Labor market movements and the indicators developed to measure them are among the most widely watched economic barometers. Therefore, a comprehensive report that can not only capture labor market dynamics at the industry, macro and regional levels but also provide a benchmark to measure human capital management for individual firms is of great importance.

ADP in collaboration with Moody’s Analytics has developed a comprehensive ADP Workforce Vitality Report using ADP’s large anonymous and aggregated payroll data set to provide insights on labor market dynamics using various dimensions and also to provide a benchmark to measure human capital management for individual firms. The report is derived from a sample of approximately 250,000 companies and 18 million employees each month, which accounts for about 15 percent of all U.S. private sector employees.

This report includes a range of metrics relating to employment, hours, and earnings. It links both the employees and firms through individual worker’s utility maximization and firm’s profit maximization.

Many existing labor market-metrics are constructed at the national level. The few metrics those do provide more detailed measures are either the overall performance for major metro areas, or selected aspects of a local economy. No existing metric assesses the labor market performance at such a diversified and detailed level as does this report. Because the industrial, geographic and demographic characteristics of a specific labor market can look quite different from the national trend, understanding and evaluating the labor market dynamics of a finely defined category will benefit job seekers and employers who operate in a specific labor market. Specific metrics will enable human resource managers to appropriately adjust their policies in response to the changing market conditions.

The WVR distinguishes four types of workers in the labor market: those who stay with the same firm (job holders), those who change jobs (job switchers), those who are newly hired by a firm (entrants), and those who left the firm either voluntarily or involuntarily (leavers). Such a distinction is meant to better capture labor market dynamics, and allows for a complete depiction of every worker available in the dataset.

The Workforce Vitality Report provides indicators calculated as both month-over-month changes as well as year-over-year changes. The report includes many metrics, such as: wages, changes in wages for job holders, changes in wages for job switchers, turnover rate, switching rate, change in hours worked for job holders, change in hours worked for job switchers, total employment growth.

These metrics are constructed by various dimensions: region, industry, firm size, age, gender, income level, tenure, and full/part time status.

Summary values reported for each category are aggregated from the individual payroll records of employees who are grouped in the category. This approach ensures that the levels of total employment and wages are consistent across various dimensions, which includes consistency among the four worker types.


Beginning from the third quarter of 2013, monthly data based on approximately 18 million employees are used. ADP’s data have two unique advantages for constructing a comprehensive Workforce Vitality Report. First, the data enable the same firm and employee to be tracked over time. Thus it is possible to measure job switching from one firm to another by computing the job switch rate and comparing pre and post job switch wages (from workers who switch between two companies in the sample). Second, the ADP data provide employee and firm demographic variables such as a firm’s industry and size, and an employee’s age and gender. These variables are used to construct metrics for many combinations of interest in human capital management (HCM). Such analysis is especially useful for HCM as it not only provides a comprehensive dimension that can fit into any individual firm’s category, but also gives firms a benchmark for comparison.

The ADP WVR model checks the raw data for outliers, anomalies and inconsistencies. The raw dataset is modified in a number of important ways.

  1. Many records with extreme or incomplete values are automatically excluded from the analysis. These include cases where:
    1. The worker is paid by the day
    2. The worker is younger than 15 or older than 85 years’ old
    3. The worker did not have earnings in the current period
    4. The worker has reported hours worked that are negative or greater than the total number of hours in the month
    5. The worker does not reside in one of the 50 U.S. states or the District of Columbia
    6. The worker’s hire date is more than 50 years ago or after the end of the current month
    7. The worker’s hourly wage is below the federal minimum wage of $7.25/hour, or $2.13/hour for certain service industries
  2. In cases where an employee has multiple wage records for a single employer, resolve these duplicates by selecting the largest hourly wage and earliest hire date.
  3. In records where the firm size and industry data fields are missing, populate using the most frequently reported values of the same firm in the current period.
  4. If the standard hours worked per pay period is missing, monthly hours worked and the hourly wage cannot be calculated for salaried workers. In such cases, impute standard hours worked by replacing the missing values with the hours worked of similar salaried workers.
  5. Since the raw data is aggregated for an entire month, the reported hours worked can be lumpy due to some months containing more pay periods than others (i.e. a fulltime worker may be credited with 160 hours worked in some months, and 240 hours worked in others). In order to prevent this volatility from impacting true estimates of hours worked, the raw hours are smoothed at the firm level (separately for hourly and salaried workers) in order to correct the issue.

Theory and Empirical Indicators

Microeconomic theory and modern labor economics guide the selection of the total wage bill as the basis of vitality because it links employees with firms. An individual worker maximizes utility by consuming a bundle of commodities subject to a budget constraint, defined as wage income. By choosing the optimal combination of goods and services, the worker can obtain the highest level of utility, which is proportional to real income. As the labor market is composed of millions of such individual workers, the total wage bill characterizes the maximum utility of all employees.

From the employer’s perspective, in order to maximize profit, firms seek to minimize the cost of labor. However, firms do not want to pay so little as to affect the productivity and morale of current workers or lose talented workers to competitors. As the labor market recovery gains momentum, firms are willing to pay higher wages to retain and recruit workers and employees benefit from rising real income. As such, the total wage bill is a good indicator of the vitality of the labor market.

The ADP WVR model distinguishes four types of workers: those who stay with the same firm (job holders), those who change jobs (job switchers), those who are newly hired by a firm (entrants), and those who left the firm either voluntarily or involuntarily (leavers). Such a distinction is meant to better capture labor market dynamics because the wage changes for job switchers are more responsive to changes in labor market conditions, and the difference between the entrants and leavers captures the dynamism of the labor market. The ADP WVR model distinguishes workers into the various types by matching workers from one month to the next.

In practice, in order to separate workers into the various types, the ADP WVR model tracks two months: the current month t, and the previous month t-1:

  1. Holders: jobs where the workers were hired prior to t-1 or hired in t-1 with positive pay in that month. If workers in these jobs are determined to have worked their job for the entirety of period t, all other ADP employment recorded for the same worker (e.g., second jobs) is also designated as holder.
  2. Switchers: current period jobs not assigned to holders where workers were employed by a different ADP firm in t-1.
  3. Entrants: jobs not assigned to holders where workers were not employed by an ADP firm in period t-1.
  4. Leavers: workers who were with an ADP firm in t-1 but are no longer in the ADP sample in period t

The ADP Workforce Vitality Report includes indicators which are calculated using two general methods: group method and individual method.

Group Method Average Hourly Wage (AHW):

where i indicates a worker in ADP sample, ni is her weight (for example, n=5 means this ADP worker represents 5 workers in US), wi is her hourly wage, and hi is her monthly hours worked.

In order to produce either month-over-month or year-over-year growth rates using the group method, the AHW is also calculated in period t-1 or t-12:

As such, year-over-year AHW growth is simply: AHWt / AHWt-12 – 1.

In addition, a similar formula is used to calculate growth in average weekly hours.

By contrast, the individual method eliminates the possibility of compositional effects by matching specific employees between the current period and previous periods and only using records which appear in both. An additional benefit of this method is that it allows for wage and hours growth calculations to be done separately for job holders and job switchers.

The wage growth formula is similar for switchers and holders, the difference being that wage growth for job holders (JH) is typically examined on a year-over-year basis, while wage growth for job switchers (JS) must be calculated month-over-month at the time the job change occurs. In order to make a valid comparison between the wage growth for job holders and job switchers, we calculate a 12-month moving average of the month-over-month wage gains for job switchers.

where i indicates individuals in the given dimension in time t, ni is her weight, wi is her hourly wage, and hi is her monthly hours worked. Note that the weights (nit) applied to each individual for time t and time t-12 are both the weights in the current period. That is, we fix the individual weights for the current and previous period.

In addition, a similar formula is used to calculate growth in average weekly hours for both job holders and job switchers.

Following indicators also provide vital information about the health of labor market.

  1. Total employment growth: As a widely watched labor market indicator, employment growth is the fundamental measure of labor market vitality. Higher employment means more job opportunities for job seekers and more competition among employers. This too can be disaggregated for different dimensions. While this component is readily available from government sources, growth in combined dimensions is less readily available.
  2. Job switchers’ wage growth: Change in nominal hourly wage rate between the new job and the old job. This is a unique indicator not available from government sources. The initial wage offer is more sensitive to labor market conditions compared with wages of job holders. A higher wage offer is often a result of a tighter niche labor market where employers compete for talent. The ability of job switchers and, in particular, a specific job switcher defined by age, tenure, gender, pay scale, industry and region to command a higher wage is of particular importance for HCM. For example, the wage change for job switchers between minimum wage earners and high income switchers can be very different. While the former may be hard pressed to find better paying new jobs, the latter can boost their wages by moving to a new position.
  3. Job holders’ wage growth: Although the magnitude of wage changes of job holders is often smaller than that of job switchers, an improving labor market will eventually increase the pay of job holders as employers use monetary incentives to retain talent and raise productivity. Both the job switch rate and the change in nominal hourly wage rate for job switchers is likely to lead changes in the hourly wage rate for job holders. However, workers across specific dimensions, such as particular industries or wage levels, may be able to bargain for higher wages sooner in the business cycle than workers in other dimensions.
  4. Job holders’ hours growth. This variable characterizes the intensity of labor utilization. When a specific labor market is sluggish, weak demand causes under-utilization of labor. When the specific labor market picks up, working hours increase as production expands. Hours worked have leading properties as workers in certain industries or wage scales or parts of the country may be working more hours earlier in the business cycle than other workers. In addition, firms that employ high-skilled workers tend to retain their employees by cutting worker hours during downturns so as to avoid the search and rehiring costs when the economy rebounds. As such, fluctuations in hours worked vary across industries with different concentrations of skilled workers.
  5. Switching rate: The percentage of workers who successfully changed their jobs within one month. In contrast to the separation and quits rates reported by the BLS, the job switch rate gives a clear indication of labor market conditions by isolating those workers who successfully changed their jobs and excluding those whose departure was due to retirement or company closure or who dropped out of the labor force. Because the majority of those who land a new job left the previous job voluntarily, the job switching rate increases when labor market conditions improve. This component is of particular interest in distinguishing labor market conditions for different dimensions of workers to ascertain whether, for example, higher or lower wage workers are switching jobs, whether workers of a certain age or tenure on the job are switching or staying put, or whether workers in certain industries or certain regions of the country are switching jobs.
  6. Turnover rate: the percentage of workers who either successfully changed their jobs or left their job within one month. The switching rate (above) is one piece of the turnover rate, which provides a more holistic view of the churn in the labor market. The turnover rate captures all separations from a job held in the previous month, whether the separation was voluntary or involuntary and whether the separation resulted in another job being acquired or not. A higher turnover rate can sometimes imply a strengthening economy, but since it also includes retirements and involuntary separations, sometimes the relationship is less clear.


The metrics described above are constructed for as many dimensions and combinations of dimensions as the data will allow. A list of the definition of dimensions is as follows:

  1. Geography (15 total areas, including U.S.):
    1. Region: Northeast, Midwest, West and South
    2. State: New York, New Jersey, Pennsylvania, Texas, Florida, California, Illinois, Washington, Michigan and Ohio
  2. Industry (11 total industries, including total private) - based on North American Industrial Classification System (NAICS):
    1. Resources and mining (21), construction (23), manufacturing (31, 32, 33), trade, transportation, and utilities (42, 44, 45, 48, 49, 22), information (51), finance and real estate (52, 53), professional services (54, 55, 56), education & healthcare (61, 62), leisure & hospitality (71, 72), other services except public services (81)
  3. Firm size: 1-49, 50-499, 500-999, and 1,000 and above
  4. Age: 24 or younger, 25-34, 35-54, 55 and older
  5. Gender: male and female
  6. Full and part time: Full-time workers are defined as those whose weekly hours is greater than and equal to 35
  7. Wage tier: based on nominal annual wage; less than $20,000, $20,000-$50,000, $50,000-$75,000, and greater than $75,000
  8. Tenure tier: Two years or less, 3-5 years, 5-9 years and 10 years and above

In addition to individual dimensions, the ADP WVR model also produces the metrics for many combined dimensions. The feasibility of various combinations is dictated by data availability and quality. An outline of the dimension coverage is below and the resulting output produces estimates for more than 60,000 individual dimensions in a given month:

Dimensions are crossed in the following combinations (# of dimensions in parentheses):

-Geography (15) / Industry (11) / Size Class (5) / Gender (3) / Age Tier (5) / Wage Tier (5)

-Geography (15) / Industry (11) / Size Class (5) / Tenure Tier (5)

-Geography (15) / Industry (11) / Size Class (5) / Full-time/Part-time (3)


The total wage growth for the entire sample period (2013Q3 – current) is not adjusted to equal the government data. The metrics are calculated at each dimension level directly from individual record-level data. The growth rates of wages or hours at higher levels are not necessarily between the ranges of the growth rates at the lower level.

Sample Weighting

Although ADP data has wide industry coverage, the composition of those employees/firms does not exactly match the size and industrial composition of the total population. For example, ADP data has an over-representation of manufacturing and professional services and an under-representation of healthcare and retail trade compared with published estimates from the Current Employment Statistics (CES) survey. In addition, ADP’s sample contains an over-representation of large companies and under-representation of the smallest companies. Therefore, the ADP sample data is weighted to align more closely with the true universe of workers and companies. The weighting is completed using data from three sources:

  1. Employment levels by geography and industry are adjusted using the output from ADP’s National Employment Report (NER) and Regional Employment Report (RER). These reports are benchmarked annually to the Quarterly Census of Employment and Wages (QCEW) which ensures accuracy.
  2. The distribution of workers at companies of various sizes is adjusted using data from the Census Bureau/Small Business Administration. Data on employment by firm size are available by state/industry combination, and the ADP sample is adjusted accordingly.
  3. The distribution of workers into various employment types (job holder/job switcher/entrant/leaver) is adjusted using data from the Current Population Survey (CPS). The proportion of workers in each group is calculated by industry and age tier, and the ADP sample is adjusted accordingly. Because the ADP sample covers about 15% of the total U.S. workforce and the coverage varies by region the identification of entrants, leavers, and switchers will be biased if no adjustment is made. Specifically, a leaver could in reality be a job switcher who leaves a job at an ADP firm to join a non-ADP firm, so they are no longer included in the sample. Conversely, a worker who newly appears in the ADP sample could be a true entrant to the labor force, or simply a worker who switched from working at a non-ADP firm to an ADP firm.

Seasonal Adjustment

To more closely align with published data from government sources, and given that most wage and employment variables exhibit seasonality, data in the Workforce Vitality Report are seasonally adjusted, where necessary. Much of the seasonality in published wage and employment data is the result of changing workforce composition over time, which often follows a seasonal pattern. For this reason, individual method calculations (such as job holder or job switcher growth rates) do not require seasonal adjustment, since the calculations are based on a constant set of workers. Group method wage calculations for all worker types are seasonally adjusted, as well as switcher and turnover rates.

Seasonal adjustment is done using standard X-12 ARIMA methodology with automatic outlier identification. More detailed documentation on seasonal adjustment methodology is available from the Census Bureau, which develops and maintains the X-12/X-13 ARIMA programs.

Due to the limited data history available in the Workforce Vitality Report, seasonally adjusted series are re-estimated for each release in order to incorporate the newest historical data. As a result, small changes may occur in the historical seasonally adjusted data between publications.