COMPENSATION PERSPECTIVES

Update on Agency Suggestions

Updates on IMT's review of agency suggestions from the Compensation Reform Survey were published in the April and July 2004 issues of the AHRS Periodical. The status of the six most frequently suggested improvements are as follows:

Improving the precision within the job structure while maintaining its flexibility

DHRM has formed a small project team to assess the job classification process and ways to bring more definition to the job structure without reducing flexibility. The group is comprised of one human resource representative from each cabinet area.

Expanding the salary reference data to include more local, versus metropolitan sources

DHRM is adding additional sources of local salary data to the Salary Reference Data web site.

Reviewing the adequacy of caps on In-band Adjustments

Draft revisions to Policy 3.05 include a provision for exceptional adjustments that may exceed 10%.

Amending the definition of the In-band Adjustment for Retention

Draft revisions to Policy 3.05 expand the definition to include “ individual employees assigned to key functions ” as well as “all employees in a particular occupation or functional area”.

Requiring expiration dates for Temporary Pay

Draft revisions to Policy 3.05 include this requirement.

Increasing the cap on Recognition Awards

DHRM is in the data collection phase of the review.

Questions about these initiatives should be directed to your assigned DHRM Consultant.

Percents and Percentiles

When reading salary survey reports we often encounter references to percents and percentiles. The term “percentile” has a very specific meaning and cannot be substituted for “percent.” “Percentile” is not just a fancy way of saying “percent,” but refers to a very specific concept.

Percent
“Percent” is a general concept that allows any whole to be thought of in terms of 100 parts. Percents are formed by expressing fractions in decimal form and then multiplying the decimal fraction by 100. So if we have one-half of a pie, we have ½ = 0.5 (times 100) = 50% of a whole pie.

We can also think of a salary range in terms of percents. If we have the salary range $50,000 - $60,000, we can divide it into portions. We can think of the lower 50%, $50,000 - $55,000, and the upper 50 %, $55,000 - $60,000. We can think of a person who earns $52,500 as being 25% into the range.

Percentiles
Percentiles are not computed from percentages of the salary range or from any dollar amount. Instead, they are found by examining the sample of salary earners. In the simplest case, we would have a sample of 100 salary earners, each earning a different salary ranging from $50,000 to $60,000. We would then arrange these people according to the amount of salary that they earn with the person earning the least being 1 and the person earning the most being 100. The numbers, or ranks, assigned to these people would then identify the percentiles associated with the amount of money each earns. If the first person earns $50,200, then that amount marks the first percentile. If the 100 th person earns $58,555, then that amount identifies the 100 th percentile.

Any given percentile is defined as the value at or below which the given percent of the cases in the sample is found. So the 11 th percentile in a salary survey is the salary at or below which 11% of the number of people in the sample is found. In the example, if the 11 th employee is paid $55,000, then $55,000 is the 11 th percentile even though it is 50% into the range.

Salary surveys often report some special percentiles by using different terms. The 25 th percentile is also known as the first quartile. The 50 th percentile is known as the second quartile, or the median. The 75 th percentile is the third quartile. The inter-quartile range begins at the 25 th percentile and ends at the 75 th percentile. This range includes the middle one-half of the values.

What you need to know
The explanation given above using 100 people is a highly simplified account of how percentiles are found. It is a simplification of both the computational approach and the circumstances encountered. We rarely have exactly 100 people in our samples; and often several of them do earn the same amount of money. Sometimes our data are based on groups in which individual salaries are not known. These and other departures from the simplest case make computing percentiles complicated. For a full explanation of how to find percentiles when facing specific problems, we may have to consult a textbook on statistics. Fortunately, there are tools to help us to find percentiles. Excel, Microsoft's spreadsheet program, is one such computing tool.

Aging Salary Data

When comparing current state salaries with external data, we often use salary surveys that are several months old. We may, then, want to make reasonable estimates of current external salaries from this past data. In other words, we may need to “age” the salary survey data.

There are three basic approaches to the task:

1. Just treat the past salaries as though they are current.
2. Estimate current salaries from actually observed, or historic, increases.
3. Use an aging factor – such as used on the Salary Reference Data web site by estimating the current salaries from a salary increase forecast made in the past.

The first approach requires no further explanation. We simply record the observed salaries, the survey reporting them, and the effective date of the survey. It is reasonable to use this approach if the data are fairly recent or if market salaries are increasing very slowly.

The second approach to estimating current salaries from past data is to use actually observed increases. For example, the Bureau of Labor Statistics (BLS) reports the Occupational Employment Statistics (OES) results and the Employment Cost Index (ECI) reports. There are limitations to the use of this approach. There is considerable lag between the OES data collection and the report publication. At this writing, the most recent OES data on the BLS web site are for November 2003. ECI data are more current, but changes are reported on a quarterly basis, rather than a monthly basis. If the survey data are not effective the third month of a quarter, or if the current month is not the third month, then the market movement indicated by the ECI will need to be interpolated in order to age it to the present.

The third approach is documented in the DHRM publication, “Informational Guide on the use of Salary Reference Data,” March 2001. In it, we compute an aging factor from an estimate, often found in the salary survey being used, of how much salaries are expected to increase over the next year. For example, for several years prior to 2001 this forecasted annual increase was about 4.2%. This 4.2% rate of increase has been used to age survey data in the salary data viewer on the DHRM website. We assume that the increase amount is apportioned equally throughout the year. We, therefore, divide the annual increase amount by 12 in order to have a monthly amount. For example, 4.2% divided by 12 is 0.35% per month. We estimate the current value of salaries that were reported by a survey effective 10 months ago by increasing the reported salaries by 3.5% (i.e. 10 x 0.35%). So, if a survey reported a job with a salary of $50,000 in February, we would estimate the salary in the following December to be $50,000 * 1.035 = $51,750.

The two approaches (one based on forecasting, the other on actual, or historic, data) are not contradictory, but complementary. When historic data exist, they allow us to determine how good our forecasts were. When historic data come to an end, they may be supplemented by forecasted estimates.

Whatever aging approach we use requires us to make judgments about its appropriateness. We, therefore, should always describe the aging approach we use in any specific case so that our work can be critically evaluated. Aged data is always an estimate and we should update survey data periodically; such estimates are more likely to stray from reality as the aging period increases.

back to top