bs_bs_banner
International Journal of Training and Development 17:1 ISSN 1360-3736 doi: 10.1111/ijtd.12002
Return on investment for workplace training: the Canadian experience Jennifer C. Percival, Brian P. Cozzarin and Steven D. Formaneck One of the central problems in managing technological change and maintaining a competitive advantage in business is improving the skills of the workforce through investment in human capital and a variety of training practices. This paper explores the evidence on the impact of training investment on productivity in 14 Canadian industries from 1999 to 2005. Our productivity analysis demonstrates that in 12 out of 14 industries, training had a positive effect on productivity. However, when the analysis is put within a financial context, the return on investment was positive in only four industries. Faced with negative rates of return, why should managers in most of the industries in the study promote investment in training? Probably the best explanation is that new technology requires an investment in training. The investment in training is necessary just for the firm to maintain its current labour productivity. Employee turnover necessarily impedes the efficacy of training, because trained workers leave, and untrained workers arrive. Thus, training in this instance again is necessary just to maintain current labour productivity.
Introduction One of the central problems in managing technological change and maintaining a competitive advantage in business is improving the skills of the workforce through investment in human capital and a variety of training practices. Canada’s
❒ Jennifer C. Percival, Associate Professor, Faculty of Business and Information Technology, University of Ontario Institute of Technology, Oshawa, ON, Canada. Email:
[email protected]. Brian P. Cozzarin, Associate Professor, Department of Management Sciences, University of Waterloo, Waterloo, ON, Canada. Email:
[email protected]. Steven D. Formaneck, Assistant Professor, School of Business, American University in Cairo, New Cairo, Egypt. Email:
[email protected] This work was supported in part by the Canadian Council on Learning, Work and Learning. Data were supplied by the Research Data Centres Program (http://www.statcan.gc.ca/rdc-cdr/index-eng.htm). © 2013 Blackwell Publishing Ltd.
20
International Journal of Training and Development
competitiveness, however, has been falling in recent years. The World Competitiveness Yearbook (IMD International, 2009) reports that Canada’s competitiveness has fallen from between 11–13th in the world to 16th place in 2009. One way to close the competitiveness gap is through employee training. In Canada, it was estimated that training expenditure was $746 (US dollars) per employee (Conference Board of Canada, 2007), while the USA spent US$1202 per employee (Bersin & Associates, 2009) and the UK US$2728. In many industries, relevant and effective training programmes have been important factors driving the growth of firms and improving their performance. Although all HRM practices support the creation of a competitive advantage through human capital, training is the primary activity which prepares employees and creates a qualified, flexible workforce (Bartel, 1994). In particular, Aragon-Sanchez et al. (2003) argue that the investment in training is still relatively low due in part to the fact that few managers or senior executives evaluate the effects on training at the establishment level and therefore do not know or understand the economic impact of investment in training. Training is also only one way in which employee learning occurs, and therefore the investments made in training require analysis in order to determine if there is sufficient evidence to support a significant return on investment (ROI) for the firm’s investment (Tharenou et al., 2007). This paper will explore the evidence on the impact of training investment on productivity in 14 Canadian industries. We consider both classroom training expenditures and on-the-job training expenditures in establishments spanning a 7-year time frame from 1999 to 2005 to determine the cumulative return on training. We will identify the longitudinal impact of training over time and determine if there are industry-specific patterns of returns. This paper contributes to the existing literature by investigating the link between knowledge accumulation over time and productivity using establishment level data for both manufacturing and service sector industries.
The relationships between training and productivity and training and financial return Many researchers have found that there exists a positive relationship between workplace training and profitability, and workplace training and productivity (Barren & Loewenstein, 1989; Bartel, 2000). While some studies discovered that increases in training can improve labour productivity and offer increased ROI, few papers have addressed the cumulative ROI provided by complementary training practices with respect to productivity at the establishment level (Cassidy et al., 2005). Cassidy et al. (2005) argue that it is important to further analyse and understand knowledge accumulation at the establishment level as this will complement existing knowledge at the macro-level (accumulated industry or country results). It is important to have an understanding of the micro-impact as this is where the investments are made, and these decisions drive the macro-level results. There is a limited understanding of the importance of training in the service sector as well as a limited understanding of factors influencing decisions to choose between type of training activities and modes of implementation (for example, on-the-job or in the classroom). Training has often been criticized as being too expensive, not transferring to specific job tasks and not improving the profitability of the firm (Caudron, 2002; Wright and Geroy, 2001). Turcotte et al. (2003) found that employers tend to believe that the modes of training (particularly, on-the-job vs. classroom) are complementary. They also found that the size of the firm has a significant impact on the incidence and intensity of training provided. Small businesses also tend to invest in a different set of training practices and modes than their larger counterparts (Leckie et al., 2001). This could be due to the constraints on small firms to absorb the possible temporary reduction in productivity that may occur during classroom training. This may result in a higher proportion of on-the-job training practices in smaller firms. Cost is also a major constraint to small firms investing in training as they do not enjoy the ROI for workplace training © 2013 Blackwell Publishing Ltd.
21
economies of scale of larger firms for offering formal in classroom training initiatives (Rabemananjara & Parsley, 2006). The majority of previous studies have focused on macro-level analysis or the impact of training on individuals. Tharenou et al. (2007) conducted a thorough literature review and analysis of the existing evidence of the effects of training on organizationallevel outputs. They found that there is little theoretical development on the impact of training on organizational effectiveness and that the measure of training varied widely in the studies. Tharenou et al. (2007) also found that many of these studies rely on perceptual outcome variables where the number of significant results is much greater than when objective outcome measures are used. This suggests that managers’ perceptions of the impact of training may be much higher than the measurable effects. The authors also express concern that just below 60 per cent of the studies rely on the same surveys to measure training and productivity which may be inflating the number of significant findings. The present study will extend the current set of results [and address many of the limitations expressed by Tharenou et al., (2007) ] by using a different dataset as well as having a relatively large sample size over seven consecutive years. Now, we would like to highlight some analyses which measure objective outcomes relative to training investment. Cassidy et al. (2005) studied knowledge accumulation at the plant or firm level in Irish manufacturing industries. Using only 2 years of data, they found that productivity-enhancing effects of knowledge accumulation are found in domestic firms but not in foreign multinationals. Holzer et al. (1993) found that firm-sponsored training aided in reducing the scrap rate in manufacturing plants (i.e. it provided quality enhancement). Using data from 1983 and 1986, Bartel (1994) found that investment in training in 1983 resulted in improved productivity in 1986. Using data from the US National Employers’ Survey from 1993, Black and Lynch (1996) found support for a positive relationship between investments in training and productivity. Aw and Batra (1998) considered both research and development (R&D) and training in Taiwan. Due to the limitations of their dataset, they were unable to differentiate between training and R&D investment amounts. They did, however, find a positive relationship between R&D and/or training investment and productivity. AragonSanchez et al. (2003) found that training positively impacted future productivity but that the impact in the subsequent year was small. They hypothesized that the effect would continue and might grow but left this analysis for future research (as their dataset only contained 1 year of observations). Our study will extend the literature on the impact of training on productivity. We too will analyse the link between training investment and subsequent productivity. The difference is that our training data spans 7 years over thousands of establishments within 14 industries. Table 1 summarizes training expenditures in Canadian dollars (CAD) by industry sorted by expenditure per employee. Interestingly, average productivity is not perfectly correlated with average training expenditure, with a correlation coefficient of only 27.5 per cent.
Model We are interested in the objective outcome measures related to training investment. Productivity can be measured as firm output or labour productivity (output per employee). The logical choice for a conceptual model is a production function (Blinder, 1990; Lazear, 1998, 2009; Levinsohn & Petrin, 2003). See Appendix I and Lazear (1998, pp. 36–43) for an elaboration.
Yit = f (K it , Lit , Eit , Mit , TSit )
(1)
where i, t represent firm i observed at time t. Y is the firm’s output in physical units, K is the firm’s capital stock, L is the number of workers employed, E is energy use 22
International Journal of Training and Development © 2013 Blackwell Publishing Ltd.
ROI for workplace training
© 2013 Blackwell Publishing Ltd.
23
All industries Finance, insurance Communication, utilities Forestry, mining, oil, gas Information, culture Primary product mfg Capital intensive tertiary Secondary product mfg Business services Transport, warehousing Construction Education, health Labour intensive tertiary Real estate Retail trade
155 452 336 198 130 228 269 108 158 218 137 147 90 151 72
1999
152 463 306 201 258 261 239 201 168 214 162 80 115 69 75
2000
149 351 226 189 337 212 169 311 181 168 154 118 75 136 81
2001
153 300 321 268 227 176 253 157 243 152 153 120 134 107 87
2002
2004
($ per employee) 181 191 516 468 340 405 358 495 403 293 221 312 377 174 226 213 269 225 163 143 124 257 197 187 119 137 87 134 77 113
2003
169 460 353 385 203 270 147 215 184 151 187 205 155 117 84
2005
Table 1: Training expenditures from 1999–2005 (Canadian dollars)
164 430 327 299 264 240 233 204 204 173 168 151 118 114 84
Training 1999–2005
203,122 195,743 165,968 476,044 163,184 201,759 174,122 195,905 163,771 306,504 205,552 101,052 143,357 230,058 120,692
Productivity 1999–2005
measured in KWh but more often in terms of expenditure, M is materials use most often in expenditure form. The term TS is the capital stock of training, which is different from simple investment in training each year. The capital stock measures the total existing value of training expenditures past and present. Unfortunately, we do not have a training capital stock variable (at this time nobody does). The best we can do is to model the annual investment in training at a given location. This is modelled below in equation 2 where TRN represents the annual investment in training. The lagged terms indicate that investments may require time to affect productivity (i.e. learning and integration into existing system effects).
Yit = f (K it , Lit , Eit , Mit , TRN it , TRN it −1, TRN it − n )
(2)
The effects on productivity are conditional on firm-specific factors, such as whether a union is present. Thus, our hypothesis is that training investments, past and current, should increase labour productivity within the workplace. Below is the empirical equation we used for estimation of training’s effect on productivity:1
Yit = α it + βit ∗ Union + δ it ∗ Prof _ Tech + φit ∗ R & D + ϕ i 0 ∗ TRN i 0 + ϕ i1 ∗ TRN it −1 + ϕ i 2 ∗ TRN it − 2 + ϕ i 3 ∗ TRN it − 3 + ϕ i 4 ∗ TRN it − 4 + ϕ i 5 ∗ TRN it − 5 + ϕ i 6 ∗ TRN it − 6 + ε it
(3)
where Union is the percentage of employees who are under union contract, Prof_Tech is the proportion of the total workforce who are professional or technical employees, R&D is a binary variable for whether the workplace conducts research and development, and the variables TRN represent current year expenditures on training (t = 0) all the way back to training expenditures made 6 years ago (t - 6). We include the Union variable as a control variable because Acemoglu and Pischke (1998, 1999) suggest that unionization may encourage firms to fund general training. According to Booth and Zoega (2000) and Zwick (2004, 2006), firms with higher quality employees tend to have more complex tasks and thus invest more in training, and therefore we include Prof_Tech as another control variable.
Research methodology Regression analysis Stepwise regression was performed by industry. Decision makers are ultimately concerned with a financial or capital budgeting interpretation of the impact of training. So below, we outline how the regression output was transformed into net present value and internal rate of return calculations. To calculate the net present value of training investment, some assumptions are necessary. Starting in 1999, an establishment decides to invest in training. The relevant discount rate in 1999 is the prime rate which was 6.44 per cent (Bank of Canada, 2009). The use of the 1999 prime rate is based on the following reasoning. The firm in 1999 has a decision to make – whether to invest in training or not. The Net Present Value (NPV) a priori of that training will be dependent on the expected discount rate over the next 6 years. Thus, the firm cannot know the average prime rate over the 6 years, since years two through six have not come to pass yet. Our assumption is that the firm (somewhat naively) uses the 1999 prime rate to make its calculation of NPV. Furthermore, we assume that training benefits will depreciate for two reasons: workers will forget what they learn over time (Lillard & Tan, 1992) and that knowledge becomes obsolete over time due to technological change 1
As a reviewer kindly pointed out, it would be best to control for initial level of human capital, corporate restructuring, and the macroeconomic environment. Unfortunately, we were not able to include such variables in our regression model.
24
International Journal of Training and Development © 2013 Blackwell Publishing Ltd.
(Bartel & Sicherman, 1995; Dearden et al., 2000; Gerlach & Jirjahn, 2001, Zwick, 2006). We make a simplifying conjecture that knowledge will depreciate on a straight line bases to zero after 10 years. This conjecture is open to debate, however, other methods of computing depreciation of knowledge using geometric decay yield similar results (Almeida & Carneiro, 2008). To calculate the NPV of training for each industry, we must first calculate the derivative to total training with respect to productivity over the period 1999–2005. To do this, the coefficients on the training variables (Trn_Cstt, Trn_Cstt-1, Trn_Cstt-2, Trn_Cstt-3, Trn_Cstt-4, Trn_Cstt-5, Trn_Cstt-6) are summed, and we denote the result as dY/dTrain. To obtain the elasticity (which is interpreted as the per cent change in productivity due to a 1 per cent change in training), we multiply dY/dTrain by the average training expenditure divided by the average labour productivity from 1999– 2005. Then we assume that the elasticity will decay over a 10-year time span. We find the NPV of the training elasticity over the 10 years with an initial investment of $1 (at t = 0). Similarly, the internal rate of return is calculated from t = 0 to t = 10 with an initial investment of $1 in training and a series of decaying training elasticities. Data We use Statistic Canada’s Workplace and Employee Survey (Statistics Canada, 2004) which contains responses from over 5000 Canadian firms over a 7-year period from 1999–2005. Because the survey is mandatory, response rates are consistently above 80 per cent reducing the concerns about sample size voiced by Tharenou et al. (2007). Substantial information on the Workplace and Employee Survey (WES) is readily available online (http://www.statcan.gc.ca/survey-enquete/businessentreprise/8104208-eng.htm). The description here will be necessarily brief; however, the interested reader can check the Internet for more information. From its inception the WES was designed to be a longitudinal survey conducted over multiple years. As such, the questionnaire has undergone only minor changes since its inception. It is a multifaceted survey with two components: an employer questionnaire and an employee questionnaire. For the productivity analysis in this paper, we are only concerned with the employer (establishment). The surveys from 1999–2005 were linked via an establishment level identifier variable called DOCKET. Establishments were sorted by DOCKET in each of the seven surveys, and the files were then merged (not-forprofit establishments were excluded). Only establishments that had remained in the sample frame for the full 7 years were retained in the linked data file. After cleaning, the number of establishments was 3528 with 24,696 observations. There are two types of training practices reported in the survey: classroom training and on-the-job training. Each training method has 13 categories representing various training practices. The 13 categories are the same for both in-class and on-the-job training. These include: orientation for new employees; managerial/supervisor training; apprenticeship training; sales and marketing training; computer hardware; computer software; other office and non-office equipment; group decision making or problem solving; team building, leadership, communication; occupational health and safety, environmental protection; literacy or numeracy; and other training. Chief executive officers were asked to provide their total training expenditures for both on-the-job and classroom training. It was not possible to discern how much was spent on classroom training versus on-the-job training, since establishments only report total training expenditures. Thus, we only use total training expenditures in this analysis. In order to analyse a time lag effect, training expenditures over the previous 6-year period were lagged for each establishment. The means and standard deviations for all variables used in the study can be found in Appendix II.
Results The stepwise regression results by industry are reported in Table 2. Because the regression for business services had no significant variables, it is not reported. As in Table 1, ROI for workplace training © 2013 Blackwell Publishing Ltd.
25
26
© 2013 Blackwell Publishing Ltd.
International Journal of Training and Development
Trn_Cst(-2) Trn_Cst(-3) Trn_Cst(-4) Trn_Cst(-5) Trn_Cst(-6) N R-Squared F-Value Pr>F
Intercept Union Prof_Tech R&D Trn_Cst Trn_Cst(-1)
2,081 0.6756 120.37