OJT for New Hires

8/3/2019 OJT for New Hires

1/53

Cornell University ILR School

DigitalCommons@ILR

CAHRS Working Paper SeriesCenter for Advanced Human Resource Studies

(CAHRS)

1-12-1989

On-The-Job Training of New HiresJohn H. BishopCornell University

This Article is brought to you for free and open access by the Center for Advanced Human Resource Studies (CAHRS) at DigitalCommons@ILR. It

has been accepted for inclusion in CAHRS Working Paper Series by an authorized administrator of DigitalCommons@ILR. For more information,

please [email protected].

Please take our short DigitalCommons@ILR user survey.

Bishop, John H., "On-The-Job Training of New Hires" (1989). CAHRS Working Paper Series. Paper 401.http://digitalcommons.ilr.cornell.edu/cahrswp/401
http://digitalcommons.ilr.cornell.edu/http://digitalcommons.ilr.cornell.edu/cahrswphttp://digitalcommons.ilr.cornell.edu/cahrshttp://digitalcommons.ilr.cornell.edu/cahrsmailto:[email protected]://cornell.qualtrics.com/SE/?SID=SV_0vPcL4LVvkEETgEhttps://cornell.qualtrics.com/SE/?SID=SV_0vPcL4LVvkEETgEmailto:[email protected]://digitalcommons.ilr.cornell.edu/cahrshttp://digitalcommons.ilr.cornell.edu/cahrshttp://digitalcommons.ilr.cornell.edu/cahrswphttp://digitalcommons.ilr.cornell.edu/http://digitalcommons.ilr.cornell.edu/http://digitalcommons.ilr.cornell.edu/


2/53

Rev is io n 6 /6 /91

ON-THE-JOB TRA IN INGOFNEW HIRES

John H . BishopCornel l Un iver si ty

W orking Paper # 89-10

C enter for A dvanced H um an R esource StudiesN ew Y ork State School of Industrial and Labor RelationsCornel l Un iver si tyIthaca, N Y 14851-0952607/255-2742

This paper is a subtantial revision of a paper originally prepared for the M ay 1989 Conference onM arket Failure in Training at the La Follette Institute of Public A ffairs, U niversity of W isconsin-Madison, M adison W isconsin. It is based on research that was funded by the U. S. Departmentof Education and by grant # USDOL J-9-M -3-0165 from the Employment and TrainingA dm inistration, U . S. D epartm ent of Labor. The w ork w as also supported under the EducationalR esearch and D evelopm ent C enter program , agreem ent num ber R Il7QOOOIl-91, as adm inisteredby the Office of Educational Research and Improvement, U .S. Department of Education and agrant to Cornell from the Pew Charitable Trust. The findings and opinions expressed in thisreport do not reflect the position or policies of the Office of Educational Research andIm provem ent, the U .S. D epartm ent of Education or the U .S. D epartm ent of Labor. This paper hasnot undergone formal review or approval of the faculty of the ILR school. It is intended to m akeresults of Center Research, conferences, and projects available to others interested in humanreso urce m anag em en t in p relim inary form to encou rag e d iscu ssio n an d su gg estion s.


3/53

ABSTRACTThis paper presents an analysis of a unique data set containing m easures of the tim e devoted

to training during the first three m onths on a job and the productivity consequences of that training.The m ajor findings derived from the analysis of the data on new hire training may be summarizedas fol lows:

* Training investments in new hires are substantial even for jobs that are generallyconsidered unsk il led .* F orm al training provided by specialized training personnel accounts for only a sm allportion of the training received by new hires.* Productivity rises substantially during the first year on the job.* T o fill jobs requiring a great deal of on-the-job training, em ployers prefer applicantsw ho have previous relevant work experience, who are well educated and who havevocational training in a relevant field.* Large establishments invest m ore in the training of their new hires than sm all andm edium sized establishments because (1) they have low er turnover, (2) they havebetter access to capital m arkets, (3) the m arginal product of an hour of training tim eis higher at large establishm ents and (4) training low ers turnover m ore substantiallya t l ar ge e st ab li shmen ts .The elasticity of dem and for training is below unity.** When it is a binding constraint, the minimum wage lowers training investment byroughly 17 percent during the first 3 months on the job and productivity growthby 5 to 10 percent.* Inform al training by cow orkers and training by w atching others do the job appearto have a higher benefit cost ratio than informal training by m anagem ent.* Estim ates of rates of return to training derived from this data should be treatedw ith a great deal of caution. N evertheless, m arginal rates of return to trainingappear to be quite high.* The estimated benefit cost ratio for formal training depends on how the model isspecified. The productivity growth effects of formal training are bigger at largeestablishm ents. Form al training has significantly larger effects on wage grow ththan inform al training. F orm al rather than inform al training significantly increasesthe w orker's propensity to quit.Formal training's tendency to have larger effects on wage growth and quit rates thaninformal training probably results from the fact that formal training is better signaledto the labor market.* The reported generality of training has no significant effects on the its marginalproductivity or on the effects of training on turnover.


4/53

* When train in g is re po rte d to b e h ig hly g en eral, train in g h as a la rg er e ffec t o n w ag eg rowth th an whe n tra in in g is rep orted to b e sp ecific . N ev erth eless, tra in in g th atis reported to be entirely general has m uch larger effects on produ ctivity grow ththan wage growth implying that the labor market treats this training as if itw ere larg ely specific to the firm .These results provide support for the view that workers do not pay the full costs of general

training and do not receive w age increases equal to the full productivity effects of general training.They also lend support to our hypothesis that the outcomes of training, particularly informaltraining, are poorly signaled to the labor m arket. B ecause other em ployers are unaw are of its exactcharacter and unable to assess its quality prior to making hiring decisions, training that istechnically general often becom es effectively specific to the firm and em ployers choose to share thecosts and benefits of investm ents in general training. The second hypothesized reason w hy sharedfinancing of general training may be in the joint interest of employees and employers is the factthat young w orkers are typically liquidity constrained w hile em ployers are not.


5/53

ON -THE-JOB TRA IN ING O F NEW H IRESIf the G erm ans had any secret weapon in the post-1973 econom ic difficulties, it isthe technical competence of their work force, which is in turn the product of theirapp rent ic eship sys tem . --L im precht and H ayes, 1982, p.139.I think that the Japanese education system is not very good employer trainingis m uch m ore effective.--utaka K osai, President, Japan C enter for Econom ic R esearch, 1989The heart of this new [flexible] manufacturing landscape is the management ofmanufacturing projects: selecting them, creating teams to work on them, andmanagin g work ers ' in te lle ctu al d ev elo pment .--R am chandran Jaikum ar, 1986, p. 75.

A growing number of commentators are pointing to employer sponsored training as acritical ingredient in a nation's com petitiveness. A merican em ployers appear to devote less tim eand resources to the training of entry level blue collar, clerical and service employees thanemployers in Germany and Japan (Limprecht and Hayes 1982, M incer and Higuchi 1988, Koike1984, Noll et al 1984, W iederhold-Fritz 1985). In the 1983 Current Population Survey, only 33percent of w orkers w ith 1 to 5 years of tenure reported having received skill im provem ent trainingfrom their current employer (Hollenbeck and W ilkie 1985). Analyzing 1982 NLS- outh data,Parsons (1984) reports that only 34 to 40 percent of the young workers in clerical, operative,service and laborer jobs reponed that it was "very true" that "the skills [I am] learning would bevaluable in getting a better job." The payoffs to getting jobs which offer training appear to bevery high, however. In Parson's study, having a high learning job rather than a no learning jobin 1979 increased a male youth's 1982 wage rate by 13.7 percent. While the 1980 job had nosuch effect, the 1981 job raised wages by 7.2 percent when it was a high learning job rather thana no learning job.

If the payoffs to such jobs are so substantial, why aren't such jobs more common? If onewere to put this question to an employer, he would point to the high turnover rates of youth as theprimary reason why he cannot afford to train new employees more intensely. For Americanworkers with less than one year of tenure, the probability of a separation in the next 12 monthsis 59 percent. Since comparably defined turnover is only 20 percent in the United Kingdom and24 percent in Japan, national differences in turnover could be a major reason for the low levels oftraining investment in the US, if the employer's explanation is right (OECD, 1984, Table 33 and34).


6/53

2The theory of on-the-job training says, however, that if training is general, turnover

propensities should not matter. The worker pays the full costs of the training and reaps the fullbenefits w hether or not there is subsequent turnover, so the decision to undertake training shouldbe independent of prospective turnover. The problem with the prediction that workers pay all ofthe costs of general training is that analyses of large representative data sets generally fail toconfirm it. In Parson's (1985, table 7.6) study, when a youth reported that it was "very true" that"the skills [I am] learning would be valuable in getting a better job", his job paid on average 2.4to 14 percent more than when the above statement was "not at all true" even with an extensive setof controls for schooling and academic achievement included in the model. B ishop and Kang(1988) have conducted another test of this hypothesis in the 1984 follow up of the High Schooland B eyond seniors by regressing the log of the deflated starting w age of the current or m ost recentjob on indicators of the receipt of em ployer sponsored training. H ere again, the jobs offering som etraining rather than none or w hich offer greater am ounts of training paid higher starting w age rateseven w hen a w hole array of hum an capital characteristics w ere controlled. For fem ales the positiveeffect of receiving training on the starting w age w as statistically significant. A dding dummies foroccupation and industry did not change the results appreciably.

It could be argued, however, that these findings do not constitute a decisive refutation ofthe proposition that workers pay all of the costs of general training. H iring decision makers areprobably better at assessing the ability of job candidates than econom etricians w ho are lim ited tothe information in the NLS or HSB data me. The positive association between wages and trainingarises, it could be argued, because workers who are highly able (in ways not observed by theanalyst) are both paid m ore and also recruited for jobs that require large am ounts of training.

U nobserved heterogeneity no doubt contributes to the positive association betw een trainingand starting w age rates, but to transform a large negative structural relationship into a statisticallysignificant positive relationships just described, sorting of m ore able job applicants into hightraining jobs would have to be very powerful indeed. If such a selection process were operating,access to training should depend on ability factors that are visible to the analyst as well as onfactors that are not visible to the analyst. Yet models estimated by Parsons and by Bishop andKang failed to find large effects of ability proxies such as test scores, grades, and being adisciplined student on the probability of receiving training.

One possible explanation of these anomalous findings is that the training is specific andthe employer is financing all of its costs. But standard models of the sharing of the costs ofspecific training do not predict that em ployers pay all of its costs and som e of the new revisionist


7/53

3theories--S alop and S alop's (1976) adverse selection theory--predict that em ployers pay none of thecosts of specific training. A specific training explanation of the these findings is particularlyperplexing w hen to all outw ard appearances the training is largely general.

Empirical tests of the theory of on-the-job training have been severely hampered by theabsence of data on the key theoretical constructs of the theoryngeneral training, specific trainingand productivity growth. Data on wage growth and turnover have been used to test variouspropositions of the theory, but defmitive results have been elusive because the large number ofunobservables result in there being at least two explanations for any given set of phenomena(Garen, 1987). I hope in this paper to overcome some of the lim itations of previous research byanalyzing the first large-scale data set to contain m easures of the tim e devoted to training activitiesduring the first three months on the job, who does the training, the generality of training and theproductivity of the em ployees both during and after the receipt of training.

The paper is organized as follows. The first section describes how the data has beencollected and how the measures of worker productivity and of time devoted to new hire trainingw ere constructed. Section 2 presents tabulations of this data by occupation, establishm ent size,industry, previous relevant w ork experience, age and education. Section 3 contains a very sim pletheory of training investm ent and then offers a m ultivariate analysis of the determ inants of traininginvestm ent. Section 4 analyzes the effect of training on productivity grow th of new hires focusingon how the im pacts of training depend on w ho provides the training, the size of the establishm entand the generality of the training. Section 5 exam ines the effect of training on w age grow th duringthe first 2 years on the job and then com pares these wage rate effects w ith the productivity effectsestim ated in section 4. Section 6 exam ines the effect of training on turnover and prom otions. Thepaper concludes with a summary of the major findings and a discussion of how the fmdings mayilluminate the causes of the lower levels of on-the-job training for new hires in the US than inG erm any and Japan.

I. DATA ON TRAINING AND PRODUCTIVITY GROW TH

The analysis is based on data from a survey of 3,412 employers sponsored by the NationalInstitute on Education (NIE) and the National Center for Research in Vocational Education(NCRVE) conducted between February and June 1982. The survey was the second wave of a two-w ave longitudinal survey of em ployers from selected geographic areas across the country. The firstwave was funded by the U.S. Department of Labor to collect data on area labor market effects of


8/53

4the Em ploym ent Opponunity Pilot Projects (EOPP). The survey encompassed 10 EOPP pilot sitesand 18 comparison sites selected for their similarity to the pilot sites. The ES-202 lists ofcompanies paying unemployment insurance taxes provided the sample frame for the survey.B ecause of the interest in low w age labor m arkets, the sam ple design specified that establishm entsin industries with a relatively high proponion of low-wage workers be over sampled. The taxpaying units were stratified by the estimated number of low wage employees and the number ofestablishm ents selected from each strata w as roughly in proportion to the estim ated num ber of lowwage workers at the establishments in that strata. W ithin strata the selection was random. Thesurvey was conducted over the phone and obtained a response rate of 75 percent.

The second wave attempted to interview all of the respondents in the fust-wave survey.About 70 percent of the original respondents completed surveys for the second wave. Most ofthe respondents were the owner/manager of small firms who were quite familiar with theperformance of each of the firm 's employees. Seventy percent of the establishments had fewerthan 50 employees, and only 12 percent had more than 200 employees. In large organiza tionsthe prim ary respondent was the person in charge of hiring, generally the personnel officer. If theprimary respondent was unable to answer questions about the training received by newly hiredworkers in the sampled job, that pan of the interview was completed by talking to a supervisor orsome on e els e w ith line re sp on sibility .

The employers who received the full questionnaire were asked to select "the last newemployee your company hired prior to August 1981 regardless of whether that person is stillemployed by your company." Only 2594 employers had hired someone in the time framerequested and these em ployers constitute the sam ple used in the study.

The respondent was asked to repon how much time typical new hires for this job spentduring the first three months of employment in four different kinds of training activities: (1)watching others do the job rather than doing it themselves, (2) formal training programs, (3)inform al individualized training and extra supervision by m anagem ent and line supervisors, and(4) inform al individualized training and extra supervision by co-w orkers. For the sam ple of firm sand jobs, the means for the typical worker were 47.3 hours watching others do the job (Tw), 10.7hours for form al training program s (TF), 51 hours for inform al training by m anagem ent (Ts), 24.2hours for informal training by co-workers (Tc). A copy of the relevant ponions of thequestionnaire is available from the author.

A training time index was constructed by fust valuing trainer and trainee tim e relative tothat of workers with two years of tenure in that job and then combining the time invested in


9/53

5training activities during the first three m onths on the job. The em ployers reported that workersw ith tw o years of tenure in the job averaged betw een 22 and 50 percent (depending on occupationand other w orker characteristics) m ore productive than new hires during their first three m onths onthe job. This ratio was calculated for each job/worker category and used to place a relative valueon cowork er time d evote d to tra in in g. 1 The management staff members who p ro vide fo rmal andinfonnal training were assumed to be paid 1.5 times the wage of coworkers. Fonnal traininginvolves four kinds of costs: development costs, facility costs, trainer time and trainee time.Sometimes, it is one-on-one and sometimes it is done in groups but since most of theestablishm ents in this study are sm all, class size w as probably sm all as w ell. Consequently, it w asassumed that when all the costs of fonnal training other than the trainee's time are lumpedtogether--developm ent costs, training m aterials costs and the value of the trainer's tim e--they areabout 25 percent greater than the tim e costs of the trainee.2 W hen supervisors and coworkers aregiving inform al training to a new em ployee, the trainee is alm ost invariably directly involved in aproduction activity. Employers report that for infonnal training, the trainees are typically asproductive while being trained as they are when working alone (Hollenbeck and Smith 1984).Consequently, infonnal training is assum ed to involve only the investm ent of the trainer's tim e.Thus in units of coworker time the value of trainer time is:(1) V alued Trainer Tim e = Tc + 1.5*Ts + TFIn units of trainee time, the time the trainee spends not producing because of training activitiesis :(2) Trainee Tim e = Tw + TFThe total investm ent in training in trainee tim e units3 is:(3 ) T ota l T ra in in g In ve stm en t = Tw + TF + (Tc + 1.5*Ts + TF)/RP.where

RP = the productivity of the average new hire during the first 3 months divided by theproductivity of typical w orker w ith tw o years' tenureThe arithmetic mean of this index is 209 hours, implying that the value of the time invested intraining a typical new employee in the first three months is about 40 percent of the output thatthe trainee can produce w orking full-tim e during the first three m onths on the job.

The survey asked the employer (or in larger firms the immediate supervisor) to report onproductivity of the typical individual hired in the job after two weeks, during the next 11 weeks


10/53


11/53

7clerical jobs and .278 in clerical jobs w ith decision m aking responsibilities (H unter, Schm idt andludiesch 1988). This means that the estimates of the effect of training on productivity growthreported in this paper are probably conservative. The fact that the employer is reporting on thepast productivity of particular employees may also generate biases in data but it is not clear howestim ates of productivity grow th rates m ight be influenced by this problem .

Estimates were also prepared of the short run productivity penalty that results when newworkers are hired. This productivity penalty has two elements: the opportunity costs of trainertime and the lower output of the trainee resulting from the worker's lack of familiarity with thejob and the time devoted to training. W hen expressed in terms of the opportunity cost of thetime of a worker with two years of tenure at the firm , the new hire penalty during the first threemonths on the job is equal to:(4 )(5 )

P roduct iv it y Pena lty = 1 - NPNP = RPTP - Ie + 1.5*T S-i:-.IF520 520

whereNP = productivity net of training cost of typical new hireTP = time a ttempting to p roduce .

T here is som e uncertainty about the correct w ay to aggregate training tim e and productivity grow theffects, so three different estim ates of the penalty are presented. The preferred "liberal" estim ateof the penalty assumes TP = 520 - Tw - TF. This estim ate assum es there is no double countingof training costs: ie. that when the employers told us that new employees were 26 percent lessproductive than w orkers w ith 2 years of tenure, they w ere not factO ring intO that calculation the factthat about 11 percent of the new hires time was spent in a training activity which producedvirtually no output (watching others and formal training). The conservative double countingestimate of training costs assumes that TP = 520. In other words, it is assumed that the lowerproductivity reported for new w orkers reflects in part that portion of their tim e devoted to form altraining and w atching others do the w ork. The ultra conservative estim ate of the penalty uses theconservative double counting assumptions and also substitutes an average of RPand 1 for RP.This estim ate assumes that the reports of productivity growth made by respondent employersexaggerates true productivity grow th by a factor of 2.


12/53

8II. E stima te s o f th e Magn itude o f On -th e-J ob Tra in in gin the First Three M onths of a Job

We w ill begin by exam ining how the costs and consequences of initial on-the-job trainingvary by occupation, industry, establishm ent size, and the previous relevant job experience, age, andschooling of the em ployee. M ultivariate m odels of the determ inants of the length and intensity oftraining are presented in section 3 of the paper.

OccupationThe im pact of occupation on the am ount of on-the-job training typically received by a new

employee is examined in Table 1. The first four rows of the table describe how the averagenumber of hours devoted to four distinct training activities during the first 3 months after beinghired varies by occupation. Even jobs that are thought to require little skill--service jobs--seem toinvolved a considerable amount of training during the first 3 months: an average of 33 hours ofwatching others, 5.7 hours of formal training, 35 hours of informal training by management and17 hours of training by cow orkers. O ther occupations devoted considerably m ore tim e to training.T he distribution of training activities w as sim ilar across occupations, how ever. T he typical traineespent most of his training time watching others do the job or being shown the job by a supervisor.Roughly equal amounts of time were spent in each. Informal training by coworkers is next mostim portant and form al training provided by specialized training personnel accounted for an averageof only 5 to 10 percent of the time new hires were engaged in a training activities.

The fifth row of the table summarizes this information into an estimate of investment intraining during the first 3 m onths on the job. The index valued the tim e that m anagers, cow orkersand the trainee devote to training and expressed it in terms of hours of trainee time. Traininginvestment for service jobs was estimated to be 130 hours implying that the time invested intraining a typical newly hired service worker in the first 3 months was equal in value to about 25percent (130/520) of that w orker's potential productivity during that period. Investm ents in trainingwere considerably greater in other occupations. Retail (and service sector) sales and blue collarjobs had a m ean index of 185 to 200 hours respectively or 35 to 38 percent of the new employee'spotential productivity. C lerical jobs typically required the equivalent of about 235 hours of trainingor about 45 percent of the new w orker's potential output. Professional, m anagerial and non retailsales w orkers required the equivalent of about 300 hours of on-the-job training or nearly 60 percentof the new w orker's potential output.


13/53

SalesProfes- Mana- Not Retail Bluessional geral Retail Sales C lerical Collar Service

H ours Spent in Training in First 3 M onthsW atching others do the job 60.0 65.0 82.8 39.2 50.4 48.1 32.7Fon na l t ra in in g p ro gr ams 9.1 12.1 23.9 8.2 13.5 9.1 5. 7In fo nn al tra in in g b y man ag em en t 76.6 80.4 71.8 48.5 54.6 49.3 35.1I nf on na l t ra in in g b y c o-wo rk er s 31.8 23.0 33.9 23.9 26.2 26.8 16.7In ve stm en t in T rain in g T im e 293 295 350 185 235 200 13 0W eeks to becom e fully trained ifn o p re vi ou s e xp er ie nc e 11.1 13.4 9.2 6.5 6.7 9.0 3.4In cre ase in R ep orte d P ro du ctiv ity (% )Betw. first 2 wks. & next 10 wks. 28% 32% 50% 30% 40% 32% 28%Betw. first 3 mo. & end of year 2 38% 33% 56% 25% 32% 23% 17%New H ire Productivity Penality as a % ofProductivity of W kr. w ith 2 Y rs. TenureLibera l assumptions 69% 69% 74% 51% 60% 50% 39%Conserva t ive assumptions 58% 56% 59% 44% 50% 43% 33%Ultraconserva t ive assumptions 43% 43% 43% 32% 37% 30% 23%

Increase in Real W age in First 2 Yrs. (% ) 5.0% 7.7% 22.6% 9.7% 11.5% 11.5% 3.7%N um ber of cases 95 11 2 76 203 429 649 334

NOTE: Sam ple is lim ited to jobs for w hich all the necessary questions on w age rates, training tim e, and productivity w ereanswered.

TABLE 1TRA IN ING AND PRODUCTIV ITY GROWTH OF TYPICAL NEW EMPLOYEESBYOCCUPATION


14/53

9The sixth row of the table reports the geometric m ean of the answers to the question "How

many weeks does it take for a new employee hired for this position to become fully trained andqualified if he or she has no previous experience in this job, but has the necessary school-providedtraining." Service jobs were reported to require an average of only 3 to 4 weeks of training, retailsales and clerical jobs slightly under 7 w eeks, and professional and m anagerial over 10 w eeks.4

This training seem ed to have the hoped-for result of increasing the productivity of the newemployees. The reported productivity of new employees increased quite rapidly (by roughly athird) during the first month or so at the firm (see row 7). Despite the much greater time interval,the percentage increases between the first quarter and the end of the second year (see row 8) weresmaller than those during the earlier period for blue collar, service, clerical and sales jobs. Forthese occupations training investments and learning by doing seem to be large in the first fewmonths on the job but to dim inish rapidly thereafter. In the higher level, managerial andprofessional jobs, reported increases in productivity w ere larger betw een the third and 24th m onththan in the first few months. This reflects the more prolonged training period for theseoccupations. The occupations w hich devote the least tim e to training--the service occupations--were the occupations with the smallest increase in productivity with tenure. The reportedproductivity of service workers improved an average of 28 percent in the first month or so and afurther 17 percent in the next 21 months. Occupations for which a lot of time is devoted totraining in the first 3 m onths--professionals, clerical w orkers, m anagers and sale representativesoutside of retail and service industries--also seemed to have larger than average increases inreported productivity as the w orker gains in tenure. C lerical w orkers, for instance, w ere reportedto be improving their productivity by 40 percent in the first month or so and by a further 32percent by the end of the second year on the job.

These very rapid rates of productivity growth suggest that the ratio of the productivityincrease to the costs of training (com bining both w orker and em ployer benefits and costs) m ay beextremely high during the first months of employment. For clerical workers the total costs oftraining during the first 3 months was 235 hours or .113 of a year's output by a worker whoseskill level is equal to that of a new employee. Since this figure is an upper bound on theinvestm ent that contributed to the 40 percent gain during the first months on the job, the averagegross rate of return m ust have been above 354% per year (.40/.113). Since the intensity of traininginvestm ent falls w ith tenure at the firm , the cost of training investm ent during the next 21 m onthscannot have exceeded .7875 (1.75*235/520) of a year's productivity by a newly hired worker.This implies that the average gross rate of return to training investments during this 21 month


15/53

10period exceeded 40% per year (.32/.7875). However, marginal gross rates of return to traininginvestment are lower and some of the gain in productivity results from learning by doing nottraining. M ultivariate cross section models of productivity growth which yield evidence on them arginal productivity of training are presented in section 4 of the paper.One of the consequences of the heavy investm ents in the training of new hires is that newem ployees m ake significantly sm aller contributions to the firm 's current output than other w orkerswho have been with the firm for a couple of years or more. The time specifically devoted toformal and informal training activities is not the only penalty incurred when a new employee ishired. In most jobs, skills are developed and refined through practice. Learning by doing as it iscalled may not actually involve spending time away from a directly productive activity. It iscostly, nevenheless, for the new worker is less productive than experienced workers. Thus theproductivity penalty when a new worker is hired has two components: training investm ents andthe lower productivity of the new worker and the tim e others devote to raising the new worker'sproductivity.

Estimates of the shon run productivity penalty when a new worker is hired are presentedin row 9-11 of the table. These numbers provide a rough guide to the m agnitude of the adjustm entcosts associated w ith expan sions carried o ut by hiring additional w orkers rather than b y schedulingextra hours. The other m ajor com ponent of adjustm ent costs--recruitm ent and selection costs--tend to amount to only about 1 percent of a year's output by an experienced worker. The new hireproductivity penalty is much larger. During just the first 3 months, it was equivalent in value forservice workers to an average of about] months output by an experienced worker usingconservative assumptions about double counting. For professional, managerial and salesrepresentatives outside the retail and service sector, the penalty averaged about 1.65 months ofoutput by experienced occupants of the job. The large m agnitude of these costs helps explain w hyemployers tend to hire new employees only when the increase in demand is perceived to be longlasting.

E sta bli shmen t S iz eThe relationship between establishment size and training was curvilinear (see Table 2).

The very largest and very sm allest (10 or few er em ployers) establishm ents invested the greatestamount of time in training. M anagers spent 59 hours training the new employee at the smallestestablishments and only 44 hours at establishments with 11 to 50 employees. The very sm allestestablishment invested 43 percent of a new hire's potential productivity (224 hours) during the


16/53

TABLE 2TRAINING AND PRODUCTIVITY GROWTH OF TYPICAL NEW EM PLOYEEBYEST ABL ISHMENT SIZE

Number o f Emplo yees0-10 11-50 51-200 201+

Hours S pent in T rainingin First 3 M onthsW atching others do the job 48.7 45.4 48.3 55.4Formal tr ain ing p rograms 11.8 7.4 9.2 17.0In fo rmal tra in in g by man ag emen t 59.1 44.4 52.8 48.0In fo rma l tra in in g by c owork ers 23.3 24.3 27.5 32.4Inv estmen t in T rain ing T ime 224 1835 213 248W eeks to becom e fully trained ifno p revious exper ience 8.1 6.4 6.1 8.3In crea se in ReportedP roductiv ity (%)Betw. first 2 wks. & next 10 wks. 29% 33% 37% 49%Betw. first 3 mos. & end of year 2 26% 24% 26% 34%N ew Hire Productivity Penality as a % ofProductivity of W kr w ith 2 Y rs. TenureLiberal assumptions 55% 50% 55% 61%Conservative assumptions 46% 42% 46% 51%Ultraconservative assumptions 34% 30% 34% 37%Increase in Real W age inFirst 2 Yrs. (%) 12.1 7.3 8.7 9.6N um ber of cases 792 678 29 6 12 3

NOTE: Sample is lim ited to jobs for which all the necessary questions on wage rates,tra in in g time , a nd p ro du ctiv ity were a nswere d.


17/53

11fIrst 3 m onths in training w hile the next largest size category (11-50 em ployees) invested only 35percent of the new hire's time. Those with more than 200 employees invested 48 percent of thenew hires time in training. The curvilinearity rem ains when other determ inants of training arecontrolled. Reflecting the pattern of investment in training, wage increases also exhibited acurvilinear pattern being bigger in the very sm allest and very largest establishm ents.

Reported increases in productivity did not, however, have a curvilinear pattern. Ratherthere w as a consistent tendency for the reported increases in productivity to be larger at the largerestablishm ents. The very sm allest establishm ents reported a 29 percent productivity increase in thefirst few months and a further 26 percent increase by the end of the second year. The largestestablishm ents reported a 49 percent increase in the first few m onths and a 34 percent increaseduring the next 21 m onths. Such a dram atic contrast betw een the pattern of training investm ents(input) and training outcom es is unusual. The relationship betw een training investm ent m easuredin time units (line 5 of Tables 1 - 5) and returns to that investm ent, the increase in productivity(line 7 or line 8) is described by:(6 ) ~YR-PlQ = %t;? = AGROR j(S )(T otal T ra in in g In ve stm en t)P1Qwhere AGRO~ is the average gross rate of return on dollars of investment in the training ofstayers at the jib establishment

Sj is the opportunity cost of training time at the t establishmentThe low er percentage productivity grow th to investm ent ratio of tiny establishm ents im plies thateither they have a lower ARj or a lower Sjo It is unlikely that tiny establishments have lowerAGRORj for they have higher turnover and poorer access to capital markets. The probableexplanation o f their sm all %t;? is a lower opportunity cost of time devoted to training (8). Theopportunity cost of m anagerial, cow orker and trainee tim e devoted to inform al training are likelyto be low er because sm all establishm ents are unable to spread the risk of stocastic dem and as w ellas larger establishments and so m ust typicaI1y operate with a higher ratio of capacity (staff onhand) to demand (staff interacting with a customer or engaged in production). Scheduling oftraining is also probably m ore flexible so training can be done during periods of slack w ork w henopportunity costs of trainer and trainee tim e are low .


18/53


19/53

12Rele va nt Wo rk E xp erien ce

The association between training investments that are typically made in new hires andprevious relevant experience of the individual actually hired is presented in Table 3. Jobs whichw ere filled by new hires w ith less than one year of previous relevant experience, typically involvenew hire training investm ent that w as 45 percent of the new hire's potential productivity. For jobsfilled by new hires w ith 10 years of previous relevant experience training investm ent averaged 29percent of potential productivity. This occured in the face of a strong tendency for the jobsobtained by those with a great deal of relevant experience to be jobs that require a considerablylonger training period (see line 5). C learly w hen em ployers filled jobs that require a great deal oftraining if w orkers have no previous experience, they tended to give preference to candidates thatbecause of their previous experience were less costly to train. Note also that jobs filled by newhires w ith greater previous relevant experience received substantially higher w age rates (see line10).

The pattern of productivity and wage increase follow the pattern of investment. Thosew ith the least experience statted out considerably less productive but their productivity grew fromthis lower base at a faster rate. Their wage rates start lower but rise faster. The new hires withmore than 10 years of previous experience, started out more productive and were paid a higherwage. Their productivity rose but at a slower rate and they received no increase in their realwage.6

AgeThe association between the training normally given to new hires and the age of the new

hire is described in Table 4. The relationship was curvilinear. The 25 to 29 year old age groupappears to obtain jobs offering the greatest amount of training to typical new hires--235 hours.T eenagers typically entered jobs requiring about 206 hours and those over forty typically enteredjobs requiring the least training--156 hours. Productivity growth and wage increases seem tofollow an irregular pattern that was roughly curvilinear with a peak in the 20-24 age group. Theaverage wage of a worker with 2 years of tenure in the firm was curvilinearly related to age withthe peak in the 30 to 39 age bracket.Schooling: Type and Am ount

The relationship between type and amount of schooling of the new hire and the on-the-job training typically received by the typical occupant of the job is explored in Table 5.


20/53

TABLE 4TR AININ G A ND PRODUCTIV ITY O F TY PICA L N EW EMPLOYEESBY AGETyp ica l N ew Emplo ye es 16-19 20-24 25-29 30-39 40+H ours Spent m Tram ing in FIrst 3 M onthsW atching others do the job 43.7 52.6 52.0 45.5 38.9Fon na l t ra in in g p ro gr ams 8.9 7.8 17.2 12.1 2.9In fo nn al tr ain in g b y man ag em en t 54.7 52.8 58.4 45.9 43.3I nf ormal t ra in in g b y c owo rk er s 23.8 29.4 23.1 23.3 20.4I nv estm en t in T ra in in g T im e 206 220 235 192 15 6W eeks to becom e fully trained ifn o p re vi ou s e xp er ie nc e 5.6 7.4 7.4 8.2 7.0In cr ea se in R ep orte d P ro du ctiv ity (%)Betwn. f irst 2 wks. & nest 10 wks. 33% 38% 30% 31% 28%Betw . first 3 mos. & end of year 2 27% 29% 24% 23% 23%New H ire Productivity Penality as a % ofProductivity of W kr w ith 2 Y rs. TenureLibera l assumptions 53% 37% 56% 51% 46%Conservative assumptions 45% 47% 46% 42% 39%Ultraconservative assumptions 33% 34% 34% 32% 28%Wage R ateCurrent wage $4.12 5.25 5.84 6.20 5.80Increase in real wage 11.8 12.1 9.3 7.5 3.6Number of cases 346 582 409 33 2 22 9N OTE: Sam ple is lim ited to jobs for w hich all the necessary questions on wage rates, training tim e, a nd p ro du ct iv it y we reanswered.


21/53


22/53


23/53

14III. The Determ in ants of T rainin g

The amount of training that is provided to typical new hires (Ij) is influenced bycharacteristics of the job and the firm w hich influence the increase in w orker productivity resultingfrom training investments (X), the cost of capital to the firm and the worker (r), the rate ofobsolescence of skills (0), the separation rate (s), the share of training that is effectively specificto the firm (I-g), and the opportunity cost of training time (8). Let us assume that the impact oftraining investm ent on the hourly productivity of a w orker can be represented by the follow ing:(7 )(8 )

aP = ~(X )lJ J J where 0 < a < 1~, a-I~ = P (I) = af(X)lj~Ij

The present discounted value of future productivity gains from training a worker who works Hjhours per month is a perpetuity that is discounted at a rate reflecting the cost of capital,obsolescence, the firm specificity of the skill and turnover. It can be expressed as:

00

PV(O = H*P) e-[Ij+Oj+(1-gj)Sj]t =J J~9) Hpfl LJ'LUJr+o+ ( I- g .) s.J J ~ J(10) ~Pv.-J -~l -J

HP'(I L-J"r+o+ ( I- g .) s.J J ~ JSince the marginal productivity of training declines as training increases, the level of traininginvestment is determ ined by the point at which the marginal cost of training investment (8) isequal to the discounted value of its future m arginal products (


24/53

15the length of the training given new employees. The second is a measure of training intensity--the value of the time devoted to training during the first 3 months of a worker's tenure at a firm .Table 6 presents the results of the regressions predicting the logarithm of the two measures oftraining investm ent. M ultiplying a coefficient by 100 gives a rough estim ate of the percentageim pact of a right-hand-side variable.

Both of the measures of training analyzed are indicators of the resource cost of training aparticular individual and not of the learning that has occurred as a result of the training. M ost ofthe determ inants of training that are available in the data set are indicators of demand for and thepayoff to training or are variables that influence both the payoff and costs of training. Factors thatraise the payoff to training w ill increase both the cost of training (input) and the learning (output)that results. W hen one looks across jobs, theory and previous em pirical w ork predict that on-the-job training is com plem entary w ith capital, com plem entary w ith the skill level of other w orkers inthe firm , and com plem entary w ith previous general and occupationally specific training of newhires. A ll of these hypotheses are supported. W orkers who use expensive machinery typicallyreceive a greater amount of training than other workers. The elasticity of response is .066 fortraining intensity and .081 for w eeks to becom e fully trained and qualified. The skill level of otherworkers seem s to have a positive effect on training. Evidence of this is the large positive effectthat the proportion of the w ork force in skilled occupations (w hite-collar or craft) has on training.

Jobs for w hich previous school-provided vocational training is im portant in selecting newhires tend to involve m uch m ore training on-the-job than jobs for w hich previous school-providedtraining is not important. Jobs that are considered to require an extensive general educationalbackground also typically involve longer periods of on-the-job training. These results im ply thatstudents w ho take m ore years of schooling and w ho obtain vocational training typically fm d jobsthat offer greater on-the-job training as well. W hen jobs requiring a great deal of training arefilled, em ployers seem to be particularly interested in hiring applicants w ith a strong educationalb ac kg ro un d a nd re le va nt o cc up atio na l train in g.

It is generally thought that very large establishments invest more training because thediscounted value of future payoffs to training is higher due to low er turnover (s), low er requiredrates of return (r) (resulting from better access to capital m arkets) and lower m arginal trainingcosts due to econom ies of scale (one trainer can teach m any w orkers sim ultaneously). The resultspresented in T able 2 suggests the follow ing additional, hypoth eses, regarding training investm entsat establishments with fewer than 10 employees. New hires in very small establishments arehypothesized to spend m ore tim e in training than new hires at m edium sized establishm ents for tw o


25/53

Table 6THE DETERMINANTS OF THE TRAINING OF THE TYPICAL NEW HIRE

CharacteristicsLog Weeks toB ecom e FullyTrained

Log Train ingIntensityIn F ir st3 Months

Job Charac terist icsImpo rta nc e o f v oa catio na l e du ca tio nSpec if ic vocat iona l p repa ra ti onGene ra l e du ca tio na l r eq uiremen tsClericalSalesR eta il S ale sProfessionalManagerialServiceCraftLog cost of m achineH ours per w eeksT empo ra ry jo bS ta rti ng wageW age at or below legal m inimumEmployer Charac terist icsL og es ta blis he d emp lo ymen tL og em ploym ent squaredL og ra tio firm /e sta blishmen t emp lo ymen tPropo rt ion ski ll edP ro po rti on c ra ftProportion under 25P ro po rti on unio nSales growth last 2 yearsSales grow th last 2 years if positiveEmp lo ymen t G rowthPast em ploym ent grow th if positiveMarket Charac terist icsLog a lte r emp lo ye rs u sin g s ame sk illsL og lab or market sizeS tandard error of estim ateR squaredNumbe r o f o bs erv atio ns

.384***-.020.176**-.257**.046.038-.082.073-.332***.136.081***.0161***-.344***.023-.072

-.206**.0273**-.016.452***.302*-.088-.155-.858***.962***-.035-.306

-.016.017

(3.72)(.67)(2.53)(2.06)(.27)(.21)(.43)(.39)(2.83)(1.19)(3.87)(3.82)(3.63)(1.55)(.85)

(2.19)(2.17)(.60)(3.76)(1.92)(.70)(1.18)(2.70)(2.70)(.17)(.99)

(.79)(.62)1.257.1821659

.522*** (5.58)-.009 (.31).067 (1.37).250** (2.21).645*** (4.26)-.344** (2.11).121 (.71)

.066 (.63).076 (.71).066 (.63).066*** (3.49).018*** (4.58)-.295*** (3.54)-.035*** (2.64)-.170** (2.22)

-.3 17 ** * (3 .7 2).0 49 ** * (4.2 5).038 (1.60).47 0* ** (4.31 )-.127 (.89).237** (2.07)-.114 (.96)-.058 (.20).065 (.20)-.041 (.22)-.270 (.97)

-.04 9* ** (2.6 3).042* (1.70)1.14.1591659


26/53

16reasons. First, their employees must be taught a broader range of skills because very smallestablishm ents have m uch m ore lim ited scope for division of labor. S econdly, the opponunity costsof informal training time are lower because it can be scheduled during slack periods (e.g., whenno customers are in the store) and these periods are more frequent in very sm all establishments.M ultivariate analysis suppons the hypothesis that the size of an establishm ent exens a non-lineareffect on the time that is devoted to training. Large establishments devote m ore time to trainingnew employees than very sm all establishm ents, but they in turn devote m ore time than mediumsized establishments. The establishment size which has the m inimum level of training is 25employees for training intensity and 43 for length of training. Being a pan of a multi-establishm ent firm has no significant im pact on training tim e.

H igh rates of turnover reduce the payoff to training, so w e w ould expect it to be associatedwith lower levels of training investment per worker and to do so panicularly when training isspecific to the firm . Endogeneity prevents our using average rates of turnover as a regressor, butvariables m easuring exog enou s d eterm inants of turnover are available. A s predicted, tem porary jobsoffer significantly less training. M odels estim ated in this data have found that turnover is higherwhen there are many other local employers which make use of the same skills being taught in thejob. As predicted, such jobs offered less training.

Full-time jobs offer more training. If one assumes that hours worked per week areexogenous (ie. hours effects but is not effected by the am ount of training), the elasticity of dem andfor training w ith respect to changes in its m arginal payoff can be calculated from the coefficienton weekly hours of the job. A t the mean number of weekly hours in the sample of jobs, theelasticity estim ate is -.7 (significantly below 0 and significantly greater than -1), im plying that thedemand for training with respect to its rental cost is inelastic. This means that a governmentsubsidy equal to 10 percent of the full marginal opponunity costs of training (or a reduction inturnover or required rates of return w hich had an equivalent im pact on rental cost) would increasetim e devoted to the training of each new hire by 7 percent. An inelastic demand for training alsom eans that holding the job constant, a decrease in learning efficiency (eg. because the w orkers hiredare slow learners or the firm is not very effective in its training) sim ultaneously increases the tim edevoted to training and reduces it's value added. The analysis fmds suppon for this predictionbecause the employers who reponed that it was "difficult to find reliable unskilled workers" andw ho hired m any w orkers under the age of 25 did indeed spend significantly m ore tim e training newhires than other firm s.


27/53

17A number of economists have argued that because the minimum wage prevents workers

from agreeing to a low w age rate during training , it discourages on-the-job training of inexperiencedand unskilled w orkers (H ashim oto 1982, L eighton and M incer 1981). D irect m easures of O JT havenot been available, how ever, and the indirect tests of the hypothesis using w age grow th outcom esas a proxy for training can not be considered conclusive. The hypothesis im plies that holding theskill requirem ents of a job constant, there is a reversal in the sign of the relationship betw een w agerates and training at the minimum wage. Above the minimum wage where wage rates areunconstrained, lower wage rates are associated with more training. The negative effect of them inim um w age on the intensity of entry level training should be visible in the jobs w hose startingwages are at or below the $3.35 minimum that prevailed in 1983.5 Many of these jobs will havehad to be redesigned to minim ize the costs of initial training. This m ight be accomplished byassigning the individual to a very narrow job and teaching only what is absolutely essential toachieve acceptable perform ance in that job. Training in other tasks m ight be postponed and spreadover a longer period of tim e. These hypotheses were tested by including continuous m easures ofthe wage rate and a dummy variable for wage at or below the minimum in the training models.A s hypothesized, both of these variables had significant negative effects on training intensity andno significant effects on the length of training. Sim ilar m odels predicting productivity grow th w ereestimated (without including training investment on the right hand side) and the dummy form inim um w age constraint had a significant negative effect (-10 percent) in the linear specificationand a sm all (-4.7 percent) non-significant negative effect in the logarithm ic specification (B ishop1985 Table 6.2).

IV . Impact of Training on W orker ProductivityNew employees experience dramatic increases in productivity in the first 2 years of

em ployment at a firm . A part of this productivity increase is due to learning by doing and wouldoccur even if no training is provided. Formal and informal training is responsible for a majorponion of the productivity growth, however. In this section, an effon will be made to determ inew hich training m ethods are m ost effective and to m easure the rate of return to training investm ents.

T he 1 98 2 Emplo ye r S urv ey d istin gu is he d fo ur d ifferen t ty pe s o f emplo ye r-p ro vid ed tra in in g:(1) form al training (provided by a training professional), (2) tim e spent w atching others do the job,(3) inform al on -the-jo b training by su pervisors, and (4 ) inform al on-th e-job training by co -w orkers.The impact of training on productivity growth of typical new employees was estimated by


28/53

18regressing productivity grow th during the first 2 years on the hours spent in each training activity,the duration of training and a vector of control variables. Since dim inishing returns are to beexpected, the square of the total cost of training w as included in the m odel. Productivity grow thduring the first 2 years was defined in 2 different ways: the log of the productivity growth ratioand the change in productivity ratings on a 0-100 scale.6

The measures of time spent in specific training activities in the first 3 months on the jobare m easures of training intensity rather than of aggregate training investm ent during the first 2years on the job. Consequently, the reported required length of training--the log of the weeksbefore a new employee becomes fully trained and qualified--was also included in the model. Afull set of controls for job, occupation, and firm characteristics w as included in each m odel. W iththe exception of the wage rate and minimum wage variables, the control variables used wereidentical to the independent variables used in table 7. The specification used was the follow ing:

(13) P2YR-P2WK= AX + a1lnL + a2TF + a3Ts + a4Tc + asTw + ~T2 + uwhere X = a vector of control variables listed in Table 3 (A is a vector of coefficients onth es e con tro l v aria ble s)

lnL = logarithm of the required length of trainingTF = Hou rs devo ted to formal training during the fIrst 3 mon ths ('OOs).Ts = H ours spent in inform al training by supervisors during the first 3 m onths COOs).T c = Hours spent in inform al training by cow orkers during the fIrst 3 m onths ('OO s).Tw = H ours spent training by w atching others do the work during the fIrst 3 m onths('O Os).T = Training Intensity is a weighted sum of the four different types of training where thew eight reflect the assum ed costliness of this form of training.T = 1.8*TF +1.5*Ts + Tc + .8*Tw.

P2YR= Productivity of the typical worker at the end of 2 years. In the linearmodels P2YRis the productivity rating on the 0 to 100 scale divided by 80, them ean productivity rating for workers with tw o years of tenure. In the logarithm icm odels, P2Y Rs the logarithm of the productivity rating plus 5.P2WK = Productivity of the typical worker during the fIrst 2 weeks. In the linear models P2WKis the productivity rating on the 0 to 100 scale divided by 80, the m ean productivityrating for w orkers with tw o years of tenure. In the logarithm ic m odels, P2WKs thelogarithm of the productivity rating plus 5.

The results are reported in Table 7. The regression with the logged productivity growthas dependent variable is in colum n 1. Regressions predicting the linear m easure of productivity


29/53

Productivity Growth Wage Growth ( log)Typical Typical Particular Typical ParticularWorker Worker Individual (2 Y rs) Individual(log 2 Y rs.) (linear 2 Yrs.) Linear (1.1 Yrs)( 1. 1 Y rs ).068*** (6.43) .032*** (6.09) .025*** (4.36) .010*** (2.84) .008**(2.49)

.133*** (3.06) .046** (2.14) .048** (2.10) .043*** (3.13) .027**(2.05)

.130*** (3.85) .067*** (4.01) .043** (2.44) .020* (1.83) .017 (1.61)

.145*** (4.92) .077*** (5.30) .057*** (3 .70 ) .001 (.15) -.002 (.25)

.149*** (7.37) .053*** (5.30) .046*** (4.32) .017** (2.54) .016** (2 .55 )-.0085** (2.27) -.0049** (2.61) -.0050* (2.53) -.0011 (.92) -.OOIl (.97)

.597 .295 .308 .187 .178

.171 .129 .135 .198 .2332Il6 2Il6 2002 1986 1963

Ln Leng th o f T ra in in gH rs. o f T ra in in g in f ir st Q ua rte rF o rm a l T ra in in g ( lOO 's )

T ra in in g b y Sup er vi so rs ( IOO 's )T ra in in g b y Co -wo rk er s ( IOO 's )Wa tc hi ng O th er s ( 10 0' s )T ra in in g I nt en si ty Sq ua re d ( IO ,OOO 's )

S ta nd ar d E rr or o f E st im a teR2Numb er o f Ob se rv at io ns* S ig nific an t a t th e 1 0% lev el (tw o-sid ed )** S ignificant at the 5% level (tw o-sided)** * S ignificant at the 1% level (tw o-sided)

T ab le 7Imp ac t o f T ra in in g o n Wa ge a nd P ro du ct iv it y G rowt h


30/53

19growth are in columns 2 and 3. In both models, the coefficient on the square term is negativeand statistically significant indicating that there are dim inishing returns to training intensity. Whenthe square of total training intensity is included in the model, all four of the linear terms for aparticular form of training have positive and statistically significant effects on productivity grow th.The effect of training intensity on productivity is quite large. An increase in any of the trainingactivities from 0 to 100 hours raises the worker's productivity by 13 to 15 percent in thelogarithm ic m odels and by 4 to 7.7 percent (calculated at the m ean level of productivity at the endof tw o years) in the linear m odels. C learly w hen training intensity is low , increases in its intensityw ill produce large increases in w orker productivity.

The total effect of training on productivity growth was calculated by multiplying the sixestim ated coefficients by m ean values of the corresponding variables. The calculated increase inproductivity was 22 percent (32 percent of the gain over the first two years) in the logarithmicmodel and 12 percent of final levels of productivity (28 percent of the gain) in the linear model.

A n alternative approach to estim ating the im pacts of training is to exam ine the productivitygrowth of particular new hires. Column 3 of Table 7 presents results using productivity data ona particular new hire rather than a typical new hire. M issing data reduces sample sizes by about100. The variance of productivity growth across firms is larger when actual individuals are thedata rather than typical individuals. R squares of the m odels are slightly higher, how ever, becausecharacteristics of the worker and the worker's tenure at the time of the interview are included inthe structural model of productivity growth. In order to minim ize simultaneity problems, thetraining variables used in these models were for a typical new hire rather than for that particularnew hire. Comparisons of the coefficients in column 3 and 2 reveal that substituting data onproductivity grow th outcom es of particular individuals for data on typical hires and controlling forpersonal characteristics does not change the estim ated effects of training.

The impacts of each type of training are remarkably sim ilar. This was not anticipatedbecause som e form s of training (e.g., form al training) have m uch higher hourly costs than others(e.g., watching others do the work), and this was expected to result in the more expensive formsof training having larger impacts on productivity than the cheaper forms. M easured in the unitsof productivity of a worker with 2 years of tenure on the job, the hourly cost of learning bywatching others is .8. Formal training with a cost factor of 1.8 is the most expensive becauseit requires the tim e of both the trainee and the trainer. The cost of inform al training by supervisors(a cost factor of 1.5) and by co-workers (cost factor of 1.0) lies between these two extremesbecause the trainee is engaged in production and only the time of the supervisor and co-worker


31/53

20must be charged off as a cost of training. If one accepts these estimates of the relative costs ofdifferent forms of training, the results imply that informal training by co-workers and trainingyourself by watching others have a higher rate of return than inform al training by supervisors.7F actors fufluencing the M arg inal P ayoff to T rainingEquation 11 im plies that the im pact of an additional hour of training on productivity grow th,P'(I), w ill be higher at com panies w ith high required rates of return (r), high separation rates (Sj),high skill obsolescence rates (~), high opponunity costs of training time (8), and low hours perweek (H). Since workers reap benefits from training even when there is a separation, traininginvestm ents should, in theory, be carried funher (ie. to a point where m arginal benefits are low er)when a job requires general skills rather than specific skills (ie. as g ~ 1). This suggests that anhour of general training will typically have a smaller effect on productivity growth than an hourof specific training. On the other hand, training that is general must be fmanced by the workernot the firm . Since young entry level workers are generally liquidity constrained, the rates ofreturn required by w orkers are likely to be considerably higher than the rates of return required byem ployers. T his has the opposite im plication. T he inability of w orkers to finance general trainingm ay substantially depress such investm ent and m arginal payoffs to such investm ent m ay be veryhigh as a result. The relative im ponance of these tw o effects can be tested by interacting trainingintensity with a m easure of the proponion of skills that are general (g).

Another job characteristic that is likely to influence the marginal product of an hour oftraining is the size of the establishm ent. Large establishm ents are likely to have higher opportunitycosts of training time (8) and to be more efficient trainers (because of economies of scale). Thissuggests that marginal impacts of training may be higher at large establishments than smallestablishments. Form al training is considerably m ore common at large establishments and thissuggests that the marginal impact of formal training may be particularly high at theseestablishm ents. To exam ine these issues, the m odels w ere respecified so as to allow for three-w ayinteractions betw een training intensity, generality of training, size, and the share of training that w asformal, watching others, and informal OJT by a co-worker. The specification used was thefollowing:(14) P2YR-P2WK= BX + bjlnL + b2lnT + b3(lnT)2 + b4ElnT + hs.s.lnT + .!1;E.s.lnT+ b7glnT + uwhere E = lo ga rit hm o f (Es ta bl is hment Emplo yment/1 8.5 )

~ = a vector of shares of training that are formal, watching others, and informal OIT by co-workers. The excluded category is inform al O JT by m anagers and supervisors.g = the proportion of the skills learned useful at other firm s.


32/53

.099*** (10.35) .098*** (10.15)-.0024* (1.76) -.0023* (1.71)

.178 .178

.233 .2381963 1963

T ra in in g We ek sTra ining Intens ityT r. I nte ns ity x S iz eT r. I nte ns ity x S ha re G en er alT r. I nte ns ity x S ha re F ormalT r. I nte ns ity x S ha re C o-Wo rk erTT. I nte ns ity x S ha re W atc h

n ure S q.

a nd ar d E rr or o f E st imat eS qu areb er o f Obs er va ti on s

* S ig nif ic an t a t th e 1 0% le ve l* Significant at the 5% level* S ig nif ic an t a t th e 1% le ve l

T able 8Im pact of T raining on W age and Productivity Grow thof a Particular N ew H ire

Produc tivity Growth(Log)

. 053*** (4 .55 )

. 115*** (9 .35 )

. 083*** (2 .58 )-.0078* (1.73)

.628

.1642002

(two-sided)(two-sided)(two-sided)

.051 ** * (4.44)

. 092*** (6 .82 ). 0099*** (4 .11 ).0043 (.45). 064*** (3 .66 ).025 (1.51). 038*** (2 .70 )

. 104*** (3 .20 )- .0 094** ( 2. 11 )

.622

.1812002

Wage G row th(Log)

.0 07 3* * (2 .1 9)

.0 15 ** * ( 4.2 6).0067* (1.93). 0089*** (2 .92 )-.0001 (.12).0029 (1.08).0 14 ** * (2 .8 6)-.001 (.23).007* (1.82)

Produc tivity Growth(Linear)

.0 19 ** * ( 3.3 8)

.041 ***(5.46)

. 0064*** (5. 48 )

.0018 (.40)

.011 (1.24)

.011 (1.37)

.009 (1.34)

.091** * (5.78)- .0088*** (4 .07 )

.303

.1622002

Wage G row th(Linear)

.0046 (.90)

.0064 (.96)

.0007 (.68)

.0 07 3* ( 1.7 6)

. 022*** (2. 86 )-.004 (.52). 017*** (2. 71 )

. 098*** (6 .31 )- .0 04 5* * ( 2.1 5)

.269

.1261963


33/53

21The results of estimating various versions of equation 14 are reported in tables 8 and 9.

Table 8 reports the results of models predicting the productivity growth of a particular new hirein w hich coefficients b3 and .l1;have been constrained to be zero. These m odels provide evidenceon the effect of the generality of training and establishment size on the marginal product oftraining. T he coefficient on the interaction betw een the generality of training and training intensityis positive but very close to zero. The two effects discussed above appear to have canceled eachother out. It appears that the difficulties that workers face in financing general training are assevere a barrier to investment in general training as high separation rates are to investments inspec if ic t ra in ing .

The coefficient on size interacted w ith training is positive and highly significant in boththe logarithm ic (colum n 2) and linear (colum n 5) m odel of productivity grow th. The logarithm icresults im ply that the elasticity of productivity w ith respect to training is 0.092 at establishm entswith 18.5 employees and about 0.1156 for companies with 200 employees.8 The positive andsignificant coefficient on interactions betw een intensity of training and the share that is part of aform al training program or that is w atching others do the w ork im plies that these form s of traininghave significantly larger effects on productivity grow th than O JT by supervisors, the excludedtraining category. C learly, the earlier conclusion that m arginal rates of return to w atching othersand to co-worker OJT are higher than marginal rates of return to supervisor OJT is pretty robustw ith respect to substantial changes in specification (alternative w ays of defining the independentvariable, alternative w ays of specifying the training variables and the use of productivity grow th ofparticular new hires rather than a typical new hire as the dependent variable). Findings regardingthe payoff to form al training, on the other hand, appear to depend upon specification.

Table 9 presents the results of testing the hypothesis that the size of the establishmentdifferentially effects the rate of return to specific types of training. The m odels presented in thistable included interactions of size w ith (share tim es log total training). W hile the coefficients onthese interactions are not significant in the particular w orker m odels, interactions betw een form altraining and size are significant in the typical w orker specifications. A s hypothesized, the payoffto form al training increases m ore rapidly w ith establishm ent size than the payoffs to other form sof training. These results help explain why form al training programs are more common at largecom panies than at sm all com panies. In the linear typical w orker specification, w atching others dothe work seems to be a less effective learning technique at large companies than at smallercompanies. The coefficients on this variable in other specifications are negative but nots ig nific an tly d iffe re nt from z ero .


34/53

Table 9Im pact of T rain in g on W age and P rod uctiv ity G ro wth

S ta nd ar d e rr or o f e stima teR2Numbe r o f O bs er va tio ns

.591.1892116.291.1562116

.178.2381963* S ign ifican t at th e 10% level (tw o-sided )** Significant at the 5% level (tw o-sided)*u Significant at the 1% level (two-sided)


35/53

22Past efforts to assess rates of return to OJT have focused on the wage payoff to worker

investm ents in training (M incer 1989). This effort is fraught with difficulties, how ever, becauseit is very difficult (a) to measure what employees [as opposed to employers] invest in trainingand (b) to distinguish w age increases caused by training from w age increases caused by selectiveturnover or the need to discourage shirking by back-loading com pensation packages.9 The tO talreturns to employer and employee investm ents (both general and firm specific:) have not beenevaluated because data on productivity effects was lacking. This study has generated tentativeestim ates of both the opportunity costs and the productivity effects of training (general and specific,w orker and firm financed com bined). It would appear, therefore, feasible to calculate m arginalgross rates of return (for general and specific training combined) necessary to cover the cost ofcapital, losses due to turnover and obsolescence. The data was not collected for this purpose,how ever, so there are gaps that can only be filled by som e judicious assum ptions. C onsequently,the estimates of m arginal gross rates of return for each form of training that are reported in table11 m ust be view ed as very tentative results w hich will hopefully be displaced shortly w hen betterdata sets become available. Because the period for which training intensity is measured is muchshorter than the period over w hich productivity grow th is m easured, an assum ption m ust be m adeabout the strength of the correlation between training intensity during the fIrSt 3 months andtraining hours during the rest of the 2-year period. W hen the two year productivity gain of thetypical new hire is being analyzed, a unit increase in a training activity during the first 3 monthswas assum ed to be associated with a further 2-unit increase in that training activity during the restof the 2-year period.lO When the productivity gain during the first fourteen m onths for a particularnew hire is being analyzed, a unit increase in a training activity during the first 3 months wasassumed to be associated with a further 1.2 unit increase in that training activity during theremainder of the first year on the job. Marginal GRORs are the ratio of the increment toyearly productivity generated by a sm all increase training divided by the cost of increased training(A detailed description is in the notes of Table 10).

The estim ated m arginal rates of return dim inish as the intensity of training increases. Themean training intensity for the first 3 months expressed in units of the time of trained workers is148 hours. As intensity during the first 3 months rises from 100 hours to 300 hours (double them ean), the m arginal rate of return (ROR ) for inform al O JT by co-w orkers drops from 43-45 percentto 25-32 percent in the two linear models for typical new hires presented in table 8. The linearmodel's ROR drops from 38-43 percent to 25 percent for watching others and from 17-23 percentto -1 to 10 percent percent for training by supervisors. The ROR of formal OJT is estimated to


36/53

Table 10Sensitivity of M arginal G ross R ates of R eturn Estim ates to SpecificationTraining TrainingFormal by by WatchingTraining ~ervisors Co-Workers Others100 hrs 300 hrs 100 hrs 300 hrs 100 hrs 300 hrs 100 hrs 300 hrs

Table 8Typica l I nd iv idua lLinear 11% -3% 23% 10% 45% 32% 38% 25%Logarithmic 38% 15% 46% 24% 85% 63% 113% 90%Part icu lar Ind ividualLinear 15% -3% 17% - 1% 43% 25% 43% 25%

Table 9Typica l I nd iv idua lLogarithmic 118% 54% 99% 48% 112% 53% 128% 58%Linear 43% 16% 41% 16% 48% 18% 50% 18%

Part icu lar Ind iv idualLogarithmic 156% 68% 109% 52% 130% 59% 146% 64%Linear 46% 16% 38% 13% 47% 16% 46% 16%

Estim ates of the m arginal gross rates of return to increases in the intensity of training at tw o different levels oftraining intensity: a 100 hour investment during the fIrst quarter of the job and a 300 hour investm ent duringthe first quarter on the job. Hourly cost factors are assumed to be 1.8 for formal training, 1.5 for training bysupervisors, 1.0 for training by cow orkers, and 0.8 for w atching others. W hen productivity grow th over 2 yearsfor the ty pical individual is being m odeled, duration adjusted cost factor is calculated by m ultiplying by th e ho urlycost factor by 3 for the reasons given in the text. W hen productivity grow th of a particular individual during thefIrst 14 m onths is m odeled, the duration adjusted cost factor is calculated by m ultiplying the hourly cost factorby 2.2. The results presented in the first panel are calculated by taking the derivative of the estim ated regressionequations reported in tables 4 w ith respect to hours of the specified kind of training, then m ultiplying by 2000,the assumed number of hours worked in a year, and then dividing by the duration adjusted cost factor. As anexample of the calculation, the form ula for formal OJT using the coefficients from the linear m odel in table 4for training intensity (T) equal to 300 hours w as as follow s:[(.0 00 46 - .0 00 00 04 9*T* 2* 1.8 )* 20 00 ] / [3 *1 .8 ] = -.0256 and the cow orker training form ula is:[ (.00077 - .00000049*T*2)*2000]/[3]= .3173. {N ote that the coeffIcients m ust be divided by 100 and 10000 inorder to scale them in hours of training}. The GROR estimates presented in the second panel assume that thefirm has 18.5 em ployees (this zeros out the 5th and 7th term s of equation 3) and that all of the training receivedis of the type indicated. For inform al training by supervisors, the form ula is:(b2 + b/lnT *2)*2000/(T*duration factor) w hich is [(.003 +.0064*4.605*2)*2000] / (100*3) =.4176 at T=I00 forthe linear productivity grow th m odel for typical w orkers. For training by w atching others, the form ula is (b2 +bsw + b3*lnT *2)*2000/(T *d uration factor) w hich is [(.003 + .013*Sw + .006 4*4.605*2)*2000] / (100*3) = .504.Obsolescence of skills and turnover mean that these cash flows do not have an infmite duration and shouldtherefore be compared to the sum of the interest rate, the obsolescence rate and the turnover rate times theproportion of skills that are effectively specific to the firm .


37/53

23drop from 11-15 percent at 100 hours to -3 percent at 300 hours. Estimated rates of return forparticular workers are generally slightly higher than those calculated for the typical worker.Estimated rates of return calculated from models based on logarithmic specifications areconsiderably h igher than those b ased on linear specificatio ns of productivity grow th. A t the trainingintensities that typically prevail during the first quarter, m arginal rates of return seem to be ratherhigh. Since the im pacts of training intensity w ere calculated w hile holding the length of trainingfixed, these G RORs should be view ed as placing low er bounds on the true relationship.

It m ust be rem em bered, how ever, that these m arginal G RORs include cash flow s necessaryto com pensate for tum over and obsolescence and are, therefore, not directly com parable to the realrates of retum to schooling and financial assets that typically lie in the range from 5 to 10 percent.If all training investments are specific to the firm and must, therefore, be written off if workersleave and turnover is high, GRORs of 30 percent or more may be required to induce the firm toinvest in specific training. Lillard and Tan (1986) have estim ated that training depreciates at 15 to20 percent per year. This also would imply that equilibrium in the training m arket would likelyyield marginal GRORs of 30 percent or more. W ith all the uncertainties regarding the bestspecification of the productivity grow th m odel, m easurem ent error in the training variables, thespecificity of the training, turnover rates, and the obsolescence rates, it is my view that robustestim ates of net rates of retum to on-the-job training are not now feasible and w ill not be feasibleuntil better data sets b ecom e available.

R esu lts U sin g In strume nta l V ariab lesThe discussion so far has assumed that the causation runs from training to productivity

growth. It m ight be argued that when one is examining relationships for a typical worker thatfirm s hiring w orkers w ith very low initial productivity w ill fm d it profitable to provide m ore thanaverage am ounts of training. C onsequently, w hen initial productivity is not controlled, there m aybe sim ultaneity bias in our models. A second econometric problem that is likely to be effectingthe results is errors in measuring training. Measurement error is probably biasing down ourestim ates of the effect of training on productivity growth. To test for these biases, we estim atedthe model of productivity growth using instrumented values of training rather than the actualt ra in ing investments .

The X v ariab les used in estimatin g the models p red ictin g inv estmen t in trainin g in T ab le6 we re d iv id ed in to two part s: thos e th at t heory p redic ts d ir ec tly in flu ence p roduct iv ity g rowth andth os e whic h in flu en ce th e c ost o f train in g w ith ou t d ire ctly a ffe ctin g ra te s o f p ro du ctiv ity g row th


38/53

Table 11Comparison of OLS & Instrumental Variable Estimatesof the Impact of Training

Training Training Log W eeks of R2Intensity Intensity TrainingProductivi1Y.-Growth (lOO 's hrs.) Squared(Linear) (lO,OOO's)

OLS .112*** -.012*** .026*** .142Typical H ire (9.3) (6.5) (4.9)2SLS .333*** -.034* -.058* .076(3.1) (1.8) (1.7)

Particular OLS .107*** -.014*** .017*** .152N ew H ire (8.) (6.8) (3.2)( 1.2 Years ) 2SLS .423*** -.058*** -.064* .115(3.6) (2.8) (1.7)

Wage G row th(Linear)OLS .028*** -.0023* .0082** .197T yp ic al H ire (3.5) (1.8) (2.3)2SLS .147* -.025* .010 .181(1.9) (1.9) (4 )OLS .022*** -.0019 .0072** .232Particular (2.8) (1.6) (2.1)N ew H ire(1.2 Y ears) 2SL S -.009 -.0039 .048** .223(.1) (2.1)

* Significant at the 10% level (two-sided)** Significant at the 5% level (two-sided)** * Significant at the 1% level (two-sided)


39/53

24conditional on training. The variables in this latter category were the number of alternativeemployers, dummies for industry, the growth rate of employment, the growth rate of sales, thenumber of employees at the establishment, the size of firm , the wage rate, a dummy for wage ator below the m inim um w age, a dummy for tem porary job, dummies for no probationary period, thelog of length of the probationary period, dummies for not know ing if there is a probationary period,a measure of the difficulty of firing a worker after the probationary period is ended, a measure ofthe importance of seniority in determ ining who is laid off, and characteristics of the local labormarket. These variables were used as instruments for the training variables. This involvesmaintaining the hypothesis that these variables influence the cost of training investm ents, andtherefore, the level and composition of training without influencing the rate at which newem ployees learn. The X variables assumed to have direct impacts on productivity growth weredummies for occupation, the specific vocational preparation (SV P), and the general educationaldevelopm ent (G ED ) that the D ictionary of O ccupational Titles (DOT ) specified is necessary for thejob, percent of work force skilled, percent of work force who are crafts workers, the im portanceof vocational education in selection, cost of m achinery, unionization, hours w orked per w eek, andcharacteristics of the hires (i.e., percent under age 25), and an em ployer response that it is hard tofind reliable unskilled w orkers. W hen outcom es for particular individuals w ere being m odeled, thenew hires' education, sex, and w ork experience w ere included in the structural m odel.

The results from a variety of specifications are reported in Table 11. In most cases,estim ating by intrum ental variables (IV ) rather than O LS has the effect of increasing the m agnitudeof coefficients but reducing their statistical significance. T he IV results also reverse the sign of thecoefficient on length of training. The fact that the IV estim ations increase rather than reduces theestim ated effects of training intensity suggests that m easurem ent error biases are m ore serious thansimultaneity bias and lends support to our general conclusion that marginal rates of return toem ployer-provid ed training are very high.

v. Impact of Training on Wage GrowthThe costs and benefits of investments in on-the-job training are shared by employer and

employee. This implies that jobs with a great deal of training will tend to have lower startingw age rates than w ould otherw ise be predicted and higher w age rates once the training is com pleted.In other words, jobs with a heavy training component--either because it requires great skill orbecause the people being hired for it are com pletely inexperienced--w ill have higher rates of w agegrowth than other jobs. The more general the training the greater w ill be the share of training


40/53

25costs that is paid by the new employee and the greater w ill be the resulting rate of wage growth.Since som e types of training are m ore effective than others, som e are m ore general than others andsom e are m ore visible to other em ployers than others, one w ould expect different types of trainingto have different effects on wage growth. Are the im pacts of different types of training on wagegrowth sim ilar in pattern to their impacts on productivity growth? Or, is the pattern of wagegrow th responses to different types of training m ore influenced by the generality and visibility ofthe specific type of training?

These issues w ere addressed by estim ating w age grow th counterparts to the productivitygrowth models presented in Tables 7, 8 and 9. The first dependent variable studied was the logof the ratio of the firm 's current wage for a worker with 2 years of tenure to the actual startingwage of a person who had recently been hired for the position. M odels predicting this variablecontrol for the effects of wage inflation by including the date of hire and it's square in thespecification. The results are presented in column 4 of Table 7 and column 5 of Table 9.

The second dependent variable is the log of the ratio of the current wage rate (or mostrecent wage if there has been a separation) and the starting wage rate for a particular newemployee who was hired on average 14 months earlier. The models predicting this variable arepresented in column 5 of table 8, column 3 and 4 of Table 8 and column 6 of Table 9. The thirddependent variable is the difference in dollars and cents betw een the current (or m ost recent) w agerate and the starting w age rate of a particular new hire. These m odels control tenure of the w orkeron the date for which wages are reported. The results of predicting this measure of wage growthare reported in colum n 6 of Table 8. A ll three models contain controls for the characteristics ofthe new hire, the occupation, SVP, and OED of the job, percent of craft workers and percent ofskilled workers at the firm , the cost of machinery used in the job, unionization, importance ofvocational training in selection, percentage of the firm 's work force under age 25, and reporteddifficulty in fm ding reliable unskilled w orkers.

The first conclusion that can be drawn from an examination of the wage growth results isthat training does have the hypothesized positive effect on w age grow th. The effect is statisticallysignificant in alm ost all of the m odels. Com parisons of these coefficients w ith the estim ates of theim pact of training on productivity grow th, how ever, reveal that training has a m uch sm aller im pacton wage growth than it has on productivity growth. In table 8, an increase in informal trainingfrom 0 to 100 hours raises productivity of typical em ployees by 13 to 15 percent in the logarithm icmodel and 5.3 to 7.7 percent in the linear model, but raises wage rates by only .1 to 2.0 percent.


41/53

26A doubling of the length of training raises productivity by 2.2 to 4.8 percent, but wage rates riseo nly 0 .7 p erc en t.ll

In Table 8's logarithm ic m odels for a panicular individual, doubling the length of trainingincreases productivity growth by 3.6 percent and increases wage growth by only .5 percent.Doubling the intensity of training, increases productivity growth by 8 percent but raises wagegrowth by only 1.1 percent. Productivity growth effects of training are also considerably greaterthan the wage growth effects in the linear models reponed in column 5 and 6.

For fmdings such as these, the first explanation that comes to mind is that the training isspecific and the firm is paying m ost of its costs and reaping most of its benefits. Since skills arethought to be more specific at large companies, the fact that the gap between the productivity andwage effects of training is largest at big establishments provides further support for the skillspecificity explanation. T he problem w ith this explanation, how ever, is that w hen em ployers w ereasked whether the skills learned on their jobs were specific to the firm, most reported to thecontrary that the skills were useful at other firm s. Funherm ore, the generality of skills taught hasonly very m odest effects on the m agnitude of the wage response to training. W hen training is doneby m anagers and the skills are reported to be entirely general, doubling training intensity raisesproductivity by 6.7 percent but wages by only .8 percent in the logarithmic model reported incolumns 2 and 4 of Table 8. In the linear model in column 5 and 6 of Table 8, doubling trainingraises productivity by 3 percent while increasing wage growth by only .96 percent. Analysis ofdata on the typical new hire produces very sim ilar findings. These results appear to contradict anim ponant prediction of B ecker's theory--w hen training is general, its im pact on w age grow th shouldequal or exceed its im pact on productivity growth. Even though employers claim the skills theyare teaching are general, the labor market is not treating these skills as if they were general. Howcan these puzzling results be explained?

O ne explanation of the phenom enon is that different firm s require different m ixes of generalskills. The firm that does the training concentrates on those skills it needs the most, some ofwhich may not be as highly valued by alternative employers. Skills that would be highly valuedby an alternative employer may not be taught because others on the staff already fulfill thatfunction. As a result, the package of general skills that workers develop are alw ays m ore valuableat the training firm than at other firm s even when each individual skill is correctly perceived to beuse fu l e ls ewhere.

A second reason why the market behaves as if general skills are effectively specific to thefirm is that other employers will generally be ignorant of the exact character of a new hire's


42/53

27general skills and, consequently, w ill often not assign the worker to a job that puts the skills towork. Even when a worker's next job makes use of the general skills learned, there is noguarantee that new hires with better than average skills w ill be offered comparably higher entryw ages. T hese phenom ena have the effect of transform ing som e skills w hich are technically generalint

OJT for New Hires

Documents

Transcript of OJT for New Hires