse of temporary employment if it helps isolate permanent workers from the negative effects of demand volatility and technological shocks (Saint-Paul, 1996; Bentolila and Dolado, 1994; Abraham and Taylor, 1996). Thus, a few empirical studies have found that temporary contracts are more wide-spread in countries with higher union densities or higher collective bargaining coverage (Kahn, 2007; Baranowska and Gebel, 2010; Hevenstone, 2010). On the other hand, unions may also oppose the recourse to temporary labour, either out of social cohesion considerations (Visser, 2002), or because they may perceive temporary workers as threats to their bargaining power (Heery, 2004), especially when firms strategically use externalized workers to diminish unions power (Hatton, 2014). Differences in unions impact on temporary labour may also eventually depend on the overall macroeconomic situation (Eichhorst and Marx, 2011; Devicienti et al., 2014), and on the prevalence of different types of temporary contracts: unions may be favourable to firm s demand for nontraining contracts used as buffers, but indifferent to demand for training temporary contracts used as a screening device for future permanent positions (ibid). 2.2 Cost advantages The management literature confirms the theoretical predictions from economics with respect to the role of labour costs in determining temporary labour hiring decisions. First, organizations devoting significant resources to hiring workers with highly job-specific profile, and also firms providing firm-specific training, would indeed be less likely to fill the vacancies requiring such training by temporary workers, because they would value the possibility to recoup this kind of investment (Davis-Blake and Uzzi, 1993). However, if temporary contracts can be used with for screening (Portugal and Varejao, 2009), such hiring costs can be reduced by more substantial recourse to temporary work (Faccini, 2014), provided that temporary workers are subsequently converted into permanent ones. As the pool of potential suitable job applicants is greater during recessions, firms have higher incentives to use temporary contracts for screening purposes when unemployment is high (Holmlund and Storrie, 2002). Summing up, the overall effect of the hiring costs on the use of temporary work depends on the type and purpose of temporary contracts. In addition, firm size can also mitigate the demand for temporary labour, as larger size can also allow firms to 6

7 spread various costs, including training of new employees, over larger base (Knoke and Kalleberg, 1994). Second, organizations may value the lower direct labour costs that firms incur while using temporary labour, because in a vast majority of cases, temporary workers are indeed paid lower wages as compared to permanent workers (for a review of empirical evidence, see ILO, 2015a). More specifically, the lower wages can be the result not only of different screening of temporary workers job intensity, as compared to permanent workers, but also because of probationary nature of some temporary contracts; shorter tenure of temporary workers; exclusion of temporary workers from corporate benefits, such as annual bonuses; or simply unequal treatment of non-standard workers (Lee and Yoo, 2008). Temporary workers are also less likely to be subject to social security contribution payments as compared to permanent workers, or have a paid leave. Given this, the higher the wages and fringe benefits in an organization, the more incentives firms have to use temporary workers as to offset such costs (Davis-Blake and Uzzi, 1993; Kalleberg et al., 2003; Houseman, 2001), although such relationship may be nonlinear and implying limits to the cost advantage of using temporary workers (Nielen and Schiersch, 2014). Third, the use of temporary employment, especially in the European context of the past three decades, has often been explained by the significantly lower firing costs associated with terminating temporary contracts, as compared to permanent contracts. While workers on fixed-term contracts are typically well protected during the period covered by the contract (in some instances, termination of such contracts before their end date may entail payment of all wages due until the contractual end date), at the end of the contract, generally no reasons need to be provided by the employer to justify the end of the employment relationship, beyond the end date of a fixed-term contract being reached. In contrast, terminating employment relationship with permanent workers, at the initiative of employer, usually entails certain costs, including severance payments, costs associated with notification procedures, and other compensatory payments if terminations are unfair. Starting in the 1970s, numerous European countries partly deregulated labour markets with the aim of increasing labour market flexibility, by allowing for a wider use of temporary contracts, by expanding their scope to jobs that were not temporary in nature, and by increasing the allowed duration and number of renewals. At the same time, employment protection for permanent workers remained relatively intact. As a result, the wedge in the costs associated with terminating a temporary and a permanent worker grew, leading many researchers to attribute the growth of temporary employment in some European countries to these partial 7

8 reforms (Bentolila and Dolado, 1994; Blanchard and Landier, 2002; Boeri and Garibaldi, 2007; Faccini, 2014; OECD, 2014). Similar reforms on the use of temporary labour occurred in some developing countries, particularly Peru (Vega-Ruíz,2005). Nonetheless, the low separation costs for temporary workers has to be weighed against costs associated with the frequent search for new workers (Holmlund and Storrie, 2002). Moreover, Kahn (2010) shows that not only liberalization of temporary jobs increased their incidence, but also that such temporary jobs substituted some of the permanent jobs, without increasing the overall employment. Using a Mortensen-Pissarides type model, Boeri (2011) explains that theoretically, such two-tier reforms of labour markets provide a strong incentive to employers to systematically offer temporary contracts to entry jobs a prediction consistent with evidence that the share of new hires on temporary contracts reached over 75% in Italy, Poland, Portugal, Slovenia, Spain,and Sweden throughout 2000s-2015s (ibid; OECD, 2014). One implication of this model is also that, theoretically, the share of temporary workers increases with the strictness of employment protection for permanent contracts 1. While the empirical evidence on this is relatively mild in Boeri (2011), it is more robust in Booth et al. (2002), and in Kahn (2007) for youth temporary employment 2. Polavieja (2006) also shows that the effect of higher employment protection for permanent contracts on the use of temporary employment can be amplified by greater economic uncertainty; in other words, firms cost and flexibility considerations brought about by external factors tend to reinforce each other. 2.3 Technological change Whereas the economics literature has focused on the use of temporary labour as a response to demand fluctuations, the management literature has emphasized the production model of the firm, particularly the extent to which production is standardized. The simplification of tasks brought about by technology means that tasks can be performed by less skilled workers, who need less training and thus can be brought in at short notice (Nollen and Axel, 1996). As a result, turnover is less costly for firms, thus there is less of an incentive to cultivate long-term employment relationships. Uzzi and Barsness (1998) suggest that firms using computerized technologies are also the ones that recur more often to fixed-term workers. On the other hand, sophisticated technologies may increase firm-specific knowledge 1 See Hipp et al., 2015 for a recent literature review on the role of other labour market institutions. 2 Note, however, that Kahn (2007) is based on micro worker-level data, not on micro firm-level data; while theoretical predictions of Boeri (2011) concern firm decisions. 8

9 and lessen recourse to temporary and outsourced labour, both to save on training costs and to preserve their know-how (Mayer and Nickerson, 2005). In addition, even with standardized production models, there may be advantages to having a stable workforce, in which the worker is encouraged to communicate problems and suggest innovations (Ton, 2014). Similarly, complex jobs, either from an interpersonal or a technological viewpoint, are less likely to be performed by temporary workers, and firms knowledge workers are most likely to be permanent employees (Davis-Blake and Uzzi, 1993). 3. Data description Our analysis is based on the World Bank Enterprises Survey, a representative firmlevel survey of 130,000 private companies in 135 developing countries. The data were collected between 2006 and 2014, with most countries being surveyed twice, and some countries, such as Bulgaria and Chile, surveyed three times. Because a different set of enterprises was surveyed in each wave, we use these data in a pooled setting. The survey data are collected from face-to-face interviews with top managers and business owners of formal (registered) companies with 5 or more employees, operating in manufacturing and services sectors. The survey covers a broad range of questions on firmlevel characteristics, business environment topics, and characteristics of the firms workforce, including the number of temporary and permanent workers in an enterprise. The latter questions allow computing both the use and the prevalence of temporary workers (as a share of all employees) in an enterprise. The exact wording of the relevant question is How many full-time temporary employees did this establishment employ in fiscal year?, where temporary workers refer to temporary or seasonal employees, defined as all paid short-term (i.e. for less than a fiscal year) employees with no guarantee of renewal of employment contract (World Bank, 2011). This definition is somewhat different from the one employed in the literature: it is relatively broad as it captures seasonal workers; at the same time it is quite narrow, because it excludes temporary workers employed for more than one year or having promise of renewal of their temporary contract. It, however, satisfies the key feature of temporary work from the firms viewpoint: the fact that no guarantee of renewal of employment contract is provided. Using this question, combined with the question on the number of permanent full-time employees employed in an establishment, we construct a variable temp_share_all, as a ratio of temporary workers to the sum of temporary and permanent workers. This variable is used 9

10 as principal dependent variable in further analysis. As explained below, we also test alternative dependent variables. After restricting the sample to firms containing information on the number of temporary employees and other key firm characteristics, as well as after retaining only meaningful observations (dropping firms with negative share of temporary employees, or this share greater than one; firms with negative age or unknown ownership type), the sample is reduced to 72,630 observations in 118 countries. Appendix A lists the countries used for the analysis, and the exact years of data collection. 4. Descriptive statistics Figure 1 shows the distribution of temporary employees, as percent of all employees in developing countries, both in manufacturing and services private sector. It shows that the average share of firms temporary workforce spans from under 5% in Latvia, Jordan, and Sierra Leone, to over 25% in Vietnam, Peru, and Mongolia, being broadly consistent with the data from national sources (for the most recent figures, see ILO, 2015a). The mean share of temporary workers is 11%, with roughly one-third of the surveyed countries with temporary employment around this mean. Interestingly, only about 40% of all firms throughout the world employ temporary workers, meaning that about 60% of firms do not use temporary labour at all 3. Among those firms that employ temporary workers, the average share is 27.5% (Figure 2). This may be evidence that firms first decide whether to use or not temporary labour, and only then they decide how much. This finding also has implications for the method of estimation, as discussed in the next section. Note also that there is a spike in the use of temporary labour at 50% (Figures 2, 3, 4, and 6); suggesting that numerous firms might have adopted rounding in reporting the number of temporary workers. While Appendix B contains full descriptive statistics of all independent variables used in the analysis, in this section, we also highlight the distribution of temporary workers across some of the key firm and country characteristics. 3 We checked the differences in means of characteristics of firms that do, and that do not, use temporary labour. While the differences are always statistically significant based on t-tests for differences of sample means, they are not large. The most sizeable differences are found across firm size (firms employing temporary labour being bigger in size as compared to those not employing); and depending on whether firms provide training for permanent employees or not (firms employing temporary labour being also those that provide considerably more training). 10

11 Figure 3 shows the distribution of temporary employees by firm size. Descriptive evidence is consistent with the literature: it is the smallest firms that employ the highest shares of temporary labour. Table 1 provides an overview of the incidence of temporary employment across regions and sectors. It shows that in the Middle East and North Africa, as well as in South Asia, temporary employment, on average, is more wide-spread in manufacturing than in services. The opposite is true in Africa, East Asia and Pacific, Europe and Central Asia, and in Latin America. With the exception of Middle East and North Africa, construction and transportation seem to be the sub-sectors that uniformly employ the largest share of temporary workers across the world. Temporary workers represent over 35 % of all workers in this sector in East Asia and Pacific, over 30% in Latin America and Caribbean, over 25 % in Africa, and nearly 20 % on South Asia. Within manufacturing, it is the leather industry that employs the largest share of temporary workers (average number of temporary employees per firm is 32%); in services, it is construction and transportation (average number of temporary employees per firm is 39%) (See Table 2). Figures 4-7 also examine the relationship between the share of temporary wage employment and key country characteristics identified in previous research: the level of economic development, and the role of labour market institutions governing employment contracts. From Figure 4-5, firms in lower- and lower-middle income countries have considerably higher shares of temporary employees than upper-middle and high-income countries, their number clearly decreasing with country s development. This finding supports the hypothesis that temporary wage employment may be a transitory phenomenon, as countries transition from self-employment to waged employment; it also reflects the more widespread use of casual employment in developing countries. In Figure 6, temporary wage employment distribution is shown by two types of countries: those that authorise the use of temporary labour for permanent tasks, and those that do not. While two distributions are highly similar, there is slight evidence that countries where FTCs are legally authorised for permanent tasks, also have higher number of firms using temporary workers, particularly at the higher-end of the temporary employment distribution. The regulations of regular contracts appear unrelated to the firms use of temporary labour (See Figure 7). 5. Empirical setting and results 5.1. Baseline specification 11

12 Drawing on the overview of the literature, we divide the reasons that prompt firm s use of temporary labour into those reasons that are internal to the firm (micro-level) and those that are external (macro-level). Within internal reasons, we distinguish between those that are related to flexibility, cost, and technology. For external factors, we test the relevance of the macroeconomic conditions and some labour market institutions. Our empirical set-up follows closely Davis-Blake and Uzzi (1993), Devicienti et al. (2014), and Portugal et al. (2009). The baseline specification allowing to analyse the firmlevel inside determinants of using temporary workforce is as follows: Temp_share_all ijkt = α ijk + β 1i X i + β 2i Y i + j j + k k + t t + ε ijkt (1) where Temp_share_all ij is the share of temporary labour in firm i operating in sector j country k and year t; X i is the set of individual baseline firm characteristics; Y i is the set of additional individual firm characteristics that allow testing the relevance of flexibility, cost, and technology factors. In all specifications, we control for sector, country, and year, by including the corresponding j j, k k, t t dummies; ε ijkt is the error term. Among the individual baseline firm characteristics X i, we include the total number of employees in a firm and its square, to capture both the actual firm size and its possible nonlinear effects; firm age as a difference between the year of the survey and the date of firms creation; two dichotomous variables for firm ownership: whether the firm is mainly owned by domestic private or foreign private individuals, the benchmark category being whether the firm is owned by domestic government; as well as average productivity, defined as logarithm of the ratio of the last years establishment s total annual sales converted in the US dollars divided by all employees, as a measure (or rough proxy) of efficiency. Table 3 column 1 presents the results of this baseline estimation using simple OLS regression. In line with the literature (Portugal et al., 2009), firm size exhibits a non-linear effect: firms of larger size tend to hire a greater share of temporary workers, but only up to a certain point. Both very small and very large firms tend to hire less temporary labour, in relative terms, as compared to medium-size firms. Firm age has a negative association with the share of temporary workforce, in line with the literature suggesting that with age, practices become standardized and resistant to change over the long run, but also because among older firms there are more firms that survived through unstable environments (Rousseau and Libuser, 1997; Uzzi and Barsness, 1998). In contrast to domestic mainly publicly owned establishments, domestic private firms tend to employ less temporary labour, 12

13 in relative terms, though this result is not robust to alternative specifications; while foreign ownership does not result in a different pattern of employing temporary workforce. Lastly, firms with higher average productivity recur less to temporary labour. To explore the rich data at hand, we also tested potential importance of some other variables, such as a dichotomous variable for female ownership, a dichotomous variable measuring whether a firm has been formally registered when it began operations, and a variable measuring the skill production mix (computed as a ratio of permanent full-time employees that are skilled production workers, to compared to all full-time permanent workers). We have not found any statistically significant effects on these variables (Table 3, columns 2-3), and hence did not retain them for further specifications. We did find a statistically significant effect for the variable measuring the proportion of nonproduction employees to all permanent full-time employees (Table 3, column 4); however, this variable is only available for manufacturing, thus considerably reducing the sample size. We have not retained it in further specifications either. In the remaining columns of Table 3, we check for the robustness of the baseline specification. We begin by checking alternative measure of our dependent variable, expressing it plainly as the number of temporary workers in an establishment (Table 3, column 5). This change somewhat affected the results on linearity of the firm size, and rendered firm age insignificant. Another robustness check relates to the estimation method. As noted earlier, a substantial percentage of firms do not employ temporary labour at all. This means that our dependent variable, expressed either as a share or as a number of temporary workers, features a sizeable number of zeroes. To account for this, a Poisson count model can be seen as more appropriate than simple OLS. We thus repeat the estimation of column 1, using Poisson maximum likelihood regression, and report it in in Table 3, column 6. Poisson regression, however, produces consistent estimates only if the error terms satisfy the log normality and homoskedasticity conditions, which are indeed very strong assumptions. It may even produce serious bias if the number of zeros is too large (Martin and Pham, 2009); in which case a possible remedy is to use a negative binomial model that may be interpreted as a Poisson-gamma mixture, in other words, a Poisson distribution with unobserved individual (gamma) heterogeneity that also allows for particular forms of dependence for the underlying stochastic process (Portugal et al., 2009). We fit the negative binomial model in Table 3, column 7. The results of both methods in columns 6 and 7 are nearly identical; moreover, they are in line with the baseline results in column (1). Furthermore, we also repeat the estimation of column 5 using Poisson maximum likelihood regression, and report it in in 13

14 Table 3, column 8 (in other words, as compared to column 1, this specification employs both a different dependent variable and a different estimation method). This specification gives more consistent results than the estimation in column 5. Given the consistency between columns 1, and 6-8, but also a certain computational cost of estimating a maximum likelihood or a negative binomial model in a setting with over a hundred variables, in what follows, we perform standard OLS estimations using a relatively conventional dependent variable expressed as a share of temporary workers Micro level determinants of firms use of temporary labour In this section, we explore the very rich questionnaire and construct variables most used in the literature to capture the external flexibility needs and the needs to adjust to the volatile environment; costs of labour; and technological factors. These variables constitute the Y i vector in specification (1). Table 4 reports estimation results, where, in columns 1-3, these variables are grouped by type of hypothesis that they pertain to; and in column 4 they are included jointly. Appendix tables A2 A5 also containe detailed estimations where each of the variables is included one-by-one. Flexibility hypothesis To measure the need for external flexibility, we include information on the principal market served, the extent of informal competition, demand volatility, indicators of firm expansion, and whether a firm is credit-constrained. All variables are found to have a statistically significant impact on the use of temporary labour, both in the individual and (with some exceptions) joint specifications, confirming the relevance of the flexibility hypothesis for firms in developing countries. More specifically, the information on the principal market served is constructed from the survey question regarding the main market (either local, national or international) in which the establishment sold its major product over the past year. From these responses, we construct two dichotomous variables, the first equal to one if the principal market is national, and the second equal to one if the main market is international, with local market serving as a benchmark category. From Table 4 column 1, serving an international market increases chances of relying on more temporary labour, as compared to serving in a national or a local market. Moreover, the coefficient on the international market dummy is twice as high as the coefficient on the national market dummy. This confirms the idea that larger markets are 14

15 prone to higher demand volatility and that firms with higher international activity face more environmental uncertainties (Supangco, 2008). To corroborate this idea, we also tested the relevance of an alternative variable, the percent of sales that are due to direct exports. The coefficient on this variable is also positive and significant; but the joint inclusion of this variable with the market dummy variables renders the international market dummy insignificant, because its effect is reflected in the export sales variable (Appendix A2, columns 1, 2, and 10). Another measure of the demand for flexibility is the degree of competition against unregistered or informal firms, measured as a dichotomous variable equal to one if a firm reports such competition. From Table 4 column 1, the existence of such competition clearly spurs the demand for temporary labour. If firms use more temporary labour as a buffer to adjust to fluctuations in demand, a demand volatility measure is another relevant variable. We construct such measure following Devicienti et al. (2014) as the standard deviation of the difference between current annual sales and the annual sales three years ago calculated at the 2-digit ISIC product classification of the firm s main product. Higher values of this variable indicate higher volatility of demand for the firm s main product; the computation of this measure at the product level, rather than at firm level, also helps mitigate endogeneity concerns related to this variable. From Table 4 column 1, this measure has a positive effect on temporary work use. In Table 4, column 1, we also include the ratio of current number of permanent employees to the number of permanent employees three years ago, and the ratio of current number of permanent employees to the number of permanent employees at the start of firms operations. Higher values of these variables capture a mid-term and long-term growth of the core labour force and firm s expansion. While some of this growth can happen because of the conversion of temporary into permanent contracts and investing in personnel, our results show that both forms of the core s expansion also lead to the expansion of temporary labour. If temporary employees are used to smooth the workload of the regular labour force and sustain the core, then greater core requires greater supplementary temporary labour force; the magnitude of the effect is greater for the middle-term rather than for the long-term variable. In contrast, companies that had downsized might have done so at the expense of changing organizational structure, job definitions, and terminating or non-renewing temporary labour first. Our next variable of interest is the extent to which access to finance is an obstacle to the operations of an establishment. Cagese and Cunnat (2008) argue that the payment of 15

16 wages and firing costs makes hiring and firing sensitive to the financing frictions that firms face. They classify firms as those facing financial constraints if they "answered positively to one or more questions regarding problems in obtaining additional credit". Similarly, we use the variable measuring whether firms reports that access to finance is no obstacle, a minor obstacle, a moderate obstacle, or a very severe obstacle to the current operations of the establishment. We find that this variable indeed has a strong positive association with the use of temporary labour. Finally, there is literature addressing how unionization can affect, both positively and negatively, firms recourse to temporary labour. We use information on the share of unionized workers in an enterprise (Table A2, column 9) to explore this issue. The variable has an insignificant effect, but it is also available for only a fifth of the sample (the question is asked in only 27 countries, and only in 2006; with substantial number of missing values). Thus, we do not retain this variable for further analysis. Cost hypothesis While many of the flexibility-related variables also embed the cost factors, the literature also offers more direct ways of testing for the relevance of the cost hypothesis. We follow it by including the total labour costs, as well as controls for whether firms offer training to permanent staff. In addition, we are testing whether the firm-reported perceptions of the severity of labour regulations and of finding a suitably educated workforce is associated with hiring temporary labour. Research shows that firm perceptions of regulations correlate well with the de jure level of actual labour regulations in a country (Pierre and Scarpetta, 2006). Table 4, column 2 summarizes these results. It shows that the higher is the total labour bill (expressed in logarithmic term, and converted into USD), the higher is the incidence of temporary labour. Training dummy also shows positive and significant result. The literature suggests that training effects may be different depending of the nature of temporary jobs. If they are used as screening device before workers are offered permanent positions, more training should be provided to them. If they are used as buffers, then training options will be more limited. The positive coefficient on training offered to permanent workers is in line with the screening hypothesis. We further exploit the information on two subjective variables: the degree to which labour regulations are an obstacle to the operation of the firm, and the degree to which an inadequately educated workforce is an obstacle to the operation of the firm (with higher values indicating a greater obstacle). Both variables have a significant positive effect, 16

17 moreover, both variables are equally important to firms in making their decision about hiring temporary labour. In a handful of countries, an even more detailed question was asked, whether labour regulations affect decisions of hiring or firing permanent workers. From column 4 of Appendix table A3, this variable is positive and significant, suggesting that indeed firms hire more temporary labour when hiring or firing permanent workers are perceived as costly. To check the robustness of these results, in Appendix table A3 column 5, we include all of these variables jointly, and also control for the extent to which all other possible obstacles, such as access to telecommunications, transportation, customs, facing informal competition, access to finance or land, crime, level of tax rate, tax administration, obtaining a business license, or political stability in a country are also problematic. Labour regulations and inadequately educated workforce remain positive and statistically significant in this specification; moreover, a few other variables appear also to be relevant. Technology factors To check the relevance of technology, we use three variables: the degree to which telecommunications was perceived as an obstacle to the current operations of the establishment; whether the establishment has an internationally recognized certification; and whether the establishment uses any technology licensed from a foreign-owned company (Table 4, column 3). All three variables were significant. Where telecommunications was perceived as an obstacle to the operation of the firm, the firm increased its recourse to temporary labour. This variable, however, more likely reflects volatility in production and thus the need to have a buffer of labour, rather than the effect of technology on the standardization of production and the use of different types of labour. The other two technology variables respond more directly to the literature. Firms with an internationally recognized certification are statistically less likely to employ temporary labour, perhaps reflecting previous findings that firms with more sophisticated technologies are less prone to use temporary and outsourced labour. On the other hand, establishments that use technology licensed from a foreign-company are more apt to use temporary labour. This latter case may reflect a greater standardization of production and thus a concomitant simplification of tasks. As discussed previously, firms with more standardized and simplified production processes have less to risk if they rely more heavily on temporary labour. In column 4 of Table 4, we bring the three hypotheses together. Similar results are obtained for the flexibility and cost variables, but international certification is no longer significant, most likely because of the inclusion of the training variable. 17

18 5.3. Macro level determinants of firms use of temporary labour Firms do not operate in isolation, and hence macroeconomic conditions as well as laws regulating the workplace are important influences on the degree of flexibility and costadjustments required by firms. Moreover, the level of economic development of a country may also affect the level of temporary wage employment as countries transitioning from selfemployment to wage employment may recur more heavily to temporary labour. In Table 5, column 1, we include three dichotomous variables reflecting country level of economic development. As compared to low-income countries (the omitted group), firms in upper-middle and high-income countries feature a lower level of temporary labour, consistent with descriptive evidence in Figures 4-5. As the use of temporary labour procyclical (OECD, 2012, 2014; ILO, 2015a; Serrano et al., 2014 for developing countries), we also control for GDP growth and its three-year lag, as well as for the level of unemployment (Holmlund and Storrie, 2002). 4 These variables, however, are not found to be statistically significant in developing countries, potentially because, unlike in the developed countries, in developing countries the pool of suitable job applicants is not restrained by the number of unemployed, but is also represented by a larger number of self-employed willing to switch to wage employment, thus rendering the number of unemployed less relevant 5. In the remainder of the Table 5, we test the relevance of the labour market regulations governing the termination of regular contracts and the use of fixed-term contracts. To measure the level of regulations governing termination of regular contracts, we include the newly developed EPLEX indicator (ILO, 2015c), which accounts for all aspects of employment protection regulation, such as the availability of substantive and prohibited grounds for dismissal, trial period, notification procedures, length of notice period, severance and redundancy payments, and redress provisions. While this indicator is available for 100 countries around the world and spanning ; its overlap with the World Bank Enterprises Survey country sample is not perfect; and restricts our sample to 45 countries. To measure the regulations governing the use of fixed-term contracts, we use the ILO EPLex database, complementing it with the World Bank Doing Business Indicators database, to 4 To remedy multicollinearity, in these regressions, we do not control for country dummies, but only for sector and year dummies. Thus, we also apply error-correction for error clustering at the country level. 5 These variable, however, are statistically significant, and in line with the existing literature, if no clustering at country level is applied. Thus, the relevance of these variables may be a purely technical issue in this study, but also in other studies that did not correct for error clustering. The results are available on request. 18

19 construct two dichotomous variables: one on whether FTCs are prohibited for permanent tasks; and the other on whether legislation does not limit the use of FTCs. In Table 5 column 2, the EPLEX indicator is included in addition to the controls for macroeconomic conditions; its effect appears to be statistically insignificant, in line with the descriptive evidence of Figure 7. In column 3, two measures of FTC regulations are additionally included. While the EPLEX variable remains insignificant, the dichotomous variable measuring whether FTCs are prohibited for permanent tasks shows a negative and significant sign: the prohibition to use FTCs for non-temporary tasks clearly reduces firms recourse to temporary labour. In Table 5 columns 4-7 we test a series of variations of this regression, including the number of workers in the private sector concerned by the regulation (coverage variable and its interaction with EPLEX: column 4); omitting macroeconomic controls (column 5), omitting the firm-reported perceptions of the extent to which labour regulations are an obstacle, since those may be correlated with the EPLEX (as shown in Pierre and Scarpetta, 2006; column 6), or further omitting the availability of training for permanent staff, since that may be co-determined with temporary labour when level of employment protection is too high (as shown in Pierre and Scarpetta, 2013; column 7). None of these specifications render EPLEX variable significant, or FTC regulations prohibiting FTC use for permanent tasks insignificant. While intuitive and robust, these findings stand in contrast to Pierre and Scarpetta (2013), who report the opposite results, i.e., that it is EPL, and not the regulation of FTCs, that matters for firms recourse to temporary labour. 6 It is possible that these differences are due to a different sample size, a different dependent variable (they use binary dependent variable of firm s use of temporary labour), a different set of controls (ours are substantially richer), and a different measure of EPL (they rely on the World Bank Employing Workers measures that include procedural requirements, notice period, and severance payments). To test for the pertinence of the latter, we include the disaggregated measure of EPLEX (column 8). The only sub-component that is positively and significantly correlated with the firm use of temporary labour is the component measuring the valid grounds for dismissals (lower values suggesting more permissive grounds, or reasons, for dismissing regular workers), but not the severance pay or the notice period and procedures Understanding differences across sectors and income groups 6 Note, however, that these authors only report the results on EPL, and only in one regression; they describe in text, but do not report, their results on FTC regulations. 19

20 As a last step, we explore the potentially different role of both micro and macro factors for firms operating in different sectors and in countries at different stages of development. Table 6 repeats estimations of Table 4 column (4), using manufacturing/services, and then lower/upper income divides. Along the same divides, Table 7 repeats the estimation of Table 5, column 8 (that is, building on Table 4, column 4, but including macroeconomic and institutional factors, at the expense of the sample size). From Table 6, while certain differences in the pertinence of the firm-level factors exist across countries, such differences are more strikingly pronounced across sectors of activity. For example, the degree of demand volatility associated with either the market served or with the sales of the firm s main product significantly affects the recourse to temporary labor in manufacturing, but not in services. Interestingly, the higher perceived extent to which either labour regulations or the availability of the educated workforce represent obstacles to businesses, affect only the manufacturing firms use of temporary labour; these factors seem to be irrelevant for firms in services. From Table 7, the opposite picture emerges, in the sense that the relevance of the macro-level determinants of firm s use of temporary labour varies across levels of development, rather than across sectors of activity. We continue consistently finding that the regulation of FTCs, and the level of protection afforded to regular workers by provisions on the substantive grounds for dismissal, are the single most robust factors among the institutional factors affecting the use of temporary labour. In upper-middle-income countries, the higher level of protection afforded to workers on regular contracts by prohibited grounds of dismissals, trial periods, and notice periods also increases firms recourse to temporary labour. In contrast, in lower-income countries, it is the higher protection available in case of contesting the dismissal (redress component) that is associated with higher use of temporary labour; while limitations on the length of trial period tend to reduce the use of temporary labour. That de jure labour regulations have more relevance in the upper-income countries is likely due to these provisions being more binding. While in principle, all firms that take part in the survey are registered, and hence complying with the regulations, such compliance may be partial in some instances. For example, Brusentsev et al. (2012) report that, in Indonesia, firms that pay severance pay when terminating workers, pay the full amount of entitlement at best in half of the cases. Higher level of country income may be positively related to enforcement and compliance, thus reinforcing the relevance of the laws on employment protection. 20

21 6. Conclusions This study has sought to increase understanding of the use of temporary labour by firms in developing countries. Using firm-level data from the World Bank Enterprise Survey, we have attempted to identify the reasons that motivate firms to rely on temporary labour. The use of temporary labour differs widely across firms, with the majority of firms (60%) not using temporary labour at all. Within the 40% of firms that do use temporary labour, temporary labour accounts for 28% of the labour force on average, but there is wide variability in use, with most firms employing around 11% of their workforce on temporary contracts. Our analysis indicates that firms motivation for using temporary labour is similar to that found in studies on industrialized countries. Flexibility is a key motivator as firms with greater volatility in demand and a more competitive environment, as indicated by the market they serve, the degree of competition with informal units, volatility in sales, firm growth and access to credit, all influence the use of temporary labour. Cost considerations are also important in firms decisions about the use of temporary labour, with firms with a higher total wage bill relying more heavily on temporary labour. Where managers felt that labour regulations and an inadequately trained workforce were an obstacle to their business, they also relied more heavily on temporary labour. The results of our analyses also indicated that a higher incidence of training was positively associated with the use of temporary labour, suggesting that temporary labour was used as a screening device. This is also in line with the previous result that firms that had grown in size, including with greater numbers of permanent employees employed, relied more heavily on temporary workers. We also investigated the influence of external variables such as macroeconomic conditions and legislation on the use of fixed-term contracts and employment protection. We find that, unlike many studies on industrialized countries, neither unemployment nor GDP growth influence the use of temporary labour, though temporary employment quite clearly seems to be part of economic development, as its incidence decreases with country s income level. We do find, however, that legislation limiting the use of temporary contracts to specific activities of the firms (e.g., core tasks v. peripheral activities) does have a statistically negative influence on the use of temporary labour. Employment protection legislation, on the other hand, does not seem to have an effect. We then disaggregate our findings across the manufacturing and service sectors and by level of economic development of the countries. We find that demand volatility is 21

22 relevant for manufacturing but not services, regardless of the country s level of economic development. Yet for the national regulations on employment protection legislation we find the opposite: the sector is irrelevant, but there are important differences with respect to the level of economic development of the country. Although the regulation of fixed-term contracts, specifically whether they are prohibited for permanent tasks, has a negative influence on the use of temporary labour regardless of the country s level of income, the indicators on prohibited grounds for dismissal, trial period and notice periods increasing firms recourse to temporary labour are only positively significant in upper-income countries. In lower-income countries, only laws on redress have a positive influence on the use of temporary labour. There has been a growing concern in recent policy debates about an increase in atypical forms of employment and the decline of the standard employment relationship. Although temporary contracts are just one form of atypical contract, little is known about their use in developing countries. This study was an attempt at exploring trends in the use of temporary labour across developing countries in order to better understand firms principal motivations. We confirmed that many firms indeed have a preference for a certain use of temporary labour, notably to respond to fluctuations in demand and save on labour costs. Rules governing termination of temporary contracts, however, which have been the source of extensive debate, have only a limited role in firms use of temporary labour, especially in lower income countries where compliance and enforcement issues may be at stake. Yet at the same time, laws which limit the use of temporary labour do have an effect on firm practices in all considered countries and sectors. From a policy perspective, these preliminary findings are telling as they improve policymakers understanding of business constraints, but also give evidence of the scope of legislation to tailor business practices. Nonetheless, the results are highly aggregated and more research is needed, including through case studies, before any policy prescriptions can be applied. 22

23 References Abraham, K. G. and Taylor, S. K., Firms Use of Outside Contractors: Theory and Evidence, Journal of Labor Economics, 14, 3, pp Baranowska, A. and Gebel, M. (2010) The Determinants of Youth Temporary Employment in the Enlarged Europe: Do Labour Market Institutions Matter?, European Societies, 12, Benito, Andrew, and Ignacio Hernando Labour Demand, Flexible Contracts and Financial Factors: Firm-Level Evidence from Spain. Oxford Bulletin of Economics and Statistics 70(3): Bentolila S. and Dolado J. J. (1994) Labour flexibility and wages: lessons from Spain, Economic Policy, 18, p Bentolila, S. and G. Saint-Paul (1994) "A model of labor demand with linear adjustment costs", Labour Economics, vol. 1(3-4), pages Blanchard, O. and Landier A. (2002). The perverse effects of partial labour market reform: fixed-term contracts in France, Economic Journal, vol. 112(480), F Boeri T.; Garibaldi, P Two tier reforms of employment protection legislation. A honeymoon effect?, in The Economic Journal, Vol. 117, No. 521, pp. F357-F385. Boeri, T Institutional reforms and dualism in European labor markets, in D. Card, O. Ashenfelter (eds): Handbook of Labor Economics, Vol. 4b (Elsevier), pp Booth, Alison L., Juan J. Dolado and Jeff Frank, Symposium on Temporary Work Introduction, Economic Journal 112, no. 480: F181-F188. Brusentsev, V., D. Newhouse, W. Vroman, Severance Pay Compliance in Indonesia. World Bank Policy Research Working Paper Bulow, J., and L. Summers, A Theory of Dual Labour Markets with Application to Industrial Policy, discrimination, and Keynesian Unemployment. Journal of Labour Economics, 4(1), pp Caggese and Cugnat (2008) Financing Constraints and Fixed-term Employment Contracts, The Economic Journal, 118, Capelli, P. H., and J.R. Keller, A Study of the Extent and Potential Causes of Alternative Employment Arrangements. Industrial and Labor Relations Review, 66(4). Pp Davis-Blake, A., and B. Uzzi, Determinants of Employment Externalization: A Study of temporary Workers and Independent Contractors. Administrative Science Quarterly, 38, Devicienti, F., P. Naticchioni, A. Ricci, Temporary Employment, Demand Volatility and Unions : Firm-level Evidence. CEPS-Instead Working paper. Doeringer, P. and M. Piore (1971) Internal labour markets and manpower analysis, Lexington, MA: D.C. Heath and Company. Faccini, R. (2014). Reassessing labour market reforms: temporary contracts as a screening device. The Economic Journal, 124: Eichhorst,W. and Marx, P. (2011) Reforming German Labour Market Institutions: A Dual Path to Flexibility, Journal of European Social Policy, 21, Gebel, M. and Giesecke, J. (2011) Labor Market Flexibility and Inequality: The Changing Skill-based Temporary Employment and Unemployment Risks in Europe, Social Forces, 90, Hatton, E., Temporary Weapons: Employers Use of Temps Against Organized Labour. Industrial and Labour Relations Review, 67(1), pp Hevenstone, D. (2010) National Context and Atypical Employment, International Sociology, 25,

24 Hipp, L., Bernhardt J., J. Allmendinger, Instituions and the Prevalence of Nonstandard Employment. Socio-Economic Review, advanced view, pp Hunt, Jennifer (2000), Firing Costs, Employment Fluctuations and Average Employment: An Examination of Germany, Economica, Holmlund, B. and Storrie, D. (2002) Temporary work in turbulent times: the Swedish experience, Economic Journal, 112, pp Houseman, S. (2001) Why employers use flexible staffing arrangements: Evidence from an establishment survey, Industrial and Labor Relations Review, vol. 55, no. 1, pp ILO, 2015a. Report for a Tripartite Meeting of Experts on Non-standard Forms of Employment. ILO: Geneva. ILO, 2015b. Final report: Tripartite Meeting of Experts on Non-standard Forms of Employment, ILO: Geneva. ILO, 2015c. Employment Protection Legislation Summary Indicators in the Area of Terminating Regular Contracts (Individual Dismissals). ILO: Geneva. Kahn, L. M. (2007) The Impact of Employment Protection Mandates on Demographic Temporary Employment Patterns: International Microeconomic Evidence, Economic Journal, 117, F333 F356. Kahn, L. M. (2010) Employment Protection Reforms, Employment and the Incidence of Temporary Jobs in Europe: , Labour Economics, 17, Kalleberg, A. L., Reynolds, J., & Marsden, P. V. (2003). Externalizing employment: Flexible staffing arrangements in US organizations. Social Science Research, 32(4), Ko, J.J.R., Contingent and Internal Employment Systems: Substitutes or Complements?, Journal of Labor Research, 24(3), pp Knoke, D., and A. Kalleberg, Job Training in U.S. Organizations. American Sociological Review, 59, Martin, W., Pham, C., Estimating the Gravity Model When Zero Trade Flows are Frequent. World Bank manuscript. Lee, B.-H.; Yoo, B.-S The Republic of Korea: From flexibility to segmentation, in S. Lee, F. Eyraud (eds): Globalization, flexibilization and working conditions in Asia and the Pacific (Oxford, Chandos Publishing and Geneva, ILO). Mayer, K. J., & Nickerson, J. A. (2005). Antecedents and performance implications of contracting for knowledge workers: Evidence from information technology services. Organization Science, 16(3), Nielen, S., & Schiersch, A. (2014). Temporary agency work and firm competitiveness: Evidence from German manufacturing firms. Industrial Relations, 53(3), Nollen, S. & H. Axel (1996) Managing Contingent Workers: How to reap the benefits and reduce the risks, New York: American Management Association. Oi, W. (1962) Labour as a quasi-fixed factor, The Journal of Political Economy, vol. LXX, no. 6 (December), pp Organisation for Economic Co-operation and Development (OECD) OECD Employment Outlook (Paris, OECD). Organisation for Economic Co-operation and Development (OECD) OECD Employment Outlook (Paris, OECD). Pierre, G., and Scarpetta, S., Employment Protection: Do Firm s Perceptions Match with Legislation? Economics Letters, 90, pp Pierre, G., and Scarpetta, S., Do Firms Make Greater Use of Training and Temporary Employment When Labor Adjustment Costs are High? IZA Journal of Labor Policy, 2:15; pp

25 Polavieja, J. G. (2006) The incidence of temporary employment in advanced economies: Why is Spain different?, European Sociological Review 22(1): Portugal, Pedro, and Jose Varejao Why Do Firms Use Fixed-Term Contracts? IZA Discussion Papers Bonn, Germany: Institute for the Study of Labor (IZA). Rousseau, D., and C. Libuser, Contingent Workers in High-Risk Environments. California Management Review, 39(2), pp Saint-Paul, G. (1996) Dual labour markets: A macroeconomic perspective, Cambridge, MA: MIT Press. Serrano M. R., Marasigan M. L. C., Pupos V. E. V., Between Flexibility and Security: The Rise of Non-Standard Employment in Selected ASEAN Countries. Retrieved from: March Shapiro, C., and J. Stiglitz, Equilibrium Unemployment as a Worker Discipline Device. American Economic Review, 74, pp Supangco, V., Organizational Determinants of Contingent Employment in the Philippines. Journal of International Business Research, 7(2), pp Ton, Z. (2014) The Good Jobs Strategy: How the smartest companies invest in employees to lower costs and boost profits, Seattle, WA: Lake Union Publishing. World Bank, World Bank s Enterprise Survey. Understanding the questionnaire. Available at: Accessed: July World Bank Enterprise Survey, World Bank s Enterprise Survey. Online Database. Available at: Accessed: July Uzzi, B., & Barsness, Z. I. (1998). Contingent employment in British establishments: Organizational determinants of the use of fixed-term hires and part-time workers. Social Forces, 76(3), Vega-Ruíz, M. L. (2005) La reformal laboral en América Latina: quince años después: un análisis comparado, Ginebra: OIT. 25

26 Figure 1. Incidence of temporary employment, as % of total wage employment, in private sector, circa 2010 Source: Own computations based on the World Bank Enterprise Survey, Notes: Data for 135 countries, for the latest available year, ranging from 2005 for Morocco and Egypt to 2014 for Afghanistan and Myanmar. For the majority of countries (67), data refer to 2009 or Figure 2. Distribution of the firm-level number of temporary employees, as % of total employment, in firms employing at least 1 temporary worker Kernel density estimate Density temporary employees as % of all employees, employed by a firm in fiscal year kernel = epanechnikov, bandwidth = Source: Own computations based on the World Bank Enterprise Survey, Notes: Data for 118 countries (regression sample), all survey years ( ). 26