Published on: Mar 3, 2016
Transcripts - NAPEOturnover
NAPEO White Paper Series
By Laurie Bassi and Dan McMurrer
McBassi & Company
Keeping Turnover Low
and Survival High
Professional Employer Organizations:
Keeping Turnover Low and Survival High
In the 2013 report, “Professional Employer Organizations: Fueling Small Business Growth,” a comprehensive analysis of existing eco-
nomic data showed that small businesses in PEO arrangements have higher growth rates than other small businesses, and small busi-
ness executives who use PEOs are better able to focus their attention on the core business. In further exploring the impact of PEOs and
their potential to help small businesses better meet the challenges of today’s demanding economic conditions, this follow-up study ex-
amines employee turnover and business survival
rates for businesses using PEOs and compares
them to national data available from the U.S. Bu-
reau of Labor Statistics (BLS). Applying a variety
of different data specifications, we consistently
found that PEO clients have lower employee
turnover rates and lower rates of business failure
than comparable national averages, after control-
ling for factors such as industry, size, and state of
The employee turnover rate for PEO clients is
10 to 14 percentage points lower per year
than that of comparable companies (see
Table 1), depending on data specification.
The average overall employee turnover rate in the
United States was approximately 42 percent per
year, based on 2012 data. It is 28 to 32 percent for companies that used PEOs for at least four quarters.
Businesses that use PEOs are approximately 50 percent less likely to fail (permanently go “out of business”) from one
year to the next when compared to similar companies in the population as a whole (see Table 2). The overall business fail-
ure rate among private businesses in the United States as a whole3
is approximately 8 percent per year, based on 2012 data. It is ap-
proximately 4 percent per year for those companies that used PEOs for at least four quarters.
Data broken down by specific industries point to “Professional, Scientific, and Technical Services,” “Construction,” and “Finance and In-
surance” as being three industry categories that disproportionately benefit from PEO services in both lower employee turnover rates
and lower business failure rates.
Across all industries, the results reflect clear advantages for PEO clients on two of the most fundamental issues faced by
any business: retention of employees and continued survival. PEOs significantly decrease employee turnover for their clients, al-
lowing them to retain the knowledge and skills of their employees, while simultaneously reducing direct and indirect turnover-related
costs (which are substantial). The fact that PEOs significantly increase the likelihood of client survival is likely a result of PEOs providing
a combination of services that makes it possible for businesses to focus on their core areas of expertise.
Table 1. Average differences in actual and expected employee turnover
rates, PEO clients, 2012.
Expected turnover rate (U.S. overall) 1
Difference for PEO clients, controlling for industry -9.7
Difference for PEO clients, controlling for company size group -13.5
Table 2. Average differences between actual and expected annual business
failure rates, PEO clients, using most conservative data specification.
Annual Business Failure Rate (%)
Expected business failure rate (U.S. overall) 2
Difference for PEO clients, controlling for industry -4.0
Difference for PEO clients, controlling for state -4.1
Employee Turnover Rates
Employee turnover generates a variety of costs to em-
ployers, both direct and indirect. These include all costs
related to hiring replacement employees, onboarding
costs, and opportunity costs incurred during the period
when positions are vacant.
For many positions and many
businesses, however, the (in-
direct) impact of losing the
skills, knowledge, and ex-
pertise of valued employees
may be significantly larger
than any other (direct)
The exact cost of turnover is difficult to estimate,
as it varies so significantly depending on
specifics. A frequently cited estimate based on a
“Cost of Turnover” worksheet4
provided by the
Society for Human Resource Management
(SHRM) is that costs are roughly 150 percent of
the employee’s salary, with other calculations
suggesting it is more than 200 percent for certain
positions, such as managerial and sales jobs.5
the other end of the spectrum, alternative “con-
servative” calculations by O’Connell and Kung es-
timate the average cost of replacing an employee
to be roughly $14,000 each.6
Regardless of which estimate is used, it is clear
that the costs of employee turnover are quite sig-
nificant, and that a business that enjoys a higher
employee retention rate than its competitors is in
a stronger position to survive and thrive over the
Overall annual results for 2012 are reported in Table 3, with results reported by quarter in Table 4. In 2012, PEO clients had annual employee
turnover between 10 and 14 percentage points lower than the national average of 42 percent per year, depending on the comparison
We also analyzed differences by industry. We did not have sufficient numbers to reliably calculate turnover differences for industries with fewer
companies included in the analysis data file, and due to small sample sizes, we view these industry split results as suggestive rather than definitive.
Figure 1 presents differences between expected and total employee turnover9
by industry for the six largest industries10
in the analysis data. Among
these largest industries, we found that the largest salutary effects of PEOs on turnover rates occurred in “Professional, Scientific, and Technical Serv-
ices” (23 percentage points lower) and “Construction” (17 percentage points lower). Smaller differences were observed in “Manufacturing” (2 per-
centage points lower) and “Health Care and Social Assistance” (6 percentage points lower).
Business Survival and Failure Rates
The ranges of aggregated actual versus expected survival values are reported in Table 5. The positive values throughout the table indicate that
PEO clients are more likely to survive, and less likely to fail, than similar companies in the population as a whole (controlling for year
of inception, analysis year, and other factors, as indicated in the table), regardless of which analysis specification is being applied.
The survival data indicate that 8 percent of all businesses fail each year. For PEO clients, the comparable percentage is between 2.1 and 4 percent,
depending on the exact specification. Thus, annual business failure rates among PEO clients range from 4 to 5.9 percentage points lower than the
Table 3. Average differences in actual and expected employee turnover
rates, PEO clients, 2012.
Expected turnover rate (U.S. overall) 7
Difference for PEO clients, controlling for industry -9.7
Difference for PEO clients, controlling for company size group -13.5
Table 4. Average differences in actual and expected employee turnover rates, PEO clients,
2012 Q1 2012 Q2 2012 Q3 2012 Q4
Expected turnover rate (U.S. overall) 8
9.9 10.5 11.3 9.9
Difference for PEO clients, controlling for industry -3.1 -2.5 -3.3 -0.8
Difference for PEO clients, controlling for company size group -4.4 -3.6 -3.4 -2.1
Percentage points by which PEO client’s employee turnover rate is lower than expected*
0.0 5.0 10.0 15.0 20.0 25.0
Prof, Sci, Tech
Figure 1. Differences between actual and expected employee turnover rates, PEO
clients, by industry.
* Larger numbers indicate greater advantage for PEO clients in that industry (i.e., lower employee turnover).
rates for the population
as a whole (50 percent
or more lower). The re-
sults are quite consis-
tent whether companies
are compared to ex-
pected survival for their
respective industries or
states. Even using the
most conservative ana-
lytic approach, the busi-
ness failure rate is 50
percent lower for busi-
nesses using PEOs than for busi-
nesses overall, as highlighted in
It should be noted that the sur-
vival rates calculated for PEO
clients reflect a relatively short-
term effect of using PEO services
(for example, the longest differ-
ence in the analysis file between
PEO use and calculated survival would be slightly
over four years, for a firm using PEOs for four
quarters in 2008, and then having its survival as-
sessed in 2013 Q1).
We also examined industry-specific differences
to identify which industries see the largest and
smallest impacts from PEO services. We did not
have sufficient numbers to reliably calculate sur-
vival variations for industries with fewer compa-
nies included in the analysis data file. Figure 2
presents differences between expected and total
business failure rates13
by industry for the six
largest industries available14
in the analysis data.
Even for these larger industries, due to the
smaller sample sizes, we recommend viewing
these results as suggestive rather than definitive.
Among these largest industries, we found that
the largest effects of PEOs on business failure rates occurred in “Professional, Scientific, and Technical Services” (business failure rate 6.3 percent-
age points lower than expected) and “Finance and Insurance” (5.8 percentage points lower). Smaller differences were observed in “Health Care and
Social Assistance” (0.8 percentage points lower) and “Manufacturing” (1.2 percentage points lower).
Employee Turnover Rates
We selected a stratified random sample of 1,000 companies from the Slavic database for analysis of employee turnover rates. (Because Slavic firms
tend to be larger than average establishments nationally, we chose a sample designed to include a larger percentage of the smaller firms in the
Slavic database). To enable quarterly turnover analysis, all companies in the sample were required to have at least one employee payroll record in
each quarter from 2012 Q1 to 2013 Q1. This allowed us to analyze employee turnover for 2012, the most recent full calendar year available for analy-
sis in the Slavic database.18
Based on this definition, turnover is being calculated only for companies that are current PEO clients.
We examined employee-level records from 2012 Q1 to 2013 Q1 for all employees for each company included in the sample. The periodicity of
payroll records varied across companies; some reported multiple times per month for each employee, while others reported less frequently. We
Percentage points by which PEO client’s business failure rate is lower than expected*
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0
Prof, Sci, Tech
Figure 2. Differences in actual and expected annual business failure rates, PEO
clients, by industry, using most conservative specification.
* Larger numbers indicate greater advantage for PEO clients in that industry (i.e., lower business failure rates).
Table 6. Average differences between actual and expected business failure rates (%), PEO
clients, by year, using most conservative specification.
2010 2011 2012 2013
Expected business failure rate (U.S. overall)12
9.2 7.8 7.7 7.6
Difference for PEO clients, controlling for industry -6.2 -2.7 -4.2 -3.6
Difference for PEO clients, controlling for state -6.4 -2.8 -4.6 -3.3
Table 5. Average differences between actual and expected business failure rates (%), PEO clients,
Data Used in Analysis
Only companies More conservative Most conservative
with fully valid specification (also specification (also
corporate includes invalid includes additional
status data inactives) unknowns)
Expected business failure rate (U.S. overall) 11
8.0 8.0 8.0
Difference for PEO clients, controlling for industry -5.8 -4.3 -4.0
Difference for PEO clients, controlling for state -5.9 -4.4 -4.1
chose to conduct a quarterly analysis
based on the expectation (from guidance
provided by Slavic401k) that companies
would report at least one payroll record
per quarter for each individual employee.
Any quarter in which an employee had a
payroll record in the Slavic database was in-
terpreted as a quarter in which that em-
ployee was employed by a given company.
Any quarter in which an employee did not
have a payroll record in the Slavic database
was interpreted as a quarter in which that
employee was not employed. If an employee
went from employed status in one quarter to
not employed status in the following quar-
ter, that was considered a “separation” (or
turnover) for the first quarter. For example, if
an employee was employed by Company 1
in 2012 Q1 and 2012 Q2, but not employed
in 2012 Q3, that employee was classified as
having separated (left the employment of)
Company 1 sometime in 2012 Q2.
Consistent with the BLS definition of sepa-
ration rate, we then divided the total num-
ber of separations in each quarter by the
total level of employment (number of differ-
ent employees) for each company to calcu-
late a quarterly turnover rate.
We then compared each company’s turnover rate to the overall national average broken down by two variables. For turnover data, national data are
available by industry and by size (but not by state). There are a number of key differences in the BLS industry data and the BLS size data. The industry
data are official BLS data and are not seasonally adjusted to account for regular seasonal fluctuations in employment levels. In this way, they are a
better match for the data derived from the Slavic file, which are also not seasonally adjusted. However, some of the industry groups used for the BLS
industry data do not precisely match the standard industry categories available for the companies in the data file.
The data by size are relatively new and are classified as “experimental” (unpublished) data by BLS and are seasonally adjusted.19
It should also be
noted that these data are calculated at the “establishment” (typically location) level, while the Slavic data are primarily at the firm level (and could
therefore include multiple establishments). This should not affect the capacity to compare turnover rates overall, although there could be some effect
on specific size-based breakdowns.
The analysis by company size was conducted on the full 1,000-company sample. The analysis by industry includes 742 companies, excluding those
firms for which available industry information from the Dun & Bradstreet corporate database was either not available or did not align with the BLS
industry groups used for reporting turnover.
Business Survival and Failure Rates
We analyzed annual year-over-year business survival and failure from 2010 to 2013 for all “company-years” (e.g., Company 1 in 2012, Company 1 in
2013, Company 2 in 2013) in the Slavic database that had the following characteristics:
• Starting in 2008 Q1 (the beginning of the available data from Slavic), at least four consecutive quarters of using PEO services at any point prior to
the year being analyzed;20
• Located in one of the 24 states with the largest number of company records in the Slavic database and company status data available through
• Incorporated in 1994 or later (1994 is the earliest year for which BLS has survival rates available for later years).
For each company-year that met the criteria above, we then determined the firm’s survival status and assigned a “survival status value.” Com-
panies that were still active in the Slavic file at the end of 2013 or those classified as “active” (or similar, such as “good standing”) in the state
The foundation of our data analysis was based on data
provided by Slavic401k, a major third-party administra-
tor of 401(k) retirement plans that specializes in provid-
ing such plans for PEO clients. Slavic’s large scope
ensured that these data represented a broad cross-sec-
tion of companies that use PEO services. The data in-
cluded more than 12 million employee payroll records
and more than 5,000 PEO client companies from 2008
The Slavic data were used to determine the following
information about PEO client companies:
• When PEO services were used;
• The number of unique employees, by quarter (in-
cluding information on whether each employee re-
mained employed from one quarter to the next);
• Whether the company was known to be active (i.e.,
still a Slavic client) as of the end of 2013; and
• Basic company information (name, location, and
date of incorporation).
The data were then matched with additional company-
level information, drawn from the following sources:
• Dun & Bradstreet corporate database for informa-
tion on each company’s industry, as well as supple-
mental information about date of incorporation; and
• Separate state-level databases (typically main-
tained by the office of the secretary of state in each
for information about the current corporate
status of those companies that exited the Slavic
database before the end of 2013 (i.e., whether the
company is currently “active” or “inactive”).16
Finally, we compared aggregated results from the com-
pany-level data above to national averages drawn from
the following publicly available BLS data:
• Business Employment Dynamics statistics for year-
to-year firm survival rates, based on firm date of in-
ception, and including breakdowns by year, state,
and industry; and
• Job Openings and Labor Turnover Survey (JOLTS)
data for average turnover rates by month, including
breakdowns by firm size and industry.
While we took a variety of steps in the analyses de-
scribed below to ensure that the comparisons between
the Slavic companies and the overall U.S. population
were as valid as possible, it should still be noted that it
is possible that Slavic clients are not representative of
PEO clients as a whole. Most notably, it may be that
clients that offer 401(k) retirement plans to their em-
ployees vary in other respects as well when compared
to clients that do not offer such retirement plans. How-
ever, because nearly all PEOs (98 percent) offer some
type of retirement plan to their clients,17
we are com-
fortable that the Slavic401k data are indeed reasonably
representative of the clients of the PEO industry over-
corporate database through the first quarter of a given year were considered to have “survived” and were assigned a survival status value of
Companies that were classified as “inactive” (or “not in good standing”) were assigned a survival status value of 0 for the first calendar Q1 in which
they were no longer active (this date was determined based on state-provided information on last date of corporate activity). A survival status value
of 0 is assigned for a single year to each inactive company for the year in which the business was determined to have failed and the company is then
excluded from analysis in all subsequent years.
Because the characteristics of the companies available for analysis from the Slavic database do not match national averages (for factors such as year
of inception, state, industry, and size), we compared each company’s survival status value with BLS’s reported average national “survival rate of previ-
ous year’s survivors” for the appropriate analysis year and inception year cohort, broken down by either industry or state. This allowed us to calculate,
for each company-year, the difference between its actual survival status and its expected survival status. Calculating the rates in this way ensures that
differences in industry or state distribution across companies in the analysis database do not affect aggregate results. Finally, when we aggregated the
company data to overall averages, we weighted the sample by size group to be consistent with the size distribution of companies in the United States.
So, for example, if Company 1 survived into 2012, its survival status value would be 100 for 2012. If the average survival rate of previous year’s sur-
vivors was 95.5 percent in Company 1’s industry, the difference between actual and expected survival status, based on industry, for Company 1 would
be 4.5 for 2012. If Company 1 did not survive into 2013 and the average industry survival rate of previous year’s survivors was 97.1, the difference be-
tween actual and expected survival status would be -97.1 for 2013. We then convert these survival numbers into their corresponding business failure
rates for purposes of discussion.
When differences between actual and expected survival rates are averaged across all firms in the analysis database, a positive average survival
number indicates that PEO clients have a higher-than-expected rate of survival and a lower-than-expected rate of business failure. A negative aver-
age survival number would indicate that PEO clients have a lower-than-expected rate of survival and a higher-than-expected rate of business failure.
We calculated survival rates using multiple specifications that were designed to ac-
count for the effects of various imperfections in the available data. There were two
major categories of imperfections:
• Some inactive firms had apparent mismatches between date of incorporation
and date of corporate failure (e.g., the listed failure date was earlier than the
listed date of incorporation); and
• The status of some companies was not available or could not be found in the
state corporate databases; these companies were classified as “unknown” in
the original analysis file while additional adjustments were made subsequently,
based on further company-by-company research.
Multiple analysis specifications
The alternative specifications we used were designed specifically to adjust for fac-
tors that could be overstating advantages for PEO clients.
In particular, because the first category of mismatches cited above only affected in-
active firms (it affected approximately one-third of all inactive firms in the analysis
file), it would have the effect of overestimating survival rates and underestimating
failure rates for PEO clients when those invalid records were (of necessity) excluded
from the original analysis file. So, for one (more conservative) alternative specifica-
tion, we calculated a “worst case scenario” in which all of the affected firms were
classified as “business failures” that had failed in one of the four analysis years
(randomly assigned to occur equally across the four years, with no corresponding
survival in any previous years added to the database).
The underlying effect of the second data issue above (companies with an opera-
tional status of “unknown”) was less clear, as it was not known whether these “un-
known” companies were primarily active or inactive. Nevertheless, it seemed likely
that some significant percentage of this set of companies would be inactive. For
companies affected by this issue, we therefore conducted additional research in an
attempt to ascertain the status of those companies not available through the state
databases. This research included web searches as well as telephone calls to busi-
ness phone numbers in an attempt to ascertain current status. A second (“most
conservative”) alternative specification included as many as possible of these “un-
known” companies, classified as either active or inactive. For inactive companies,
we followed the same conservative technique as we did for the mismatched invalid
companies—we assigned business failure to each firm in one of the four analysis
years (and did not consider the firms to have survived in any other years).
In addition, apart from the issues above, the BLS data on survival rates also had
quality issues that primarily affected certain industries and certain data for the year
We communicated directly with BLS to understand the source of the issues,
with a particular focus on some reported BLS data that were clearly erroneous.
Based on their explanation and guidance, we chose to exclude certain data points
from the analysis, although it was not possible to entirely remove the effects of the
problem from the BLS data in this area.23
It should be noted that all of these errors
tended to overestimate national survival rates as reported by BLS (and thus reduce
the estimation of any advantage PEOs might provide to their clients).
Overall impact of data issues and multiple analysis specifications
The data uncertainties described above make it unclear which analysis specifica-
tion is most accurate. Taken together, however, the three specifications (as well as
separate calculations using state and industry comparisons) provide a range of esti-
mates of survival rates among PEO clients that can be viewed as providing a floor
and ceiling on the actual number. The first specification in Table 5 overstates sur-
vival (because we know it excludes some inactive companies due to inconsistencies
in their data). The second and third specifications, however, likely understate actual
survival, because they assign corporate failure dates to inactive companies without
allowing for the possibility that they had first survived for any years prior to their
failure. As noted, the unquantifiable errors in the BLS data also have the effect of
understating any advantage that PEO clients might have in terms of reducing busi-
We rely on the most conservative specification (listed in the right column of Table 3
as the most analytically responsible for summarizing the analysis results) because it
yields the lowest estimates of the advantage generated by PEOs. We use this spec-
ification as the basis for all of Tables 2 and 6. This file is also the largest sample,
including 4,508 company-years when compared with industry survival rates and
4,798 company-years when compared with state survival rates.
1 Sum of monthly turnover rate average for all private employers, as reported by BLS.
2 U.S. overall number controls for year of inception, year of analysis, and industry, and is weighted identically to the sample weighting described in this report.
3 For purposes of this paper, we define “business failure rate” as the percentage of businesses that do not survive in a given year. On average, 92 percent of all
businesses from one year survive into the following year, so the business failure rate is, correspondingly, 8 percent annually.
4 The SHRM worksheet is available at www.shrm.org/templatestools/samples/hrforms/articles/pages/1cms_011163.aspx.
5 William G. Bliss, “Cost of employee turnover,” The Advisor (2004). http://hrtogo.com/pdf/turnover-cost.pdf.
6 Matthew O’Connell and Mei-Chuan Kung, “The Cost of Employee Turnover,” Industrial Management 49:1 (January/February 2007).
7 Sum of monthly turnover rate average for all private employers, as reported by BLS.
8 Sum of monthly turnover rate average for all private employers, as reported by BLS.
9 Reported differences are the averages of company differences from expected turnover compared to industry averages and size group averages. They are calcu-
lated only for companies with expected values based on both their industry and size group (this excludes some industries for which comparable data are not
available in the BLS-reported industry turnover data; “Administration and Waste Services” is the largest such industry in the data file).
10 Sample sizes for these industries (based on the size group analysis) were: “Professional, Scientific, and Technical Services” (172); “Health Care and Social As-
sistance” (112); “Finance and Insurance” (86); “Manufacturing” (81); “Wholesale Trade” (61); and “Construction” (60).
11 U.S. overall number controls for year of inception, year, and industry, and is weighted identically to the sample weighting described above.
12 U.S. overall number controls for year of inception, year, and industry, and is weighted identically to the sample weighting described above.
13 Reported differences are the averages of company differences from expected failure rates compared to industry averages and size group averages. They are cal-
culated only for companies with expected values based on both their industry and size group (this excludes some industries for which comparable data are not
available in the BLS-reported industry data; “Administration and Waste Services” is the largest such industry in the data file).
14 Unweighted sample sizes for these industries were: “Professional, Scientific, and Technical Services” (1,105); “Health Care and Social Assistance” (461); “Ad-
ministrative and Waste Services” (458); “Manufacturing” (400); “Finance and Insurance” (354); “Construction” (335).
15 This process involved searching state records by company name and location. In those cases in which exact matches were not found, we examined a variety of
other measures, including date of incorporation, similarity of name, and geographic proximity of reported location. We used a conservative methodology in
which we did not classify a company as a match unless we had significant certainty that we had located the same entity, even if some specifics were not exact
16 All companies still active in the Slavic database at the end of 2013 were, by definition, classified as active for purposes of the analysis in this paper, which ends in 2013.
17 NAPEO, 2013 Financial Ratio & Operating Statistics Survey.
18 As discussed below, to determine employee separations, turnover analysis requires employee data from at least one quarter after the analysis period, so 2013
analysis was not possible based on the available data.
19 The firm size data break companies into six categories: 1 to 9 employees; 10 to 49; 50 to 249; 250 to 999; 1,000 to 4,999; and 5,000 or more employees.
20 Use of PEO services for a given quarter was determined by the presence of at least one payroll record in the Slavic database.
21 Because company status needed to be determined through manual research in each state’s unique online corporate database, we selected the 25 states with
the largest numbers of companies to maximize the number of companies included while reducing the number of states where it would be necessary to learn
database details. Massachusetts was included in the original 25 states, but companies in that state were ultimately excluded from the analysis because corpo-
rate information provided by Massachusetts is insufficient to determine a company’s current status.
22 BLS indicated that the data problems stemmed primarily from two issues: year-to-year counts for companies with multiple establishments in a given state; and
reclassification of some establishments from one industry category to another that had not been treated consistently across states.
23 We removed all BLS data reporting a survival rate of over 100 percent (a mathematical impossibility) as well as all data from a small number of industries classi-
fied by BLS as most affected by the error. Other data also overstated survival rates, but it was not possible to quantify the extent of the error, so we used all re-
About McBassi & Company
McBassi is an independent analytics and research firm that helps clients create consistently profitable and enlightened workplaces. McBassi uses
the language and tools of business—metrics and analysis—to build successful organizations by optimizing the power of their people. McBassi’s
principals (Dr. Laurie Bassi and Dan McMurrer) are co-authors of “Good Company: Business Success in the Worthiness Era” (winner of the 2012 Nau-
tilus Gold Award for Business/Leadership) and the “HR Analytics Handbook.”
About the Authors
Dr. Laurie Bassi is CEO of McBassi and a global leader in the field of applying analytics in the world of HR. Laurie is the author of more than 90 pub-
lished papers and books and was previously a tenured professor of economics and public policy at Georgetown University. She holds a Ph.D. in eco-
nomics from Princeton University.
Dan McMurrer is the chief analyst at McBassi. An analytics expert, Dan focuses on researching the relationship between organizations’ work and
learning environments and their business results. He holds an M.P.P. in public policy from Georgetown University.
National Association of Professional Employer Organizations
707 North Saint Asaph Street
Alexandria, VA 22314
NAPEO White Paper Series