Introduction

The term “labor turnover” has explicit and implicit meanings for those who practice and those who use labor market information.  In order to produce common definitions and common understandings about turnover, four State Labor Market Information (LMI) Offices[1] have joined together to employ a standardized approach to the computation of job taking (and therefore of employer hiring), worker exits from employment, and continuous employment.  The strategy and categories of computation are fully documented in the accompanying paper by Tony Glover.[2]

The overall strategy in this phase of turnover development is twofold.  First, no two Unemployment Insurance (UI) tax entities collect and process wage records and other tax information in precisely the same way.  Producing common interstate measures means management of UI files for research purposes in a consistent manner in each LMI Office.  Identifying comparability issues and producing standard strategies to cope with them is an ongoing task. Our second objective was to lay the foundation for subsequent analysis of employer-worker interactions viewed from the perspective of job seeking and retention costs.  The concept of a re-hire in the turnover data, for example, is related to the concept of job attachment found in State UI programs.  A re-hire strategy constitutes a relatively low cost interaction between employer and worker.  Similarly, high turnover among primary education workers may signal the use of substitute teachers-- another form of low cost job finding for teachers and worker finding for schools.  The interpretation of worker-employer interaction based on wage records and UI accounts is not always straightforward.  While we cannot infer motive from market churning behavior, the analysis of wage records and employer activity in quantitative form is a useful beginning to understanding how the labor market works and why. 

The analysis of secondary data sets by themselves can produce misleading results.  At the same time, the computation of turnover data can be viewed as a gateway activity.   Turnover rates serve as the dependent variable in the analysis of the effects of employer provided training, pay, and benefit effects on worker retention.   Worker retention strategies are among the issues of greatest interest in the workforce development community.   The findings of this paper suggest that turnover can serve as a necessary intermediary for other applications as well.

Findings

Turnover and job retention appear in patterns consistent with what we know about labor force participation and unemployment rates.  Demographic segments with higher job exit rates also have higher unemployment rates.  Middle aged individuals exhibit the highest market participation rates and highest rates of continuous employment when measured by wage records linked to UI accounts.  In terms of face validity, the traditional Current Population Survey (CPS) based labor market concepts have corollaries in turnover data sets.  The consistency between CPS estimates and State administrative data based turnover represents an opportunity for State LMI Offices to develop richer profiles of  labor force estimates.

The available data on exit and unemployment rates indicate that there is a correlation between the two indicators of labor availability.  Many descriptions of the cost of acquiring labor assume a fixed cost rate of recruitment and training relative to the salary paid for the vacancy.  Exit data, however, can be interpreted to suggest that hiring costs are a variable function of market conditions.  These market conditions may also be associated with the unemployment rate.  It appears that cost of acquiring labor is a variable cost and that these costs may be inferred from either or both the unemployment and exit rates.

Some of the findings from the turnover project suggest new ways in which turnover and retention may be used in conjunction with the CPS.  On the other hand, analysis of inter-state labor movement raises a concern over the accuracy of state and local labor force estimates.  The analysis also serves as a note of caution in the use and interpretation of OES staffing pattern estimates.  Turnover analysis can be used to elaborate on the use of traditional measures of market activity and to fill gaps in knowledge.

The level of continuous employment (workers employed by the same firm for three continuous quarters or more) can be treated as an indication of the probability that workforce and incumbent worker trainees are exposed to greater opportunities for successful labor market outcomes.  However, the rates of continuous employment in the mountain west states of Wyoming and New Mexico are consistently lower than in the plains states of Nebraska and South Dakota. The industrial composition of each state, in turn, is associated with the opportunity structure for successful, continuous attachment to the labor market.  Turnover analysis can produce objective, common, and standardized ways of accounting for the structure of economic opportunity at the state and local level. They could also be used to objectively and systematically describe the “economic and demographic” characteristics of the populations to be served as provided for in negotiating performance levels under the Workforce Investment Act’s Section 136(c)(3).

Organizing the Issues and the Analysis

Figure 1 identifies the market participants, individuals and firms, the levels of organization in which each may be found, and identifies some of the market and workforce development strategies mediating between them.  Administrative data about individuals, for example, can focus on descriptions of individuals at one point in time.  Or, individuals can be organized into demographic segments, as is shown in Table 1, and their market interactions (exit rates) analyzed over time.  The data in Table 2 are presented to demonstrate the problem of stock and flow analysis over both time and space.  In Table 3, we present information on the market-based use of human resources by industrial grouping.  Each market participant’s actions are goal directed (i.e. organizing a living).  However, administrative data by themselves are insufficient to infer certain attributes of market behavior which suggest that administrative records used in conjunction with other types of research are called for to explicitly address these issues.

Lastly, the role of the workforce development system (Column 2 in Figure 1) is addressed elsewhere in this Symposium.[3]

The successful management of educational, and employment and training programs, and, to some extent, understanding the consequences of personal choice in the labor market, depends upon understanding how the labor market works.  At the same time, there is a paucity of published research on the stock and flow of labor over the business cycle.  Measuring employer-worker interactions over time, or analysis of the ecological distribution of workers and employers and the cultural, social, and economic mediums through which they seek to achieve their mutual goals is a daunting task.  In part, the lack of research in this area may be a function of the fact that those who have greatest access to the data sets needed to most efficiently conduct this research lack the resources.  On the other hand, there may also be a certain reluctance to

engage in basic research without either a certain tangible product as the outcome, or a certainty over how turnover research may be connected to traditional labor market measures.

Typically, wage records analysis is funded at the client group level (see column 1 in Figure 1) for performance measurement.  Turnover analysis, on the other hand, (at least as practiced here) is conducted at the population and total market level with a focus on industry and market mechanisms.  Turnover analysis, however, is not generally funded.  Consequently, the context for the analysis of performance for any population segment, and the concomitant role of the workforce development system is not generally available for most analysis by either researcher or Workforce Development Council member.

One of the objectives of the four state turnover project is the development of the context for the subsequent analysis of segments of individuals as they work (or are worked) through the workforce development system.  Population parameters (e.g., the turnover rate) become interpretable to the extent that we can understand the extent to which these parameters are similar to traditional, commonly understood (or taken for granted) labor market information concepts, or to the extent that we can understand why they may be different from traditional measures.  The question, then, becomes one of asking how turnover concepts might be compared to traditional labor market information measures.

Turnover and Labor Force Analysis

Does turnover measure the same market phenomenon as traditional measures, or, does it measure a different dimension(s) of a similar concept(s)?  In the first quarter of 2000, two-thirds (66.2%, n=166,115) of those who worked during the quarter in Wyoming were continuously employed.  That is, they worked for the same employer for three quarters or more.  While those continuously employed may have engaged in job changing, the bulk of all hiring and exit transactions in the market were associated with the remaining 84,814 persons who worked at any time during the quarter.[1]

As can be seen in Table 1, continuous employment rates climb and exit rates decrease as age increases until age 65 is reached.  This relationship appears for both genders in each quarter.  High continuous employment and CPS based participation and employment rates have common peaks for the 25 to 54 year old segment, and lower rates for age segments that are younger and those who are more mature.  Similarly, high exit and unemployment rates are common to younger workers regardless of gender.

Annual average 2000 Wyoming labor force estimates by demographic segment show a participation rate for 35-44 year olds at 89.2 percent and for 45-54 year olds at 89.2

percent, with the distribution of employment also the highest for these two segments.[1]  Those aged 19 and under have the highest exit rates (49.9% in 2000 Q3) and have the highest unemployment rates at 10.7 percent.  Both exit and unemployment rates taper off with age until age 65, when both increase slightly.  While exit rates lack the volitional character of the official definition of unemployment, an exit is a necessary condition of unemployment for everyone but new and re-entrants to the labor market.  It appears at this point that turnover and unemployment are conceptually related phenomenon, and that they are operationally interrelated but not precisely the same measure.

Figure 2 illustrates the relationship between turnover-based exit rates, and quarterly unemployment rates for 1996 through the available quarters of 2001 in Wyoming. (Wage records 2000 Q3 through 2001 Q3 are subject to revision with the 2001 Q4 extract. More substantial changes are likely in more recent quarters.)   The correlation between the two series may become more convincing when we remove the job attached re-hires from the exits in Figure 1.  However, given these results, we would certainly look to other states in the project to replicate and confirm these findings.  The potential to estimate unemployment rates from exit rates means that it would be possible to allocate quarterly CPS based unemployment rates (as a control) to industries for all states. 

Based on the fit between the unemployment and exit rates shown in Figure 1, it may be useful to explore the potential of computing unemployment rates from exit rates for demographic segments on a quarterly basis for those states with access to demographic information.  In theory, by controlling the total exit rate to the CPS unemployment rate, all states could compute industry unemployment rates.  In addition, some states could compute industry and demographic (and/or industry by demographic segment unemployment rates) unemployment rates at the sub-state level as well.[2]

Exit to; Job Changing and Migration as Turnover

Considering labor availability and retention as a “local” issue assumes that markets are closed systems.  This section, however, suggests that local market behavior can only be understood in terms of the broader dynamics of which they are a part.  Turnover analysis represents one of those dimensions connecting locality to the larger context and in general we include it as the type of analysis identified under “Interactions” found in Figure 1.

Some of the turnover found in the quarterly tabular data from the four states represents job changing behavior and the migration of labor between these and other states.  Turnover also represents the movement of specialty workers from location to location within the same corporate environment but between different State’s UI accounts.  Other quarterly hire and exit activity represents the movement of labor from secondary markets to higher earnings primary markets.  Exits represent a change to another status, among them: a withdrawal from the market (and return to school), unemployment, a new job, a new location, and for those with more than one job it may mean a loss of partial earnings.  Replicating in narrative analysis all of the behavior of stock and flow data means mastering the complexity not simply of the computations but of the reality they represent.  Clearly, this discussion is limited to only a few outcomes associated with exit behaviors.

The rapid increase in natural gas prices in 1999-2000 resulted in a substantial expansion of the natural gas industrial complex in Wyoming.  Oil and gas, trucking, wholesale trade, and the energy related construction employment grew rapidly during this period and throughout 2001.  Much of this development was concentrated in the Powder River Basin of northeastern Wyoming and focused on the availability of methane which could be extracted from shallow wells drilled to the surface of underground coal beds.

Two types of events recorded in the turnover data were associated with employment growth in the oil and gas industry.  (The oil and gas industry grew by 17.8 % and made up 4.1 percent of all jobs in 2000, see Table 3.) Approximately two of five growth jobs came from workers within Wyoming changing jobs, and the rest were the result of workers moving to the area from other states.  Table 2 illustrates the primary industrial sector and primary place of work in 1999 for individuals moving to the oil and gas services industry in Wyoming during 2000.  A total of 5,149 persons entered employment in SIC 13 in Wyoming in 2000.  Of these, 43.9 percent (n=2,261) were employed in another industry in Wyoming in 1999 and most, 55.9 percent, worked in the services sector.  Wyoming’s service sector employees who went to work in SIC 13 during 2000 experienced the greatest relative wage gain (36.5 %).  Wyoming workers moving from goods-producing industries (UI covered agriculture, construction, manufacturing and non-SIC 13 mining) to oil and gas saw average quarterly earnings rise from $5,027 in 1999, to $6,638 (32.1%) in 2000.  The economic advantage of job changing seems clear (see Figure 2) with most workers finding work in all four quarters (on average 3.4 quarters in 2000), up slightly from 1999. 

A significant share (2 in 5) of individuals who came to work in the oil and gas industry in 2000 from another state had at least some prior work history in Wyoming during the period 1992 to 1998.  Once in Wyoming, individuals from 7 States with data sharing agreements with Wyoming’s LMI Office worked an average of 3.4 quarters, the same average as those moving from a job in Wyoming to the oil and gas industry.  However, the 831 workers from these 7 States had earnings in 1999 that were 33 percent above ($6,317 compared to $4,748) those of Wyoming workers, and 24 percent above ($7,918 compared to $6,381) Wyoming in 2000, after both worker groups entered the oil and gas industry.  These earnings differences may reflect a higher skill and/or experience level among workers moving to Wyoming than the skills of resident workers.  Both the source data and anecdote suggests that some companies move workers with specialized skills

from work- site to work-site as circumstances demand.  These individuals may be paid by the same firm in one quarter from payroll offices (and UI accounts) in more than one state.  Hires and exits, in this circumstance, represent the movement of human resources from locality to locality as a function of the technical requirements of the job site rather than as a function of hiring a new individual.  Employees showing up in the data as a “hire” may be working for the same multi-state or multi-national company and simply represent a transfer rather than a hire.

Most notable among the 2,057 individuals relocating from work in other states with whom we do not have a data sharing agreement, is the fact that they worked on average only 1.8 quarters in the oil and gas industry in Wyoming during 2000.   Many of these workers entered the market in Wyoming and disappeared from it shortly thereafter.  Only 38.1 percent of those who worked in SIC 13 in Wyoming in 2000 were still working in the state and oil and gas during the first quarter of 2001.  This proportion declined to 30 percent in the second quarter.  In contrast, three in five workers employed in jobs in Wyoming in 1999 were still working in SIC 13 in the first quarter of 2001, and slightly over half were still employed in the second and third quarters (wage records for 2001 represent preliminary data).

Entering employment in SIC 13 for Wyoming residents (those that worked in the State in 1999) appears to be a response to earnings incentives.  Vacating the lower wage secondary labor market and entering the higher wage primary market has inspired a number of news accounts of labor shortages in the state.  This concern should not be surprising considering that this movement of labor probably has its parallel across the supporting and associated industrial infrastructure and complex. 

However, since aggregate employment in SIC 13 continued its rapid expansion in 2001, it is not particularly clear why residents, rather than non-residents were more likely to remain employed in the industry.  Moreover, we expect some inter-state movement of workers along corporate avenues.  However, the tentative attachment of half of the workers who entered the oil and gas industry in 2000, primarily those from another state, lacks explanation in the face of expanding aggregate demand for labor in oil and gas.  On the other hand, work in this industry is often physically demanding.  Given this situation, non-residents may be less willing to remain in the industry while residents may still view it as an advantage over work in the secondary market, especially since such fixed costs as housing still remain.  At the same time, the non-resident exodus from oil and gas is difficult to explain as a function of labor demand elsewhere especially since the first quarter of 2001 represented the period when most States were entering a recession.

The rapid expansion of the oil and gas industry and associated industrial complex presents several problems for the “official” statistical system.  The fact that over half of the workers responsible for employment growth in 2000 re-located from work in another state creates anomalies in statistical series based on snap shots of the market rather than stock and flow analysis grounded in administrative records.  Immigration based employment growth in 2000 and 2001 was associated with employment in the establishment survey expanding at a rate three times faster than growth in the labor force estimates.   This imbalance in two sets of “official” employment statistics is clearly a function of the migration of workers to Wyoming that represent households established outside the scope of the Current Population Survey sample frame for the State.  Since turnover analysis can be used to reconcile dissonance between the establishment and labor force series, it would certainly appear worthwhile to see the practice followed in more than one state.

Some of those new to the oil and gas industry in Wyoming represented specialty workers arriving and departing as part of a multi-state workforce.  At the same time, a number of  work positions characterized by specialized skills were included in the OES collection.  Turnover may represent the movement of workers into and out of a relatively fixed industry staffing pattern.  On the other hand, it may represent occupational differentiation or specialization as the skill mix changes to meet transitory market task requirements.  Occupational staffing patterns drawn from employment peaks may not be representative of longer term trends.  Those engaged in occupational projections tend to either fail to specify their assumptions about staffing patterns, or assume a fixed staffing pattern and geographically stable employment.  These assumptions may not be reasonable depending upon the way firms interact with the market and make seasonal and business cycle staffing decisions.  Staffing patterns estimated from business cycle troughs and seasonal peaks may not be indicative of staffing patterns found during an expansion.

Finally, place, not simply the ecology of worker-firm relationships, is an important determinant of the measurement of turnover and the allocation of labor among industries.  If we stipulate that demand is the principal explanatory variable for turnover, then it seems that we need to specify how this mechanism works and the conditions affecting its efficiency.  Otherwise, we merely affirm the consequent.  The concept of community implies access to knowledge, ownership of property, family, and other institutional ties associated with a sedentary existence. Were demand the key variable explaining the availability of labor, then all local unemployment rates would constantly be shifting to the national unemployment rate as labor migrated to accommodate the current circumstance.   The fact that this does not occur universally suggests that other factors need to be taken into consideration if we are to understand how the market works.

Worker Retention Strategies; Continuous Employment in Four States

Employer retention strategies, pay, benefits, and training are found in the fourth column of Figure 1 and are identified as employer and industry based strategies for managing labor retention.  In this section, we explore part of the market context using some of the industry data generated from the four state turnover project.

Table 3 contains quarterly continuous employment rates by industry for four states, along with the associated growth rates for 1999 to 2000 and the employment distribution for each State based on the ES 202 files.  At 3.1 percent, New Mexico exhibited the fastest growth from 1999 to 2000 while Nebraska, at 2.0 percent, had the slowest growth with Wyoming and South Dakota falling in between.

The continuous employment rate represents the proportion of all persons who worked during a quarter who worked for the same employer (he or she may have worked for other employers as well) for three continuous quarters.  Employers and industries reliant on seasonal employment often have the lowest continuous employment rates during peak periods of employment.  As was evident in Table 2, lower continuous employment rates tend to be associated with higher exit rates.  The mountain west states of Wyoming and New Mexico had consistently lower continuous employment rates across all four quarters of 2000 than did the adjoining plains states of South Dakota and Nebraska.

Nebraska and Wyoming have the highest and lowest continuous employment rates regardless of quarter (with one exception in Q1).  In the first quarter of 2000, Wyoming’s continuous employment rate was 66.2 percent while in Nebraska it was 72.4 percent or a 6.2 percent difference.  The point separation between them was 11.2 percent in the second quarter, 10.3 percent in the third quarter and 9.4 percent in the fourth quarter.  These differences appear to be associated with the distribution of employment among industries with large seasonal employment components.

Almost one-third (32.8 %) of the employment in Wyoming is found in industries with a strong seasonal component:  construction (8.6%), retail trade (20.4%), and hotels and other lodging places (3.8%).  On the other hand, these three sectors made up only a quarter (24.5 %) of employment in Nebraska, with New Mexico close to Wyoming at 29.1 percent and South Dakota at 27 percent.  In rank order, then, continuous employment rates are associated with the level of employment in these seasonal industries.   This cursory analysis should be considered as suggestive of the direction future analysis should take.  Certainly, industries associated with construction (i.e. sand and gravel mining) could be expected to be more seasonal in New Mexico and Wyoming than in either plains states.  Further, longitudinal analysis needs to be developed to identify the consistency of these relationships.  At the same time, comparative inter-state data clearly allows us to determine whether or not stock and flow events are internally consistent and seemingly explainable. 

Continuous employment rates, earnings, and migration data should allow us to identify which industries may serve as the best candidates for workforce and economic development.  Occupations in growth industries with high rates of continuous employment would appear to offer better choices for targeted job development purposes if the program objective is increased earnings and stable employment.  At the same time, it appears that program efforts aimed at these goals could be accomplished with less difficulty in Nebraska than in Wyoming.

Longitudinal analysis of continuous employment and exit rates need to be examined over the course of the business cycle in order to identify constant relationships.  Even so, the comparative analysis of subsets of employment across states can prove valuable.  As can be seen in Table 3, nursing care facilities have the lowest continuous employment rates when compared with either hospitals or the offices and clinics of medical doctors, dentists and other practitioners.  In comparison to these two industry classifications, Nebraska has the highest relative share of employment in nursing care facilities and was the only state among the four to demonstrate net employment growth in this industry from 1999 to 2000.  The issue of turnover at the occupational level, for nurses, is discussed elsewhere in this Symposium.[1] However, given a knowledge of the industrial makeup and the consistently lower continuous rates for nursing care

facilities, it seems reasonable to suggest that the retention of nurses is more problematic in Nebraska than in the other three states.  This particular illustration of the use of comparative data suggests another application of turnover data facilitating the generalization of research findings from one state or setting to another.

Turnover and retention data can be used to identify the applicability of evaluations completed in one state for other states.  A barrier to generalizing successful employment and training programs is the lack of information on the way these same innovations may interact with the markets of target states.  While only a few states may be able to monitor the progress of program completers through market processes associated with continuous employment and low turnover, all states can compute turnover in a standard way.  The availability of standard turnover computations in each state would allow LMI Offices to compute the probability of comparable success in target states from proven innovations in other states.

Turnover data can also be used as the dependent variable in explaining the role of training, benefits, and pay in employer strategies to retain labor.  Benefit survey results by themselves can provide employers and workers with information on the incidence, availability, and sometimes the cost of benefits packages.  They cannot, however, answer the question of which benefits are required to retain an existing workforce.  This question can only be answered if benefits packages and costs can be related to continuous employment rates and/or turnover rates.  From a practical standpoint, turnover and continuous rates need to be computed at the firm level by the same LMI offices collecting benefits information.  These are the necessary pre-requisites to answering the question of the role of benefits in the maintenance of high retention rates.

Conclusions

Turnover and retention data are essential to answering the questions Workforce Development Councils have asked regarding how the labor market works and how the workforce development system interacts with it.  State ability to replicate turnover and retention computations for subsets of market participants receiving services under WIA, other One Stop services and related programs, allow Councils to view the market paths taken by program participants to the achievement of program outcomes.  State capacity to aggregate these subsets into a single profile across the employment, training, and educational system as a whole permits the role of the workforce development system to be identified within the context of market functioning.  This capacity is essential if we are to distinguish and visualize uncontrolled market effects from manageable program effects on outcome measures.

Having common turnover computations for a group of states permits the computational strategy to be held constant and facilitates the identification of meaningful market and labor supply features.  Decomposing transactions into a standard typology of hire, exit, and continuous employment allows us to identify underlying dynamics obscured by traditional measures of the labor market.  The fact that continuous employment rates often vary within the same industry in different states is suggestive of the role of locality and place in market dynamics.  On the other hand, there are instances where the distribution of employment, growth, and continuous rates of employment are consistent (e.g. food stores) regardless of location.  The underlying purposes of this paper have been to highlight selected applications of turnover analysis that can potentially be used to answer the questions of Workforce Development Councils, and to begin identifying areas of research that the four states may wish to explore next.  The range of possibilities in both arenas is large indeed and any state is welcome to join us.

Wage Records Research Index  |  Labor Market Information