TOPICS

Introduction to the Salary Calculator

What are the technical platform requirements for using the Salary Calculator?

Can I explore the implications of a move to a different area?

How do I allow for the passage of time since this system was created?

How do I use the range data?

Is there a way to compare with national rather than regional or local scales?

Technical Notes on the Salary Estimation Models

What if I'm working part time?

Why do salary values decline somewhat for some of the most experienced people?

Introduction to the Salary Calculator

Here is a quick overview of this tool. Users are asked to supply information on an employment situation. This may be the user's own current job -- one of the more common uses of this system is preparation for salary reviews -- but the salary calculator can also serve other purposes. For example, a student might want to see what current compensation is like for different stages of career development. A manager may want to learn what competitive scales of pay may be for particular technical people. An experienced engineer may wish to test the implications of changes in technical specialties or job functions.

Detailed specification of employment situations is supported, including allowance for education, years of experience, professional specialty, job functions, employer characteristics (size, general types, and lines of business), and geographic location. When the position has been defined, the calculator uses data from IEEE-USA's surveys of its members to compute ranges of base pay and income from primary sources for similar jobs. "Income from primary sources" has been the principal measure of compensation used by IEEE-USA; it includes base pay plus any net earnings from bonuses, commissions, or self employment. Base salary levels are also a major topic of interest for many users, and in 2004 estimates of this component of primary income were added to the calculator system. All of these estimates of earnings are derived from information supplied by engineers working full time in their primary areas of professional specialization, that is, by appropriately employed persons.

The ranges of compensation values generated by the system center on estimates of median income; medians show the 50th percentile of a range of pay (that is, the value that is exactly in the middle of all cases); they are a preferred statistic for compensation data because they are not distorted by extreme values at either end of a distribution of earnings. In addition, the calculator also displays outlying values around the medians: 10th (low), 20th, 30th, 40th, 60th, 70th, 80th, and 90th (high) deciles. These range estimates provide a way to allow for variations in pay that are not measured by surveys -- such as a person's skill and talent as an engineer.

Many advanced features are incorporated in this salary comparison system. Most of them are invisible to users but they still influence results. These features include adherence to smoothed "maturity curve" estimation practices which have long been used by professional compensation analysts, especially in high tech situations; the ability to make simultaneous allowances for many different aspects of an employment situation, yielding a single set of results that allows for all of the conditions specified by a user; reports on the specific net impact of particular conditions on pay -- that is, the additional effect of a given condition after also allowing for the impact of all of the other factors used by the system; and the use of optional information on levels of professional responsibility, which if supplied can substantially improve the accuracy of compensation estimates. Salary figures are also automatically updated to allow for the passage of time since source data were collected.

Return To The Top Of The FAQs Page

• Microsoft Internet Explorer 5.5+ for Windows
• Safari 1.2+ for MacOSX
• Mozilla 1.3+ for Windows/MacOSX
• Firefox .8+ for Windows /MacOSX

• Cookies: Enabled
• JavaScript: Enabled

Can I Explore The Implications Of A Move To A Different Area?

Yes, but you must be aware that comparisons of pay for engineers in different parts of the USA may not necessarily capture all of the relevant aspects of your situation. Some industries may be clustered in particular regions, influencing general levels of pay for those areas. For example, pay for IEEE-USA's members in the San Francisco Bay area is strongly influenced by conditions in the Silicon Valley. If you define a situation for a position that is not in information technology, comparisons with the Bay metropolitan region may be misleading. The data here are useful and valid, but engineers who are considering a change in location will also be well advised to consult one of the web-based services for relocations, such as the international salary calculator for relocations, cost of living comparisons, and real estate, located at www.homefair.com. This service will provide additional information based on pure differences in local conditions, and will not be influenced by such factors as biases in the distribution of different kinds of engineers among geographic areas in the USA.

Return To The Top Of The FAQs Page

How Do I Allow For The Passage Of Time Since This System Was Created?

The results supplied by IEEE-USA's Salary Comparison System are based on data collected in late 2004 on income from primary sources -- that is, base pay plus any income from bonuses, commissions, or self employment -- for the year 2003. The middle of that year is taken as the specific applicable time for these annual totals. All results are then updated to the current time period before they are presented, based on changes since mid-2003 in the U.S. Consumer Price Index. In short, the information you obtain here has already been adjusted to allow for the passage of time. Note that IEEE-USA has decided to conduct its salary surveys annually, rather than every other year as had been the case in the past. This means that each year fresh information will be used to generate these comparative results.

Return To The Top Of The FAQs Page

How Do I Use The Range Data?

Pay is influenced by many important factors which cannot be accurately measured in surveys, such as a person's individual skill, the differing situations and compensation policies of particular employers, negotiating abilities, and sheer luck, either good or bad. As a result, there are very wide variations in the salaries of people with otherwise identical job situations and levels of experience. Median salaries are good estimates of typical pay levels for a position, but many people are not typical.

For example, if you believe your boss will agree that you are a superior performer, compared to most other people with jobs like yours, then you may be able to argue that you deserve to be paid at (say) the 80th percentile level. Or if the comparisons show that your pay is substantially below typical levels for your situation, this may be evidence that an adjustment is in order.

Obviously, negotiating skills are crucial in these kinds of situations. Experience suggests that the best approaches usually involve a combination of objectivity and friendliness. Bear in mind that it is management's job to contain costs, while it's up to you to ensure that your income is in line with your skills and experience.

Many users have credited IEEE-USA's salary comparison system with helping them achieve higher pay in their current jobs or a better initial salary offer. However, please note that IEEE-USA cannot assume liability for any individual's use of the data provided by this instrument. How you use the information is your responsibility.

Return To The Top Of The FAQs Page

Is There a Way to Compare With National Rather Than Regional Or Local Scales?

Yes. In some situations, especially for highly skilled experts, job markets may be national in scale. Not that one should accept pay at national levels if local standards are higher! Rather, in some situations, employers in locations that normally pay less may seek to match national levels of pay in order to attract the people they need. To get national results with no adjustments for region, divide salary values by the factor values for region and metro area, or use a region and metro combination where the factors are 1.0, such as the Pacific region and the "Other" choice for metro area.

Return To The Top Of The FAQs Page

Technical Notes On The Salary Estimation Models

Notes contributed by Richard Ellis of Ellis Research Services (ERS), Carlisle, Pennsylvania; Dr. Philip Broyles, a professor at Shippensburg University who takes on statistical consulting jobs as E-Statsman, Chambersburg, Pennsylvania; and Brett Matzke of CAMO, Inc., Woodbridge, New Jersey. Ellis directed IEEE-USA's salary surveys in 1995, 1997, 1999, and 2001, and served as an advisor and analyst for the surveys in 2003 and 2004. He was the original developer of the salary comparison system. Broyles created the specific statistics used to estimate salaries in 2001 and 2004. Matzke was the developer of the system's statistical engines in 2003, adding several new features to the models.

The system used to generate comparative information on base salaries and income from primary sources for IEEE-USA's members uses log-linear regression models. The organization began testing such approaches in 1991. By 1997, the methodology was sufficiently refined to justify the production of a publication, Salary Benchmarks: A Personal Workbook, which made these tools available to members at low costs. Today, this workbook has been replaced by the current web-based system, which automates all computations and provides range data as well as basic general results in the form of estimates of median earnings.

The cases from IEEE-USA's 2004 Survey of Salaries and Fringe Benefits that were used to create this system were limited to persons receiving $20 to $250 thousand dollars per annum, and years of experience were limited to 50 or less. Cases with any missing data for any item used in the models were also excluded from the data sets. The final size of the dataset was 9,386 cases.

In these models, annual income from primary sources is specified as an exponential function of one's work experience (years and years squared), education, geographic region, and numerous organizational and industrial characteristics. The general form of the predictive equation can be stated as follows:

P = C(Y₁^y)(Y₂^y2)(I₁)(I₂)...(I_n) where

• P = salary

• C = constant (intercept for salary)

• Y₁^y = first coefficient for years of experience, raised to the y^th power where y = years of experience (note that if y = 0, the effect of this computation is to leave the constant unchanged)

• Y₂^y2 = second coefficient for years of experience, raised to the y squared power where y = years of experience (again, if y = 0 the constant is unchanged; the effect of this second computation is to gradually reduce the incremental impact of an additional year of work, yielding results which match the traditional maturity curves for pay which have long been used by compensation analysts)

• I₁ = first adjustment value, as applicable

• I₂ = second adjustment value, as applicable

• I_n = N^th adjustment value, as applicable

• (Note: the 2003 regression models created by Brett Matzke used additional linear variables and thus a somewhat more complex general form of the equation reproduced above. In 2004, Dr. Broyles was able to achieve improved results by returning to the simpler approach used in all previous years. Some of the innovations added by Matzke are retained in 2004 as Boolean, rather than continuous, conditions. Other potential improvements suggested by Matzke remain under study and may be restored in future versions of this system.)

A "default" situation is defined which reflects the income of persons for whom no adjustments to the estimates are required. For the data on the 2003 calendar year obtained by the 2004 survey, this reflects situations for recently hired (less than one year of experience) engineers with a bachelor's degree, a technical specialty in circuits and systems, and a primary job function of systems software engineering. These engineers work for a large (over 10,000 U.S. employees) private electrical/electronic services company in IEEE's Northeast region, not located in the Boston, New York, or Rochester metropolitan areas.

Each regression model generates a constant value which is equivalent to earnings for this default case. When the lowest possible levels of experience and professional responsibility are selected as conditions for a run of the system, the resulting values become estimates for entry-level positions.

Estimates of income from primary sources are then adjusted to allow for all of the following. First, two continuous variables (years of experience and years of experience squared) adjust starting salary values for the impact of years of professional and appropriate managerial experience (time in undergraduate or graduate school is NOT counted). The years squared adjustment has the effect of slowly reducing the size of each year's increase, so that results resemble the smoothed maturity curves long used by compensation analysts who examine pay scales for technical professionals. Pay rises most rapidly in the early years and gradually begins to plateau for mature workers (for whom the effect of another year of experience is somewhat less significant).

A large set of dichotomous indicator variables is then used to reflect the presence or absence of other employment characteristics. These variables were tested in stepwise fashion, with a .05 level of significant entry criteria, beginning with the most influential predictor. For a variable to be included in the final models, it had to make a statistically significant contribution, at the 0.1 level, to account for differences in income beyond what was already explained by other variables previously incorporated into the system. These factors include:

• 3 possible higher degrees (masters, Ph.D., and JD);

• 42 technical specialties;

• 14 principal job functions;

• 2 types of supervisory responsibility, for technical and non-technical people (supervision of non-technical people does not have significant effects on earnings in the 2004 system, but is being retained because it may produce such effects in the future);

• 8 broad types of employers;

• 5 categories of employer size;

• 14 employer sectors (lines of business); and

• 31 U.S. metropolitan geographic locations, plus nine additional regions for those not located in any of these major urban districts.

A separate estimation system was constructed that adds data on levels of professional responsibility to the above list; because this is a completely independent mathematical model, factors that do not yield significant effects on pay when levels of responsibility are considered may turn out to be significant if responsibility is not considered; the converse also applies. IEEE-USA obtains data on nine such levels plus a tenth category for those for whom no level can be appropriately defined. For these models, the first two (lowest) levels are combined, as only a few members of IEEE-USA are at these lowest levels of responsibility. For details on the definitions of these levels, readers are referred to the comparative system itself, which includes full definitions for each of these classes of professional duties. As noted elsewhere, if a level of responsibility cannot be defined for an employment situation, the system will revert to the simpler approach that leaves that factor out of the calculations; in short, the use of this variable is strongly recommended but optional.

In addition, other factors, such as gender and minority ethnic status, were tested for the models but did not contribute statistically significant effects to earnings estimates.

The results of the operations of these models take the form of an estimate of the median (50th percentile) income for engineers with the specified situation. Medians are the preferred measure for all examinations of income, because they are not influenced by very high values at the high end of distributions of pay. In addition to medians, eight additional values -- the 10th, 20th, 30th, 40th, 60th, 70th, 80th, and 90th percentiles -- are also calculated, to provide estimates of the ranges of pay around these medians for better-paid or worse-paid people in similar situations (for further comments, see "How do I use the range data?" in these FAQ pages).

In addition, all results are updated to allow for the passage of time since the middle of 2003, the point taken as the reference time for IEEE-USA's data. These updates are based on changes since mid-2003 in the value of the U.S. Consumer Price Index for Urban areas (CPI-U), a widely accepted measure of inflation.

The models use adjustment coefficients to adjust salary estimates for situations that differ from the baseline. For example, to calculate estimates of earnings for a person with a master's degree and 10 years of experience, a primary job function of basic research and a professional specialty in magnetics, working for a very small (1-50 U.S. employees) private defense-oriented organization located in Austin, Texas -- all other characteristics of the situation being similar to the baseline profile -- the following adjustments would be made:

• Allowance for 10 years of experience raises estimates of base salaries to $73,697 and estimates of income from primary sources to $80,607.

• The master's degree raises the estimate of base salary by a factor of 1.077045 (roughly 7.7 percent), and the estimate of income from primary sources by an almost identical adjustment (1.077230).

• The basic research job function raises the base salary line estimate by 1.073762, or about 7.38 percent, and it also increases the estimate of income from primary sources, by 1.052077 (about 5.21 percent).

• The magnetics professional specialty lowers the baseline estimate by a factor of 0.923541 (approximately -7.65 percent), but estimates of income from primary sources are not affected by this factor.

• The employer's small size (1-50 employees of all kinds in the USA) lowers the estimate for base salaries by 0.918798 (about -8.12 percent), and the same factor also lowers the estimate for income from primary sources by a similar amount (0.911387, or about -8.86 percent).

• The employer's line of business as a defense-oriented private company lowers the baseline estimate by 0.966210 (about -3.38 percent), but a separate adjustment for basic defense industries raises it back up by 1.036949 (about +3.69 percent), so the net effect of defense-related employment is a small gain in base salary. In the separate set of calculations for income from primary sources, there is a similar reduction of 0.964185, or about -3.58 percent, for defense-related lines of business, but no offsetting upward adjustment for the broad defense industry variable, so the combined effects in this case reduce the final estimates.

• The Austin location raises base salary estimates by 1.097835, or about 9.78 percent, and it also raises estimates for income from primary sources by a similar degree (1.096646, or about 9.66 percent).

• Multiplying baseline values for the estimates by all these various adjustment factors yields final median values for this position: a base salary of $79,549 and income from primary sources (base salary plus bonuses, commissions, and any net income from self employment) of $88,036. These values are subjected to further adjustments before being displayed to the user, to allow for inflation in the value of the dollar since mid-2003.

The results pages of the web site reports the effects of the factors which have been used for the specific situation defined by the user. Those factors are reliable estimates for the typical net impact of the specified condition on income scales for electrical, electronics and computer engineers. Often, a specified condition has no significant net effect on income; that is, it does not generate statistically significant improvements in the estimates produced by the other factors used by the system. If so, the results page also notes this outcome.

How good is this approach as a way to provide data on compensation? Several standards can be applied. First, in regression models, a summary statistic, R squared, reports the proportion of all variance in a predicted value which can be attributed to the measures incorporated in the system. The value of this statistic in the IEEE-USA model for base salary comparisons using data on levels of professional responsibility is .554, meaning that well over half of all variations in the pay of these engineers is covered by the situational factors examined here -- experience, degrees, specialty, location and other relatively objective conditions. If information on levels of professional responsibility is not provided, the precision of the models drops somewhat, with the value of R squared at .494, or just under half of the variation in income. The precision of the models drops slightly for predictions of income from primary sources, to .548 and .479, respectively; the decline is probably attributable to the relative unpredictability of such factors as bonus awards.

What accounts for the remaining unexplained variations in pay? Some may be due to simple errors in data, but most of the remaining differences in what the members of IEEE-USA make are likely to be due to such factors as sheer skill, the varying generosity of different employers, negotiating ability, and simple luck, both good and bad. None of those factors is easily measured and few if any sources of data on compensation allow for them. The importance of these less easily measured factors is the reason why the salary comparison system includes information on salary ranges around the median values as well as the basic medians themselves. The range data provide a way for users to consider their own circumstances and adjust results accordingly. See "How do I use the range data" in the FAQs for further comments.

Another test of the appeal of systems for the presentation of information on compensation is the simple plausibility of the results. For example, many salary surveys present graphed maturity curves for changes in pay with increasing experience. Because of sampling anomalies which will influence the smaller data sets that must be used to compute salary levels for people in particular professional specialties, experience categories, both of these factors, or more, such data may yield obviously incorrect results, such as a finding that people with more experience are receiving lower salaries than those with less experience. To deal with this problem, curve smoothing routines are customarily used to present such data, but the use of such corrections is not foolproof and it can raise entirely new problems of its own. None of these drawbacks apply to the regression approach, which can be described as self-smoothing.

A third test is the ability of a model to provide an accurate reflection of a user's situation. Pay is influenced by a host of factors, and data that reflect only one or two of these may be useless. Separate results may be presented for variations by experience, by specialty, by job role, by degree level, or by other conditions of importance, but conventional approaches have provided little or no guidance for people who need to take all of these factors into consideration at once. The appeal of regression approaches is that they can eliminate this problem entirely. The salary comparison system developed by IEEE-USA reflects the simultaneous influence of many different kinds of factors that affect the salaries of its members, and generates a single set of results that allows for all of these factors at once.

The most telling test of all is the actual experience of users. At this writing, IEEE-USA' system has been tested by more than five years of experience with web-based salary comparisons systems and the earlier printed workbooks, and reactions have been very favorable. Members often report that the system has an uncanny ability to correctly estimate their income, especially if allowance is made for range variations. Those who are making more or less than the typical electrical or electronics engineer know that their situations are atypical and they find that the range data accurately reflect these variations from the norm.

Return To The Top Of The FAQs Page

What If I'm Working Part Time?

To provide consistent comparisons, the system generates only data on income from primary sources for full-time positions. To facilitate comparisons for part-time positions, hourly rates are also reported, by dividing annual salaries by 2,080 (52 weeks times 40 hours/week). If different standards for the number of hours in a work week apply to people in your situation, you may want to adjust this computation. Note that the number of work week hours is an "official" or conventional figure and bears no necessary correspondence to actual hours. Most engineers are treated as professional staff and are not paid for overtime. Many employers may have official work weeks of 35 hours, for example. This does not mean that people actually work that amount of time; for many years, IEEE-USA's surveys have reported that its members typically put in around 45 hours a week, but for the purposes of converting annual to hourly rates, the official and not the actual hours should be used. Also note that a full 52-week year should be assumed, with no allowance for leave.

Return To The Top Of The FAQs Page

Why Do Salary Values Decline Somewhat For Some Of The Most Experienced People?

Levels of professional compensation for groups of people usually decline a bit after peaking during the middle years of experience. To some extent, this may be an effect of the composition of the group; some of the most highly paid personnel may take early retirements or may move into other professions, so that they no longer contribute to the results. In other words, compensation for the group declines because it contains fewer of the most highly paid people. Pay for individuals does not decline (assuming they are not downsized or otherwise forced to accept inferior jobs); instead, after many years of experience, the marginal value of further time on the job becomes negligible, and scales tend to become flat. If you are a very mature engineer, this means that comparative data on pay may understate your compensation. To correct for this, try using FEWER years of experience. In IEEE-USA's 2004 results, income estimates peak at around 30 years of experience. This approach will yield a peak figure which should apply for the remainder of an individual career.

It is also possible that declines in pay are real. Many older engineers report instances of age discrimination and of being forced to accept such changes as conversions to contracted or temporary status. It would be feasible for IEEE-USA to hold baseline salary estimates in these models at peak levels, rather than allowing values to drop once they have peaked for the most experienced personnel. To date we have opted not to do this, so that users can inspect the effects of reductions in pay for very experienced groups of people. Downsizing, spin-offs of "permanent" employees into temporary or contracted personnel, or other similar practices may have actually forced scales downward for very senior engineers. The existing approach allows analysts to try either strategy of comparison, to see what these effects may be.

Return To The Top Of The FAQs Page