Wages and voluntary labor turnover: comparing it workers with other professionals.

ABSTRACT

There is a body of literature suggesting that a new labor contract between employer and employee is emerging which places less value on loyalty and long term employment relationships. Some have argued that the use of computing technology contributes to this by making skills more common and transferable across organizations, thereby encouraging organizations to hire away the skilled workers of their competitors. Concern about voluntary job change specifically among IT professionals was heightened during the IT boom of the late 1990s.

This study examines monthly employment, wage, and job change data from 1998 through the beginning of 2001. IT professionals were compared to other technical and managerial workers. In addition, IT workers employed in the computer services industry (the sector which includes software companies and IT consulting firms) were examined as a distinct subgroup of IT workers. IT professionals who changed jobs were found to have gained a small wage premium, while other professional workers did not benefit from job change. In addition, voluntary job change rates were found to be high for IT professionals in the computer services industry, while IT professionals in other industries were no more likely to change jobs than other professional workers.

INTRODUCTION

During the late 1990's, U.S. IT employers expressed substantial concern over what they perceived to be a substantial shortage of skilled IT workers with rapidly escalating wage rates and high rates of voluntary job change among IT workers with critical skills. Not surprisingly, many IT workers and groups representing them expressed a very different view of the nature and magnitude of this problem. They argued that employers were using the perceived shortage of IT workers to justify visa exemptions to bring in foreign workers at wages that undercut those required to hire and retain experienced U.S. workers. They cited the lack of evidence of rapid wage increases or unusually rapid employment growth across the IT profession, arguing that rapid employment growth was occurring only at the highest skill levels (Bernstein & Hamm, 1998).

The performance of the IT labor market during this period of market stress is important since it is likely that similar pressures will occur in future periods of strong economic growth. Perhaps more importantly, IT wage and turnover behavior during this period are of interest for what they can tell us about the nature of the employment relationship in the IT profession and perhaps for what they can tell us about changes in the employment relationship across all professional occupations. The academic literature provides a number of potential explanations of and prescriptions for dealing with the high job turnover rates of IT workers. Theories range from those that focus on unique aspects of the IT worker and IT jobs to those that suggest that the changing IT job market simply reflects changes, and perhaps represents the vanguard of changes, that are impacting employment relationships across all industries and occupations. In addition, the economic theory of human capital suggests the nature of a worker's skill set (as determined by such things as formal education and formal and informal on-the-job training) is a key factor determining the likelihood of job turnover.

In this study, we will examine alternative theories postulated to explain IT job turnover and determine the extent to which these theories are competing or reinforcing in terms of expected labor market behavior of IT workers. Then, using data from the US Current Population Survey (CPS), we will evaluate the degree to which the alternative theories are supported by US labor market data.

HUMAN CAPITAL THEORY AND TRADITIONAL ECONOMIC THEORY

In a broad sense, wage structures and labor turnover can be explained by the theory of human capital as developed by Becker (1962). According to human capital theory, investments in skill acquisition come at a cost since time that could have been used in production is sacrificed to acquire the skills. A return on this investment is generated in succeeding periods since the acquired skills enhance the productivity of the worker. Individuals can increase their productivity not only through their investments in formal education, but also by learning work skills while they are actually on the job through formal training programs or informal instruction by supervisors or coworkers (Mincer, 1962).

Becker (1962) divides on-the-job training into general training and specific training. General training increases a worker's prospective productivity in a large number of firms. For example, training in the use a word processing package or conflict resolution skills would be expected to have equal value to a broad set of firms. Competition for the services of a worker with these general skills forces the current employer to pay the full value of the skills to the workers or lose her/him to a competing employer. Therefore, the employer presumably cannot afford to pay for such training. The employee reaps all of the returns to general training and must, in a competitive labor market, bear all of the costs of such training. If the skills developed are general, the model of on-the-job training is not fundamentally different than that for a worker deciding whether to bear the costs of formal education in anticipation of higher future earnings after completing their formal education.

Specific training increases the worker's productivity only at the firm that provides the training. For example, one may learn the operating procedures and business rules of a particular firm or learn how use or maintain specialized information systems developed by the firm. The enhanced productivity from this type of knowledge is only achieved if the worker remains with her current firm. For all other firms, her productivity (and therefore her value) has not increased. Once specific training has occurred, the employee has incentive to remain with the firm as long as they receive at least some of the benefits of their improved productivity and the employer will gain from retaining the employee as long as they pay no more than the full value of the employee's improved productivity. Since the worker is now more valuable to their current firm than to competing firms, the worker is less likely to seek a job with another firm and the employer is more likely to make efforts to retain the worker. Specific training requires that the employee and employer to strike a mutually beneficial bargain to share costs and returns of training, which enhances job security for the employee and reduces voluntary turnover for the employer.

When applying human capital theory to information systems (IS) professionals and other high tech workers, it is important to note that specific skills as defined by Becker are far different than specialized skills. Highly specialized skills in mechanical engineering, JavaScript programming, or even in the implementation of a specific ERP software package are likely to be of value to a substantial number of firms constituting a competitive labor market. Thus, such skills are general in Becker's terminology. To avoid confusion, we will henceforth use the term "generally-valued" for Becker's general skills and firm-specific for his specific skills.

TRANSFORMATION OF WORK AND "THE NEW LABOR CONTRACT"

Increased global competition and rapid changes in technology of the 1980's and 1990's set the stage for dramatic changes in the nature of work and the structure of organizations. To accommodate these changes, many researchers believe that a new employment relationship between employers and employees has emerged and that IT is in the forefront of this emergence. Sommers (1995) remarks: "Employees are now loyal first to their individual careers, second to their peers, and third to their company".

Under the old labor contract, the expectation was that good work and loyalty on the part of employees would be appreciated and rewarded with good pay and job security by the employer. However, Chilton and Orlando (1996) argue That this system has "created high cost labor systems with insufficient incentives for producing high quality goods." Rapidly changing needs of business for employees with specialized skills shifts employees' focus away from job security to employment security. Hall and Moss (1998) suggest a new "Protean Career Contract" is emerging in which -"The career is managed by the person, not the organization". Development is self-directed, continuous, and often found in work challenges. The organization is expected to provide challenging assignments, developmental relationships, information, and training opportunities.

A recent content analysis of articles specifically examining the New Employment Relationship in the human relations literature (Roehling, et. al., 2000) identified the requirement for employers "to provide training, education, and skill development opportunities" and for the employee to "assume responsibility for developing and maintaining skills" as the most common characteristics of this new relationship. Thus firms are demanding, and in need of, cutting-edge skills and employees need maintenance of employability through continual training opportunities.

In applying these concepts to IT worker retention, Cappelli (2000) states: "a market-driven retention strategy begins with the assumption that long-term, across-the-board employee loyalty is neither possible nor desirable." Echoing these ideas, Stokes (2000) suggests that neither the employee nor employer desires a career length commitment, so that reward and incentive structures should be built around the desired length, such as, keeping workers through the completion of a project.

The studies cited above suggest that loyalty, job security, and the expectation of long-term employment with the same firm are less important under the new labor contract. Employers are not expected to offer continued employment across a career and employees are not expected to forego short-term wage gains from alternative offers in exchange for a long-term commitment to the firm. At the same time, the new contract requires employers to provide employees access to new skills and requires employees to remain employable by taking the initiative to maintain current skills.

Has computer technology played a special role in the emergence of the New labor contract? Zuboff (1992) argues that computer technology, in addition to automating certain functions, serves to "informate" business enterprises. That is, computer systems make it easier to acquire information about an organization. They formalize and expose data about company operation and policies and procedures. This changes the nature of relationships within an organization, and also makes it easier for those outside the firm to gain an understanding of its systems. If this is the case, computerization is in effect increasing the proportion of generally-valued skills as compared to firm-specific skills.

McConnell (1996) contends that the impact of computers is pervasive across occupation and industry lines--"moving many organizations from a pyramid shape structure to a flatter structure." Quinn, Anderson, and Fenkelstein (1996) argue the ability to leverage information and intellect is leading to new organizational structures which are centered on intellectual skill sets. They argue that "leveraging intellect" by hiring the best people, stimulating them to internalize the information knowledge, skills, and attitudes needed for success, creating systematic organizational structures to capture, focus, and leverage intellect, and demanding and rewarding performance is the key to success. Workers in such a firm will stay because the intellect-nurturing climate helps them increase their value, so they are less likely to be lured away by offers from other firms.

If changes in IT are an important factor in the emergence of new organizational relationships, we would expect the IT labor market to be on the leading edge of this change. The reduction of loyalty under the new labor contract and the increased proportion of generally-valued skills in computer-based systems suggest that IT workers would have higher turnover rates than other professions.

TRAINING, AND THE NEW LABOR CONTRACT

Empirical studies of training suggest that differences in schooling, work, and occupational choice decisions can be explained by unobserved variations in skill level with more able individuals investing more in both formal education and on-the-job training settings because their expected return for skill acquisition is greater (Keane & Wolpin, 1997) and the likelihood of receiving training increases with the education level of the worker (Neal, 1997; Royalty, 1996).

Recent training studies imply that access to training opportunities is often viewed as a form of fringe benefit that firms can use to attract and retain higher quality workers. A survey of US firms (Barron, Berger & Black, 1997) found that there was no evidence that starting salaries were lower when workers received greater on-the-job training, although on-the-job training did cause subsequent increases in wages and productivity. In addition, large firms are much more likely than smaller firms to provide formal on-the-job training than smaller firms and firms offering more numerous and generous benefits to their employees are more likely to provide training (Frazis, Herz & Harrigan, 1995. The incidence of training is also likely to be higher in firms where employee turnover is low (Frazis et.al., 1998) and which use high-performance work systems (such as just-in-time or MRP) (Lynch & Black, 1998).

Traditional human capital theory holds that the employer can only afford to offer training that builds firm specific skill. However, if workers view training opportunities as a significant form of fringe benefit, employers offering strong training opportunities can select the most able workers out of a pool of candidates and pay them the same wage that firms offering less training must pay to less able workers remaining in the candidate pool. This allows employers to in effect recoup training costs by getting a greater stock of ability at the same cost.

Viewing training as a cost of attracting and retaining quality workers is very consistent with the concepts of the new labor contract. The lessening of loyalty under the new labor contract is perhaps, in part, explained by the fact that firms are forced to provide access to training that is not specific to the firm, but enhances the worker's value in the competitive labor market (generally-valued training). Although human capital theory indicates that firms are not willing to pay the costs of providing generally-valued training, firms may pay the direct costs of such training, in return for obtaining and retaining higher quality workers at a given cost.

LABOR MARKET CHARACTERISTICS OF IT WORKERS

A number of studies of the IT field over the past 20 years have attempted to explain job satisfaction and job turnover in terms of the unique characteristics of IT workers and the job characteristics they desire. The motivating factors identified for IT workers are similar to those described for workers in general under the new labor contract.

Couger, Zawacki, and Oppermann (1979) studied motivational levels of MIS managers and other managers. They found stronger growth needs scores and lower social needs scores for MIS managers. A later study (Zawacki, 1992) confirmed these results and suggested than managers must recognize this need for growth and new challenges to motivate IT workers. With respect to more specific aspects of job motivation, Couger (1988) identified the top motivating factors for analysts and programmers as: the work itself, opportunity for achievement, opportunity for advancement, pay and benefits, recognition, and increased responsibility. Job security, working conditions, company policies, and interpersonal relations were ranked lower. Smits, McLean, and Tanner (1993) found that the desire for interesting and challenging work and task variety and autonomy are key motivational factors for high achieving IS professionals and a recent survey IS professionals in the Midwest (Mak & Sockel, 2001) found that the employees' assessments of career development prospects with their current firm had a substantial impact on retention.

It is worth noting that there is a distinction between worker satisfaction with the work environment and worker turnover intentions. Guimaraes and Igbaria (1992) found that job satisfaction was the primary factor for IT workers to feel organizational commitment. However, they found that job satisfaction had no direct effect on turnover intentions.

Stokes (2000) comments that reasons workers stay or leave an organization "vary as much as these professionals' skills and personalities and organizations' cultures and IT functions". He argues for the use of employability training and notes that: "Successful firms accept the marketability of employability". While it may be beneficial to use training as part of a retention strategy, Sheehan (2000) contends that organizations tend to use training for reward purposes rather than as a discipline-driven, reskilling strategy.

Several of the studies reviewed above suggest that employers wishing to minimize voluntary turnover of IT workers should ensure that jobs are varied and challenging and that they involve the opportunity for acquisition of new skills. The new labor contract literature also suggests that those characteristics are important for obtaining high-quality professional workers, but is less clear on retention. Workers who are provided good access to skill improvement opportunities may have higher job satisfaction, but this may not translate into higher retention.

Shortages of workers with specialized skills may force employers of IT workers to provide training in these areas even though these skills are immediately valuable to other competing firms. Firms must balance training costs and an increased risk of losing trained workers against the increase in the ability level of the workers they attract and the enhanced productivity of the workers during their employment with the firm. The choice may become one of: fail to provide training and retain a set of less qualified workers for a longer period of time, or provide training and retain more able and skilled workers for a shorter period of time.

IMPACT ON EXPECTED VOLUNTARY LABOR TURNOVER

The alternative theories described above offer differing expectations with respect to voluntary labor turnover of IT workers. Under the new labor contract with its lessened emphasis on loyalty, it would be expected that voluntary turnover would be increasing across the board. However, if IT workers are on the leading edge of this trend, we would expect their voluntary turnover rates to be higher than those of other professions that are less affected by the new economy. Human capital theory reaches the same conclusion. Job skills that are more generally-valued have a more competitive labor market and thus there is less room for employers and employees to strike a mutually beneficial bargain for the employee's labor. Labor turnover would be expected to be higher the more generally-valued a worker's set of skills. If IT makes the internal workings of a business more transparent, it makes skills more generally-valued and IT workers, on the leading edge of that transformation would be expected to have a higher proportion of generally-valued skills, and thus higher rates of voluntary turnover.

On the other hand, the literature on job motivation of IT workers suggests that they are likely to be better satisfied and thus (probably) less likely to move to another job to the extent that their current job gives them access to the opportunity to learn new skills and the ability to perform interesting and fulfilling work. Thus, workers would generally be less likely to voluntarily leave jobs with more access to learning new skills.

TURNOVER AND WAGE RATE

Also of interest in this study is the impact of voluntary job turnover on the wage of employees. Other things being equal, an employee would voluntarily change jobs only if he or she received a wage premium for doing so. Labor market analyses that have measured wages longitudinally (so that a worker's current wage may be compared to a former wage level for that same employee) have found the expected positive wage effect for a voluntary job change (Keith & McWilliams, 1999; Tomkins & Twomey 2000). However, cross sectional studies that compare the wage levels of workers who have changed jobs with the wages of comparable workers (equalizing for effects of age, sex, race, education level, and total work experience) who have not changed jobs have often concluded that job-changers tend to have lower wages. Light and McGarry (1998), for example, observed a negative impact of job change on wage in a purely cross sectional study based on data from the NLSY. They postulated that jobs could be viewed as "experience goods" where workers will change jobs if the experience of the job is less favorable than expected.

Some researchers have suggested that cross sectional results showing lower wages for job-changers than for job-stayers are due to adverse selection in the "used worker" market (job-changers). Current employers are better informed about the abilities of their employees and thus are able to concentrate their retention efforts on their better workers. Thus there is adverse selection in the market for "used" workers (Acemoglu & Pirchke, 1998; Greenwald, 1986; Waldman, 1984). Workers who quit one job to move to another are likely to be those workers who were passed over for promotion or whose previous employer was unwilling to match an alternative offer. Quitters may increase their salary, but still make less on average than those who do not change jobs.

Countering these adverse selection effects, aspects of the new labor contract seem to imply a measure of positive selection. The emphasis on the employee's responsibility to maintain employability suggests that companies believe that they have numbers of current employees whom they would not be willing to hire on the open market. Presumably, this set of employees would not have access to voluntary job change and only those employees maintaining a current skill set would be able to find alternative job offers.

Another factor contributing to positive selection among voluntary job changers is "poaching." A study of French workers (Hocquet, 2000) found that workers with trained in certain types of skills by their former employers earned a substantial wage premium if they left that employer and went immediately to another employer in the same sector. This suggests that workers are receiving training that produces skills that are valued by competing firms. To cite just a couple of examples from the IT field, United Health Care reported losing "about half" of 600 new IT hires in 1997-98 to other employers (McGee & Davey, 1998) and a Gartner group study (Bittman, 1998) suggested that "Services providers like IBM, EDS and CSAA grow their NT know-how by stealing talent away from their end users with a minimum 35 (pay) percent increase."

The specificity of skill needs in IT is also an important factor that may contribute to poaching behavior. Moore and Burke (2002) suggest that, "At any given time within the IT profession some skill-sets are more marketable than others. In today's environment, skills associated with say, network and database management (such as in Oracle database administration) and with enterprise resource planning (ERP) systems (such as those in people-soft and SAP) are in demand." The sets of highly demand skills also tend to change rapidly. As Memishi (2000) notes "workers proficient in Novell were in huge demand about five years ago and today that's not a problem. There are multiple skill sets and so if something new comes down the pike and people aren't trained in that application, that's when there's a shortage."

Overall, theoretical models and past empirical results do not provide a clear-cut prediction as to the effects of job change on the wage level. Some evidence suggests that in purely cross sectional studies (as will be the case for our analysis) wage rates of job changers have tended to fall below those of job-stayers. However, the new labor contract research suggests that this is changing and that positive selection factors, including incentives for poaching those workers with the most highly desired skill-sets, may begin to outweigh the adverse selection factors leading to higher wages among voluntary job-changers.

CPS SURVEY DATA

The empirical component of this study is based upon data from the Current Population Survey (CPS which is household based. If a person changes their residence, they drop from the survey and the new occupant of that residence is surveyed. Thus, only job changes not involving changes in residence are included in this study. Workers can be in the survey for 8 months. However, they are asked questions about their wages only in months 4 and 8 in the survey. The data set used here is based upon data for the 4th and/or 8th months in the survey plus data measuring voluntary job change over the prior 3 months--workers who reported changing employers in any one of the 3 previous months were coded as job changers. Individuals are included in our survey data only if they were surveyed throughout months 1 through 4 or throughout months 5 through 8. The data set used in this analysis covers a three year period (April 1998 through March 2001). It is restricted to individuals 25 years of age and above, working full time, and reporting weekly earnings of at least $150. This allows us to focus the analysis on full-time paid employment of workers who have completed their initial formal education.

Most of the variables used are standard demographic and economic measurements. The wage variable used is an average weekly wage estimate compiled by the CPS by converting wage values reported over differing time periods (hourly, monthly, annually, etcetera) to a weekly bases (and, if necessary constructing values) using a standard procedure (US BLS, 2000). This wage is top-coded at $2881.61, or approximately $150,000 per year. Following Autor, Katz and Kreuger (1998), top coded wages are reported as 1.5 times the top code value or $4322.42 per week to adjust for truncation bias. In the analyses that follow, less than two percent of any of the groups studied had top coded values.

Comparison of IT workers with workers in other professions is the major focus of this study. The detailed occupation code of workers was used to separate IT workers (codes 64 and 229) from other business and technical professionals. Professional workers and technicians in all other non health care related fields were identified as the base group of other comparable professional workers.

Data identifying the industry in which a worker is employed are also of interest. Industry code 732 Computer and Data Processing Services is the industry category that includes software and computer consulting services companies. Arguably, firms in this industry tend to provide their employees with more access to new technology and provide workers with greater professional growth opportunities than companies that are not in computer related industries. Comparison of IT workers in this IT related industry with IT workers in less IT focused industries allows us to determine whether this industry has distinct wage and turnover behavior.

EMPIRICAL RESULTS

Descriptive statistics for a number of qualitatively measured demographic characteristics that might be expected to impact wage and voluntary labor turnover rates are presented in Table 1.

The education level of IT workers appears to be somewhat higher than that of other technical occupations and management related occupations. Distribution of employment across industries shows that just over one-third of IT workers are employed specifically in the Computer services industry, which constitutes only a very minor component of employment in the other occupation categories. IT workers tend to be more concentrated in large cities than workers in the other two categories. Finally, female workers are much more prevalent in the management related occupations than in the IT and other technical occupation groups, whereas, Black and Hispanic workers are about equally represented across the three occupation categories, but are underrepresented across all of these occupation groups relative to their prevalence in the general population.

Age is measured in individual years in the CPS data set. Table 2 shows the distribution of employment by age grouped in 5-year age ranges and compared across occupation groups. Clearly the proportion of workers in the age groups up to age 40 is substantially higher in IT related occupations, than in the other occupational categories.

Table 3 presents summary data showing the average weekly wage for workers at each age range across the occupation groups. The data suggest that young workers in IT related professions experience a substantial wage premium compared with the other occupation groups. However, this premium shrinks substantially with age.

We have not yet presented data about the rates of voluntary job change. Since these rates are one of the principal targets of our analysis, we present the job change data in Table 4 along with results of a statistical test of whether the proportions are equal across occupation groups.

Table 4 shows that the 3 month rate of voluntary job change is between 1 and 1.5 percent higher for IT related occupations than that for other technical workers and for management professions and these differences are statistically significant (based upon separate comparisons of IT worker to each of the other occupation groups).

In Table 5 similar statistics are presented with a breakdown of the turnover rate of workers in the computer services industry versus all other industries, as well as a breakdown of IT versus other professional occupations.

It is clear that job turnover in the computer services industry is much higher than the average across other industries for both IT workers and other professional workers. It appears that all of the difference in job change rates found in Table 4 is due to the strong prevalence of IT workers in the computer services industry. Within industry comparisons showed no significant difference in the rate of job change. IT professionals working in the computer services industry experienced a job change rate of 9.22 percent over the three month observation period, but employees from other professions working in the computer services industry experienced an even higher, 10.20 percent, job change rate. In all other industries, IT professionals experienced a job change rate of 5.37 percent compared to 5.23 percent for other professions.

Table 6 presents regression results for determinants of the rate of job change. Job change is a dichotomous variable. A worker either changed jobs over the prior 3 months (1) or did not change jobs in the prior 3 months (0). For a number of reasons, standard linear regression is not appropriate when using a dichotomous dependent variable. Most prominently, the error term cannot be normally distributed.

The logit function is a commonly used method that produces maximum likelihood estimates for equations with dichotomous dependent variables (Green, 1997). The Logit function essentially makes the log of the odds ratio (the ratio of the probabilities of the two outcomes of a dichotomous variable) the dependent variable and performs a "standard" linear regression on this transformed value. In our case the odds ratio is the probability of changing jobs divided by the probability of not changing jobs. This is illustrated in the following equation where ln is the natural logarithm, pjc is the probability of changing jobs, a is an intercept term X is a vector of independent variables, B is a vector of coefficients on those variables and e is the error term.

ln([p.sub.jc]/(1 - [p.sub.jc])) = a + BX + e

Taking the antilog of the equation above yields:

[p.sub.jc]/(1 - [p.sub.jc]) = expa x expB x expX x exp e

The B coefficients above can be interpreted as percentage changes in the odds-ratio of the dependent variable, and this interpretation is appropriate for the coefficients of Table 6.

Data for all of our occupational categories were pooled into a single group for this analysis with occupation and industry groupings being treated as independent variables. The likelihood ratio test for this model shows it to be statistically significant. Individual parameter estimates and their standard errors are also shown.

Age, and the square of age are the only continuous variables in this model. They are statistically significant with opposite signs. These coefficients suggest that the odds of a worker changing jobs initially increase at a decreasing rate with age and eventually begin to decrease with age.

The remaining coefficients in Table 6 are based upon sets of dummy variables. In each category there is a base group and a set of coefficients reflecting how a worker's probability of job change is affected by belonging to some other demographic group. The base group is white male workers in non-IT related technical professions who live in cities between 500,000 and 5 million in population and work in the services and government industry sector (but not the computer services industry). The results suggest that college graduates are less likely than workers at other education levels, both higher and lower, to change jobs. Workers in manufacturing industries are significantly less likely to change jobs and workers in the computer services industry are substantially more likely to change jobs than workers in other industries, which is consistent with the base proportions in Table 5 Not surprisingly workers in small cities are less likely to change jobs (without relocation) and workers in large cities are more likely to do so. The race and sex of workers does not appear to have a significant impact on their job change characteristics. Finally, consistent with the Table 3 results, the worker's occupation category does not significantly impact the rate of job change when the impact of working in the computer services industry is treated as a separate factor. Together the results of tables 4 through 6 suggest that the higher rates of voluntary job change for workers in IT occupations are almost entirely due to the high incidence of job change in the computer services industry where over a third of IT professionals work.

Some of the results of Table 6 seem at first counter intuitive and are inconsistent with past studies of voluntary job change. However, it must be remembered that only job changes not involving relocation are included in our sample. With respect to age, younger workers may change jobs more frequently overall (the expected result) but a higher proportion of their moves may involve relocation. Rates of job change have generally been found to decline with increases in the education level, but the markets for workers with advanced degrees may be too small to allow them to easily change jobs without relocation.

Table 7 summarizes linear regression results examining the determinants of wage rates. In this case separate results for IT occupations, other technical professions, and management related professions are shown. Values for the log of weekly wage were found to be approximately normally distributed, while untransformed wage values were not. Thus, the log of weekly wage was used as the dependent variable in this analysis.

In standard linear regressions with a dependent variable that is measured in log form, coefficients can be interpreted as percentage impacts on the dependent variable in non-log form. Thus, for instance, the--.15896 coefficient for female workers in IT occupations means that, other things being equal average weekly wages of female workers are about 16 percent less than those of their male counterparts.

The regression model is statistically significant for each occupation group. Adjusted R-Square values are modest suggesting, not surprisingly, that there are many factors influencing individual wage rates which are outside the scope of the broad set of demographic factors available for this study.

The age and age squared coefficients are similar across the occupation groups and reflect a pattern of wage rates increasing at a decreasing rate in younger age levels and decreasing in absolute terms for older workers. The somewhat larger magnitudes (in an absolute sense) of these coefficients among workers in IT occupations means that wages are expected to rise more rapidly with age initially, but that this rate of growth will also drop off more rapidly as age increases.

The month variable is an index that begins with a value of 1 in April of 1998 and increases by one for each month thereafter. It is designed to capture the general time trend of wages. This rate is about 0.4 percent per month or 4.8 percent per year for workers in IT occupations. The rate for management occupations is almost identical to the rate for IT, while that of other technical occupations is slightly, but not significantly lower.

The Education level parameters suggest that the wage premiums for increased education are systematically (and in many cases significantly) lower for IT occupations than for the other technical and management occupations. This may suggest that formal education is less important, relative to subsequent on-the-job and nonacademic training and experience among IT professionals.

Industry differences are less pronounced for IT workers than for the other occupations. However, it should be noted that the services and government sector (the base sector in this analysis) includes the computer services industry that has a large number of highly paid IT workers. This industry will be examined in detail in Table 8 below. City size has consistent effects across the occupation groups workers in small cities on average earn less while workers in large cities earn more. The wage penalty for living in a small city is somewhat smaller for non-IT technical occupations. The wages of women and minorities are systematically lower across all of the occupational categories. The wage differential for female IT workers is significantly smaller than that for the other two occupation categories, while the wage rate differential for Blacks is a bit larger than that for management related professions. Finally, the job change parameter shows that workers who had changed jobs in the previous three months earned slightly less than those not changing jobs among both management and non-IT professional occupations. Neither of these coefficients was statistically significant. For IT related occupations, on the other hand, job changers on average earned just over 5 percent more than those who had not changed jobs and this difference was statistically significant.

The results indicate that workers in IT occupations could expect a premium for changing jobs that was not present in other occupations. At the same time, the rate of increase in wages over time was not significantly greater for IT occupations. Together these results show a substantial incentive for job change among workers in IT occupations. Perhaps positive selection factors, including poaching effects, were on balance more important than adverse selection factors among job changers in IT occupations. This would account for the combination of a wage premium for job changers with very little increase in the overall wage trend.

In Table 8, we further break down the data, for workers in IT occupations only, to see if the wage determinants in the computer services industry differ from those for other industries. Results suggests that wages of workers in the computer services industry tend to rise more rapidly with age initially, but to fall of more quickly at older ages. With respect to the effect of job change on wage, job changers in the computer services industry gained a premium of 5.8 percent while IT workers in other industries gained about 3.7 percent. However, the job change coefficient was not significantly different from zero for either group. The time trend of wages for workers in the computer services industry is a bit stronger than that for IT workers in other industries (5.6 percent versus 4 percent). Differences across the remaining demographic groupings provide few systematic patterns. Overall, the results in Table 8 suggest that factors affecting wage rates of IT professionals in the computer services industry do not differ substantially from the factors affecting IT professional's wages in other industries.

CONCLUSIONS

This paper has examined alternative theories of the impact of training and skills on the structure of wages and voluntary job change. Human capital theory suggests that firms can only afford to pay for firm-specific training. However, the dynamics of the IT market suggest that firms are often forced to pay the direct costs of specialized training that has a very competitive market. The new labor contract model suggests that firms should view such training as a cost of attracting more able employees. By providing frequent opportunities for performance appraisal and salary adjustments, the firm can recoup a sufficient return on this enhanced ability to recoup its training costs, although job change rates are expected to be higher overall due to the highly competitive skill sets developed.

Job satisfaction surveys within the IT industry have found that IT workers are generally more satisfied with their jobs when they have access to learning new skills. However, this greater job satisfaction may not translate into longer retention, if the skills learned have a highly competitive market.

Generally, the empirical results presented here support the concepts of the new labor contract and suggest that IT workers, or at least IT workers in the computer services industry, may be on the leading edge of the movement to this form of relationship between employees and employers. The higher rate of voluntary job change among computer service industry workers and the positive returns to job change among IT workers (suggesting a predominance of positive selection factors among job changers) are both compatible with new labor contract ideas that: neither workers nor employers desire career length relationships and that the employer is expected to provide opportunities for training and skill development.

The nature of available data has imposed major limitations on the hypotheses that could be developed and the conclusions that could be drawn in this study. The ideal data set for the analyses presented here should: follow workers if they relocate due to job change, provide estimates of the wage before and after job changes, track cumulative job changes over a longer period of time, and provide data about specific skills developed and training provided on the job. The costs of developing such an ideal data set at large scale may be prohibitive, but it should be possible to test the impact of some of these elements through survey research and the use of other predefined data sources.

It is also important to examine the responsiveness of the high-tech labor market across the business cycle. This study covered a time period that was predominantly part of an economic boom. If the new labor contract is rooted in changes that are making skills more generally-valued and less firm specific, firms will have less incentive to retain skilled workers through economic downturns, since the ability to share in the returns of a firm-specifically trained workers skills during good times provides incentives for employers to retain them during recessions.

Finally, this paper and the literature it reviews suggest that the process of skill development in the new economy is important area for future research. If skills are indeed becoming more generally-valued, who will pay for them and how will they be provide? Will firms be less willing to bear the costs of obtaining these skills (as economic theory suggests)? If the individual must bear all, or nearly all, of the costs of acquiring skills will there be substantial under-investment in these skills? What is the role of the higher education community in the provision of these skills and the re-skilling of workers as technology advances? All of these are important questions that need to be addressed as we work to understand the implications of the changes in work relationships implied by the new labor contract.

REFERENCES

Acemoglu, D. & Pirchke J. (1998), Why do firms train? Theory and evidence. Quarterly Journal of Economics, 113(1), 79-119.

Autor, D., Katz, L. & Krueger, A. (1998) Computing inequality: have computers changed the labor market? Quarterly Journal of Economics, 113, 1169-1212.

Barron, J., Berger, M. & Black, D. (1997) Do workers pay for on-the-job training? Journal of Human Resources, 34, 235-252.

Becker, G. (1962) Investment in human capital: a theoretical analysis. Journal of Political Economy, 70(5), 9-49.

Bernstein, A. & Hamm, S. (1998) Is there really a techie shortage? Business Week, Issue 3584, 93.

Bittman, T. (1998) Do you know where your NT specialists are? Insurance and Technology, 23(8), 7.

Cappelli, P. (2000, January) A market-driven approach to retaining talent. Harvard Business Review, 103-111.

Chilton, K. & Orlando, M. (1996) A new social contract for the American worker, Business and Society Review, 24, 23-26.

Cougar, D., Zawacki R. & E. Oppermann, E. (1979) Motivation levels of MIS managers versus those of their employees. MIS Quarterly, 3(3), 47-56.

Cougar, D. (1988) Motivators versus demotivators in the IS environment. Journal of Systems Management, 39(6), 36-41.

Frazis, H., Herz, D. & Harrigan M. (1995) Employer-provided training: results from a new survey. Monthly Labor Review, 118(5), 3-17.

Frazis, H., Gittleman, M., Harrigan, M. & Joyce, M. (1998) Results from the 1995 survey of employer-provided training. Monthly Labor Review, 121(6), 3-13.

Green, W. (1997). Econometric Analysis, 3rd Edition, Upper Saddle River, New Jersey: Prentice Hall.

Greenwald, B. (1986). Adverse selection in the labour market. Review of Economic Studies, 53, 325-347.

Guimaraaes, T. & Igbaria, M. (1992) Determinants of turnover intentions comparing IC and IS personnel. Information Systems Research, 3(3), 273-303.

Hall, D. & Moss, J. (1998, Winter) The new protean career contract: helping organization and employees adapt. Organizational Dynamics, 23-36.

Hocquet, L. (2000) Vocational training and the poaching externality: Evidence for France. In Education and Training in a Knowledge-Based Economy, Heijke H.& Muysken, J. (Eds.), New York: St. Martin's Press.

Keane, M. & Wolpin, K. (1997). The career decisions of young men, Journal of Political Economy, 105(3), 473-522.

Keith, K. & McWilliams, A. (1999). The returns of mobility and job search by gender. Industrial and Labor Relations Review, 52(3), 460-477.

Light, A. & McGarry, K. (1998). Job change patterns and the wages of young men, Review of Economics and Statistics, 80(2), 276-286.

Lynch L. & Black, S. (1998) Beyond the incidence of employer-provided training. Industrial and Labor Relations Review, 52(1), 64-81.

McConnell, S. (1986) The role of computing in reshaping the work force. Monthly Labor Review, 119(8), 3-5.

Mak, B. & Sockel, H. (2001) A confirmatory factor analysis of IS employee motivation and retention, Information and Management, 38, 265-276.

McGee, M. & Davey, T.(1998) IT skills drought worsens. Information Week, 665, 30.

Memishi, R. (2000) Facing the shortage of information technology professionals. Electronic Commerce World, 10(1), 24-29.

Mincer, J. (1962) Laborforce participation of married women. In Aspects of of Labor Economics, Davis, G. (Ed.) NBER Conference Series No. 14, 63-97.

Moore, J. & Burke, L. (2002) How to turn around the 'turnover culture' in IT. Communications of the ACM, 45(2), 73-78.

Neal, D. (1997) The link between ability and specialization: an explanation for observed correlations between wages and mobility rates. Journal of Human Resources, 33, 173-200.

Quinn, B., Anderson, P. & Finkelstein, S. (1996) Leveraging intellect. Academy of Management Executive, 10(3), 7-27.

Roehling, M., Cavanaugh, M., Moynihan, L. & Boswell, W. (2000) The nature of the new employment relationship: A content analysis of the practitioner and academic literatures. Human Resource Management, 39(4), 305-320.

Royalty, A. (1996) The effects of job turnover on the training of men and women. Industrial and Labor Relations Review, 49(3), 506-521.

Sheehan, L. (2000) Variables in the hiring equation: ensuring the right candidate is hired. Information Systems Management, 17(3), 21-32.

Smits, S., McLean, E. & Tanner, J. (1993) Managing high-achieving information systems professionals. Journal of Management Information Systems, 9(4), 103-120.

Sommers, P. (1995) Workforce 2000. Industry Week, 244(10), 21.

Stokes, S. (2000) Attracting and keeping IT talent. Information Systems Management, 17(3), 8-16.

Tomkins, J. & Twomey, J. (2000). Occupational mobility in England. Applied Economics, 32, 193-209.

U.S. BLS(2000). Current population survey technical paper 63--design and methodology. Zawacki, R. (1992) Motivating the IS people of the future. Information Systems Management, 9(2) 73-75.

Zuboff, S. (1992) Automate/informate: the two faces of intelligent technology. Organizational Dynamics, 20, 5-18.

James Morgan, Northern Arizona University Wendy Campione, Northern Arizona University Max Jerrell, Northern Arizona University

Table 1: Distribution of Demographic Variables

                            IT Related
                            Occupations

                         Count   Percentage

EDUCATION LEVEL

  High School or <        401      6.90%
  Some College            818      14.07%
  Associates Degree       656      11.28%
  Bachelors Degree       2873      49.42%
  Masters or Higher      1066      18.34%

INDUSTRY

  Manufacturing          1142      19.64%
  Trans., Util. & Fin.   1109      19.07%
  W & Retail Trade        360      6.19%
  Computer Services      1941      33.38%
  Other Serv & Govt.     1262      21.71%

CITY SIZE

  Small (<500,000)       1256      21.60%
  Medium                 2294      39.46%
  (500,000-5,000,000)
  Large (>500,000)       2264      38.94%

FEMALE AND MINORITY STATUS

  Female                 1692      29.10%
  Black                   369      6.35%
  Hispanic                182      3.13%

TOTAL COUNT              5814

                         Other Technical
                         Occupations

                         Count   Percentage

EDUCATION LEVEL

  High School or <       1701      14.33%
  Some College           1945      16.39%
  Associates Degree      1665      14.03%
  Bachelors Degree       4612      38.86%
  Masters or Higher      1946      16.40%

INDUSTRY

  Manufacturing          5260      44.32%
  Trans., Util. & Fin.   1810      15.25%
  W & Retail Trade        392      3.30%
  Computer Services       363      3.06%
  Other Serv & Govt.     4044      34.07%

CITY SIZE

  Small (<500,000)       3860      32.52%
  Medium                 4508      37.98%
  (500,000-5,000,000)
  Large (>500,000)       3501      29.50%

FEMALE AND MINORITY STATUS

  Female                 2670      22.50%
  Black                   622      5.24%
  Hispanic                538      4.53%

TOTAL COUNT              11869

                         Management Related
                         Occupations

                         Count   Percentage

EDUCATION LEVEL

  High School or <       2200      17.75%
  Some College           2356      19.00%
  Associates Degree      1288      10.39%
  Bachelors Degree       5116      41.27%
  Masters or Higher      1437      11.59%

INDUSTRY

  Manufacturing          2108      17.00%
  Trans., Util. & Fin.   3438      27.73%
  W & Retail Trade       1303      10.51%
  Computer Services       186      1.50%
  Other Serv & Govt.     5362      43.25%

CITY SIZE

  Small (<500,000)       3939      31.77%
  Medium                 4598      37.09%
  (500,000-5,000,000)
  Large (>500,000)       3860      31.14%

FEMALE AND MINORITY STATUS

  Female                 7251      58.49%
  Black                  1037      8.36%
  Hispanic                537      4.33%

TOTAL COUNT              12397

Table 2: Age Distribution of IS and Other Professional Workers

             IT Related    Other Technical   Management Related
Age Range    Occupations     Occupations        Occupations

25 - 29        16.37%          12.55%              13.65%
30 - 34        19.49%          14.65%              14.90%
35 - 39        19.80%          17.75%              15.86%
40 - 44        16.05%          18.24%              16.83%
45 - 49        13.50%          14.79%              14.99%
50 - 54         9.24%          11.64%              12.79%
55 - 59         4.16%           7.03%               7.43%
60 or Over      1.39%           3.34%               3.56%

Table 3: Average Weekly Wage of Workers by Age and Type of Profession

             IT Related    Other Technical   Management Related
Age Range    Occupations   Occupations       Occupations

25 - 29            $926              $783                 $765
30 - 34          $1,059              $909                 $873
35 - 39          $1,189            $1,001                 $985
40 - 44          $1,177            $1,063               $1,002
45 - 49          $1,181            $1,059               $1,006
50 - 54          $1,158            $1,080               $1,003
55 - 59          $1,201            $1,166               $1,047
60 or Over       $1,090            $1,054                 $973

Table 4: Three Month Job Change Rate IS VS OTHER OCCUPATIONS

                          IS Related    Other Technical   Management
Value                     Professions     Professions     Professions

No Job Change               5,427           11254           11716
Job Change                    387            $615            $681
Percent Changing Jobs        6.66%           5.18%           5.49%
Test for =                                  -3.98 **         3.11 **
  Proportions (Z-Value)

** Signifacant at .05 level.

Table 5: Three Month Job Change Rate Computer Service Industry
VS All Other Industries (In IT and Other Occupations)

                         Information Technology Occupations

                             Computer
Value                    Services Industry   All Other Industries
No Job Change                  1762                  3665
Job Change                      179                   208
Percent Changing Jobs          9.22%                 5.37%
Test for = Proportions                              -5.56 **
(Z-Value)

                         Other Technical and Management Occupations

                             Computer
Value                    Services Industry   All Other Industries
No Job Change                   493                 22477
Job Change                       56                  1240
Percent Changing Jobs         10.20%                 5.23%
Test for = Proportions                              -5.12 **
(Z-Value)

** Significant at .05 level.

Table 6: Determinants of Job change Rates
(Logistic Regression Results)

                               Chi-Square       DF       Pr > ChiSq

  Likelihood Ratio Test         163.0729    17.0000      <.0001
  Parameter                     Estimate    Std. Error
  Intercept                      0.6320     0.4549
  Age                            0.0787     0.0215 **
  Age Squared                   -0.0008     0.00026 **

Education Level (Bach. Degree Base)

  High School or <               0.0152     0.0416
  Some College                   0.0792     0.0364 **
  Associate Degree               0.0795     0.0413 *
  Masters Degree or >            0.0639     0.0376 *

Industry (Services Base)

  Manufacturing                 -0.1275     0.0351 **
  Trans. Utilities & Finance    -0.0066     0.0348
  Wholesale and Retail Trade     0.0437     0.0513
  Computer Sevices Industry      0.2416     0.0469 **

City Size (Medium Size Base)

  Small (< 500,000)             -0.1029     0.0328 **
  Large City (> 5,000,000)       0.0486     0.0293 *

Race and Sex (White and Male Base)

  Female                         0.0160     0.0281
  Black                          0.0063     0.0493
  Hispanic                       0.0116     0.0599

Occupation (Non IT Technical Base)

  Information Technology        -0.0156     0.0385
  Management                    -0.0059     0.0315

** Significant at .05 level

* Significant at .10 level

Table 7: Determinants of Log Wage Rate Across Occupations

                    Info. Systems
                    Occupations

Observations            5813

Adj. R-Square         0.1833

Variable       Parameter      Std.
               Estimate      Error

Intercept       5.43964    0.10453 **
Age             0.06788    0.00522 **
Age Squared    -0.00073    0.00006 **
Month           0.00397    0.00054 **

Education Level (Bach. Degree Base)

  H S or <     -0.27048    0.02237 **
  Some Col     -0.19804    0.01660 **
  Asso Deg     -0.16726    0.01820 **
  MSDeg >       0.07959    0.01515 **

Industry (Services Base)

  Mfg          -0.00159    0.01451
  T. U. & F    -0.01323    0.01463
  W/R Trade    -0.00879    0.02328

City Size (Medium Base)

  Small        -0.13145    0.01478 **
  Large         0.08849    0.01246 **

Race and Sex (White and Male Base)

  Female       -0.15896    0.01214 **
  Black        -0.17494    0.02270 **
  Hispanic     -0.07429    0.03167 **
  Job Change    0.05381    0.02206 **

                      Other Technical
                      Occupations

Observations            11783

Adj. R-Square          0.2799

Variable       Parameter      Std.
               Estimate      Error

Intercept       5.61670    0.0625 **
Age             0.05227    0.00332 **
Age Squared    -0.00051    0.00004 **
Month           0.00333    0.00039 **

Education Level (Bach. Degree Base)

  H S or <     -0.37738    0.01215 **
  Some Col     -0.25708    0.01147 **
  Asso Deg     -0.26005    0.01211 **
  MSDeg >       0.11004    0.01142 **

Industry (Services Base)

  Mfg           0.07786    0.00875 **
  T. U. & F     0.10698    0.01178 **
  W/R Trade    -0.02975    0.02212

City Size (Medium Base)

  Small        -0.08511    0.00927 **
  Large         0.08529    0.00951 **

Race and Sex (White and Male Base)

  Female       -0.24555    0.00960 **
  Black        -0.16250    0.01748 **
  Hispanic     -0.07187    0.01867 **
  Job Change   -0.01617    0.01741

               Management Occupations

Observations            12386

Adj. R-Square          0.2822

Variable       Parameter      Std.
               Estimate      Error

Intercept       5.62659    0.06829 **
Age             0.05131    0.00328 **
Age Squared    -0.00053    0.00004 **
Month           0.00395    0.00038 **

Education Level (Bach. Degree Base)

  H S or <     -0.30675    0.01134 **
  Some Col     -0.24375    0.01097 **
  Asso Deg     -0.25583    0.01351 **
  MSDeg >       0.15588    0.01286 **

Industry (Services Base)

  Mfg           0.05393    0.01100 **
  T. U. & F     0.04671    0.00934 **
  W/R Trade    -0.06161    0.01328 **

City Size (Medium Base)

  Small        -0.12754    0.00935 **
  Large         0.11648    0.00941 **

Race and Sex (White and Male Base)

  Female       -0.23889    0.00811 **
  Black        -0.13547    0.01413 **
  Hispanic     -0.07267    0.01844 **
  Job Change   -0.02448    0.01688

Table 8: Determinants of Log Wage Rate of Workers in it Occupations
(Computer Services Industry Vs All Other Industries)

                            All Other Industries

Observations                         3872
Adjusted R-Square                  0.2009

                        Parameter Est.   Std. Error

Intercept                  5.42152       0.12238 **
Age                        0.06399       0.00607 **
Age Squared               -0.00068       0.00007 **
Job Change                 0.03688       0.02819
Month                      0.00336       0.00062 **

Education Level (Bach. Degree Base)

  High School or <        -0.23423       0.02446 **
  Some College            -0.19243       0.01887 **
  Associate Degree        -0.14452       0.02055 **
  MS Degree or >           0.10454       0.01808 **

Industry (Services Base)

  Manufacturing            0.08097       0.01633 **
  Trans. Util. & Fin.      0.07108       0.01642 **
  W & R Trade              0.07563       0.02377 **

City Size (Medium Size Base)

  Small (< 500,000)       -0.10100       0.01649 **
  Large (>5,000,000)       0.10261       0.01470 **

Race and Sex (White and Male Base)

  Female                  -0.15457       0.01379 **
  Black                   -0.16580       0.02616 **
  Hispanic                -0.05514       0.03596

                        Computer Services Industry

Observations                         1940
Adjusted R-Square                   0.155

                        Parameter Est.   Std. Error

Intercept                  5.29579       0.19433 **
Age                        0.07688       0.00976 **
Age Squared               -0.00082       0.00012 **
Job Change                 0.05861       0.03563
Month                      0.00475       0.00104 **

Education Level (Bach. Degree Base)

  High School or <        -0.34835       0.04863 **
  Some College            -0.18578       0.03257 **
  Associate Degree        -0.20132       0.03613 **
  MS Degree or >           0.04312       0.02689

Industry (Services Base)

  Manufacturing
  Trans. Util. & Fin.
  W & R Trade

City Size (Medium Size Base)

  Small (< 500,000)       -0.15897       0.03092 **
  Large (>5,000,000)       0.05502       0.02270 **

Race and Sex (White and Male Base)

  Female                  -0.13975       0.02411 **
  Black                   -0.16004       0.04320 **
  Hispanic                -0.12964       0.06222 **

** Significant at .05 level * Significant at .10 level.