Voodoo Pay Management - or the Perils of Using Structure Data to Track the Market

In yesterday's post on aging survey data, I advised against using structure adjustment data as the basis for aging.  I thought it worth a follow-up post to expand on this topic and share what I believe are the perils associated with using structure and structure adjustment data to manage your salary program.

A quick pause for definitions:

By structure, I mean the salary ranges or schedules used to manage base pay.

By structure adjustment, I mean the amount by which that set of ranges or schedules is adjusted upward, typically on an annual basis, in response to competitive trends.

Structure and structure adjustment data is often offered by surveys in addition to data on actual salaries or actual salary increases.  The key difference?  Structure data represents policy, where salary data represents actual practice.

I think this distinction is an important one.  Having practiced in this field long enough to see how a wide range of organizations actually manage pay, I can tell you that there may be sizable discrepancies between policy and actual practice.  A couple of extreme (but very real) examples to illustrate my point:

Organization A - This organization is facing financial challenges, which have translated into a struggle to keep employee pay competitive.  The organization has faithfully reviewed and adjusted its salary ranges every year, so that they are kept at market levels, but has lagged behind in delivering employee increases.  The result is that while their ranges (as evidenced by their midpoints) are competitive, actual salaries are very low in comparison to both ranges and the external market.  And so, you see, their structure tells a very different story about their pay program and practives than their actual salaries do.

Organization B - This organization has been in business for over 50 years in a relatively stable industry.  Turnover is low to nonexistent, and the average employee has been in their job for over 8 years.  The organization, because it has not experienced much competitive pressure for people, has chosen to manage its salary program and ranges at the lower end of market practices.  Employees, though, being such a long-tenured group, tend to be paid at or near (or even over) the top of their assigned ranges.  Here again, the organization's structure might tell you a very different story about the pay program and practices than their actual salaries would.

My point?  Structure data (e.g., average minimums, midpoints and maximums) and structure adjustment data (how much the average organization has moved its structure) can paint a very different - and even misleading - picture of market pay practices in comparison to using actual salary data.  Salary data - actual salaries and salary increase percents - tell you what's really being done, regardless of what kind of policy and structure posturing might be happening. 

Better to base your analysis and decisions on reality - what and how people are really being paid.  Pay decisions based on anything else ... well, my title speaks for itself!

Aging Survey Data is About Mimicking Market Pay (Not Structure) Movement

Reader Christine asks about the best way to age survey data; whether to use salary structure movement or salary (merit) increase budgets as the basis for aging.

Aging survey data, for the uninitiated, is the process used to update or bring the survey data forward, covering the gap between its actual effective date and the date you need to apply it.  Because most pay surveys are conducted on a periodic basis (the most prevalent schedule still being an annual one) and because in most cases there is a gap - minimally - between the collection/effective date and the date of survey publication, the need to age the survey data is simply an accepted fact of life.

The question that Christine poses, then, is what rate to use in aging the data.  It's a good question, since most salary planning surveys give you both sets of data (salary increases and structure adjustments) to select from. 

Since our ultimate objective is to mimic the rate at which market pay is moving, using the average salary increase amount or salary increase budget is a better indicator of how much the average employee's salary is moving in a year.  Average structure adjustments, on the other hand, reflect the degree to which organizations are moving their pay ranges in order to keep them competitive, which can and often does differ from the average increase planned or delivered.  My experience in watching this data over time would suggest that, on average, structure adjustment numbers tend to run at 60%-70% of salary increase numbers.  Using structure adjustment rates to age your survey data, therefore, puts you at risk of underestimating actual market pay growth.

(In fact, I would argue that structure adjustment data, while certainly interesting, ought to be disregarded entirely when making pay programs decisions, including how much to adjust your own salary structure(s).  Pinning your salary range movement to a rate that lags actual salary movement by 30% to 40% will ultimately leave you with a structure that lags the market.  But perhaps that's a topic for another post.)

My advice in summary: Use salary increase - not structure adjustment - numbers as your rate for aging survey data to ensure that you are simulating market pay movement as accurately as possible.

Selecting Survey Data: Follow the Labor Market

When you are market pricing jobs and find yourself in the enviable position of having multiple surveys and/or multiple survey cuts to select from, the question becomes what data do you use?  Do you focus on the local geographic area where the job is located?  The industry?  Organizations similar in size?  Some combination of the above?  The right path may not be clear, particularly when you are new to market pricing and/or compensation.

I'd like to suggest a rule of thumb to guide you in your selection:  Follow the labor market.

The labor market is your particular market for talent.  It represents where you draw your talent from and where you potentially lose it to.  And, most importantly, it is the external "stake in the ground" against which you set and assess your pay practices.  Which makes its definition a key philosophical question underlying the design of your compensation program.

For most organizations, the labor market is not a singular one-size-fits-all-jobs phenomenon.  The typical organization draws from different labor markets for different job groups.  Executives might come strictly from within your industry, from organizations comparable in size/scope, but from anywhere in the nation.  The labor market for specialized jobs like sales engineers, alternatively, might be any size/scope organization within your industry, but strictly the local or regional geographic area.  Administrative support staff, on the other hand, might be drawn from the local geographic areas, from any industry or organization size.

If these represent the labor markets for your different job groups, then your selection of survey data should mirror this to the extent possible.

I sometimes find it helpful to chart out an organization's labor markets in a matrix, like the one shown below, to guide survey selection and market pricing efforts.

Clearly defining your organization's labor markets can provide helpful guidance through the all the choices involved in the market pricing process.

Creative Commons Photo "Arrow" by Paul Downey

Free Salary Data Resource

The Bureau of Labor Statistics has just released its May 2007 Occupational Employment Statistics Survey, which provides mean and median wage data for more than 800 occupations across 375 MSAs (metropolitan statistical areas). 

This is essentially a free salary survey, and one which I have found it to be a good supplemental source of data.  My experience is that it syncs up well with other sources that cost me a lot of money, and it has the additional benefit of covering metro areas and occupations that are not well covered by other professionally published compensation surveys.  I have found it particularly valuable for its wide range of unskilled and semi-skilled "hourly" occupations, which I often have trouble locating in other sources. Further, it can be a limited but helpful source in examining geographic differentials for select groups of jobs.

So, something very good created with our tax dollars ... and a no-cost addition to your bag of tricks.  Cheers!

Post update:  Just want to add to the body of this post - a couple of savvy readers (thanks, Carla and R) have noted some limitations to this data that are worth mentioning here.  First, there are no "levels" for any of the jobs covered.  In other words, there is an Accountant, but not a Junior/Associate Accountant and not a Senior Accountant.  One level only.  Second, the data is not broken out by company size.  For most jobs, I think that this is OK - but I do not use this source for management jobs for this reason.  Otherwise - a gold mine of free salary data.  Enjoy!

The 70% Rule: A Guide to Market Benchmarking

When working to match an organization's jobs to published compensation surveys, and especially when involving line managers and supervisors in this matching process, I have found it helpful to use something I call "the 70% rule". 

The rule works in this way:  When trying to assess whether or not a survey job description fits, or matches to, one of the organization's jobs, we use 70% as our guidelines.  If the survey job description appears to capture 70% or more of the job content, then we call it a match.  If it doesn't meet the 70% criteria, we don't use it. 

I am a fan of involving line managers in survey matching for the jobs reporting to them, and I find that this rule of thumb helps managers with that fuzzy, challenging  and sometimes disconcerting task.  Of course, it also helps to remind them that the benchmark jobs found in surveys are not "real" jobs, but rather common roles that exists across many organization.  For this reason, survey job descriptions are purposefully brief and generic.  They must be defined broadly enough that they can provide comparisons across a range of organizations, so that a critical mass of pay data can be collected.  If they were defined in a way that exactly matched the jobs in your organization, the likelihood that any other employer would see them as matches and provide data is pretty slim.  Then you have a perfect match - but no competitive pay information.  So, yes, at the end of the day, the act of market benchmarking must involve some compromise.  And because this involves art as well as science, it helps to have a guideline like the 70% rule to help us make the calls.

Note that there are multiple versions of this rule floating around the rewards/compensation profession, including "the 75% rule" and "the 80% rule".  The concept is the same, the percents vary slightly.  I was raised on 70% and that works for me.

National Median Total Cash for Small Business CEO is $276,327

Finding solid executive compensation data appropriately scoped for small businesses can be - in my experience - like looking for the proverbial needle in the haystack, so I was interested to learn of the annual Small Business Executive Compensation Survey published by Salary.com, and thought it worth mentioning here.

This year’s survey covers 12 executive positions and includes data from 2,237 organizations, representing an average company size of 92 employees and a broad range of industries, geographies and ownership structures.

A few summary statistics, featuring national median total cash (base salary plus any short-term incentive), courtesy of Salary.com:

More information on the survey can be found here.

Survey Quartiles Are Not Salary Ranges, And I'll Tell You Why

Most professionally published salary surveys report, in addition to mean or average salaries, salary quartiles.

Quick stats lesson (with a little help from Wikipedia):  A percentile is the value below which a certain percent of pay rates fall for a particular reported survey job.  So the 75th percentile is the level below which 75% of the reported pay rates may be found.  The 25th and 75th percentiles are known as the first and third quartiles, respectively, and the 50th percentile is known as the second quartile or median. 

There is a tendency, a growing one in my experience, to simply lift the quartiles reported in a salary survey and use them - as is - as a de facto salary range for your job.  In the most common variation of this, the first quartile becomes the minimum, the median becomes the midpoint and the third quartile becomes the maximum.  While this is an easy (and apparently common) way to establish a salary range for a job, I advise against it.

Here's why.

One of the primary reasons for creating salary ranges for jobs is to establish some structure to guide salary decisions in order to ensure that they are made in a fair and consistent manner.  And if you use a standard approach to develop ranges - e.g., a midpoint established at market value, minimum at 80% of midpoint and maximum at 120% of midpoint for a symmetrical range 50% wide from bottom to top - you will accomplish this.  But if you pull quartiles from surveys and create ranges from them without ensuring a consistent pattern, you are setting up arbitrary ranges which may vary substantially in width and pattern from job to job, and which will - invariably - produce erratic and capricious salary decisions.  Sound like a good idea to you?  Me neither.

Better instead to use a consistent and well-thought-through approach to creating ranges - either from market values or from values generated through an internally-based job evaluation approach.

Too Fancy? Thoughts on Weighting Schemes for Pay Survey Data

Whether we are using sophisticated software products or basic Excel spreadsheets to organize and analyze compensation survey data, we compensation professionals have a tendency toward using elaborate weighting schemes in developing overall market values for jobs.

What do I mean by this?  Well, let's say you are "market pricing" (gathering competitive market pay data on) a Software Development job.  Let's further say that you've identified three good "matches" from three separate salary surveys for this job, as noted below:

Survey 1:  Job title - Software Developer

Survey 2:  Job title - Software Developer - Intermediate

Survey 3:  Job title - Software Engineer 2

A simple approach would be to weight each of these survey jobs equally - or, essentially, calculate a simple mean or average of their values.  A more complicated approach would involve developing a "composite" (a weighted average of sorts) which reflects weighting each survey job differently.  For example:

Survey 1:  Job title - Software Developer (10%)

Survey 2:  Job title - Software Developer - Intermediate (60%)

Survey 3:  Job title - Software Engineer 2 (30%)

Why, or on what basis, would we treat these survey jobs differently?  Reasons abound and can include (but are not limited to):

1. A desire to reflect the fact that some survey matches are a better fit than others
2. A desire to reflect the fact that some survey sources are a better fit than others
3. A desire to place more or less emphasis on a particular industry sector or geographic area
4. A desire to weight each survey job proportionate to the number of employees it represents (e.g., a piece of data that reflects the pay rates of 400 employees would be weighted twice as much as one that reflects the pay rates of 200 employees) or, in other words, to develop an "employee-weighted" average.

My take?  I have to say that I lean toward simplicity, and advice my clients accordingly.  The simpler the survey weighting scheme, the better - unless there is a logical and compelling reason to get "fancy" (and sometimes there truly is).  In my experience, however, getting fancy often becomes a slippery slope to a place that is difficult to explain and defend.  Not a good place to be in these times of greater pressure for pay program transparency.

P.S. If you must get fancy, make sure and clearly footnote your weighting approach and rationale.  You - and anyone who must follow behind you and unravel your logic - will be glad you did.

Mothers Let Your Babies Grow Up to Be Computer Science & Engineering Majors

Basic tenets of economics, like supply and demand, account for a lot of the variances we see in the market pay level for different skill sets and disciplines.  A shortage in a particular area, which may exist for a host of different reasons from a particularly rigorous educational requirement to a simple case of not enough people feeling attracted to the field, will cause a spike in the price of that talent.  Often, but not always, that spike in pay will spark interest and the flow of people to the field will increase.  If that happens, the imbalance between demand and supply will lessen and the growth in pay will level off.

At least, that's how it's supposed to work.

But now the National Association of Colleges and Employers (NACE) tells us, in a recent release based on their Fall 2007 Salary Survey, that this basic economic rule is failing to kick in.  Some of the highest demand educational fields, which command the highest starting salaries for graduates, are facing declining enrollment.  Computer science graduates, for example, according to the NACE, are earning average starting salaries of $51,992 and yet the field is seeing a negative growth rate (-0.6%).  The situation is even worse for engineering disciplines, where the average starting salary is $53,710 and the growth rate is -2.2%.

So what fields are attracting more students?  Visual and performing arts tops the list, with an annual growth rate of 2.6%, followed by psychology at 2.3%.  This in spite of starting salaries - $30,174 and $31,857 respectively - significantly below those associated with the aforementioned more technical disciplines.

This could spell trouble ahead, as most organizations will be seeking more - not fewer - employees in technical fields like computers and engineering in the future.  Pay alone, as the NACE points out, may not be the answer - but we may have to prepare ourselves for premium pay levels in a number of these disciplines.  Employers, particularly those in industries which demand a significant supply of technical talent, may also have to work harder to make these jobs and opportunities more appealing - and to become an "employer of choice" for this particular set of people.

In the meantime, never too early to encourage those kids along in their math and science classes.  And for those of us with kids in and/or approaching college, real food for thought and discussion!

Postscript:  Perhaps I am dating myself with the title of this post, which I expected most people would "get".  But my teenage daughter, who occasionally reads this blog, had to ask for an explanation, to which she replied, "WHO is Willie Nelson?"

Mean Versus Median in Survey Statistics

When using compensation surveys to determine the "going rate" for a particular job - or what I like to call the estimated market value - we are typically presented with a number of descriptive statistics to choose from.   Since most of us are looking for information on the market center, we will have the choice of a couple of measures of central tendency:

Mean - the mathematical average, calculated by adding up all the pay rates in the data set and then dividing by the number of pay rates in the data set.

Median - the value of the pay rate that falls in the middle of all the rates in an ordered data set (that is to say, a data set which is ordered from lowest to highest rate).

(The other measure of central tendency, which rarely if ever shows up in compensation surveys, is the mode, or the pay rate which occurs most often in the data set.)

I have a strong preference for the median, over the mean.  And I found a great explanation of my preference over this past weekend in my daughter's psychology textbook (Psychology, the 8th Edition, by David G. Myers, Worth Publishers).

With income distribution, the mode, median, and mean often tell very different stories.  This happens because the mean is biased by a few extreme scores.  When Microsoft Chairman Bill Gates sits down in an intimate cafe, its average (mean) patron instantly becomes a billionaire.

In other words, the mean pay rate in a compensation survey will be affected by any extreme pay rates in the data set, where the median will not.  For this reason, I believe the median (where it is offered; some surveys only provide the mean) is a better and more reliable measure of central tendency to use in pay program assessment and design.