Estimating Percent Effort

Śaunak SenŚaunak Sen is professor and chief of biostatistics in the University of Tennessee Health Science Center Department of Preventive Medicine. His research interests are in statistical genetics, analysis of high-throughput data, and statistical computing.

Many statisticians in American academic institutions have “soft money” appointments funded by a mutable portfolio of research grants. A fundamental aspect of holding such a position is proposing and negotiating “percent effort”—a proxy for money without being called so.

Misunderstandings and disagreement regarding percent effort likely underlie much interpersonal conflict involving soft money statisticians. Negotiations regarding percent effort conducted indelicately can degenerate into unpleasantness, harming both the statistician and the funding collaborator. On the other hand, to negotiate effectively and have a productive working relationship, the statistician needs to build trust with the collaborator. A key ingredient is being transparent regarding what percent effort means and how to come up with a number that appears fair to both parties.

I propose a few principles for percent effort estimation so the statistician can justify the number he/she proposes to a collaborator. Note that these are principles, and not formulas. So, what may appear reasonable to one may not appear so to another, and therefore one has to negotiate. I suggest you come up with a number that satisfies the following three principles framed as three questions:

    1. How many projects like this can I handle? Divide 100 percent by that number.

    For example, if you can handle 10 similar projects, then that means the estimate should be 10 percent. This means most collaborative projects should be at least 5 percent effort, preferably at least 10 percent. Very senior personnel may sometimes accept less effort, say 2 percent.

    2. How many hours per year (month or week) would the project consume? Divide that by the FTE (full-time equivalent).

    The biostatistics consulting unit uses an FTE of 1,600 hours per year. This assumes you work 50 weeks a year, 40 hours a week (2,000 hours), and about 20 percent of your time is spent on activities essential to your position but not attributable to any specific project. This would include attending seminars, water cooler talk (not too much), writing letters of recommendation, attending conferences, and reviewing grants and papers.

    The above mentioned 1,600 hours FTE may or may not apply to you, but is a good place to start. Please use your judgment to estimate what is appropriate for your position.

    So, if you think you might spend about three hours a week on a project, that would come out to about 10 percent effort.

    3. What is the understanding with your collaborator? Higher percent effort should mean faster response times and dedicated time slots.

    By agreeing to a certain percent effort, the statistician and collaborator enter into a contract. This is not always clearly spelled out, unfortunately, and can lead to misunderstandings. I am not suggesting both parties treat this too formally, but it may be worth discussing mutual expectations. These expectations will also shape the percent effort.

    For example, one might expect faster response times for a project covering 20 percent effort compared to 5 percent effort. A project covering 30 percent effort may ask the statistician to hold in-person office hours once a week.

Case Studies

Minimal Effort
Bob would like Alice to provide statistical advice on a study of smoking and asthma. An experienced analyst in Bob’s lab will perform the analysis under Alice’s direction.

This may be a candidate for 5 percent effort for a mid-level or senior statistician. A junior statistician may consider 10 percent effort.

  • Principle 1. A moderately experienced faculty member would probably be able to handle 10–20 such projects, assuming they don’t all use specialized techniques and most involve judicious use of existing methods.
  • Principle 2. A 5 percent effort means Alice can set aside about an hour and half weekly, or about 20 hours per quarter.
  • Principle 3. Alice would be somewhat available to Bob to answer questions or give feedback on an abstract or manuscript. It would not be unreasonable for Alice to take two weeks to respond to a query.

Advice and Occasional Analysis
Now, if this study involved analysis of longitudinal data of smoking patterns using mixed effects models that Bob’s analyst is not fluent with, then one may expect Alice to request 10 percent effort. The understanding might be that Alice would perform the more complex analyses herself using R or Stata.

  • Principle 1. If Alice is fluent in analysis of longitudinal data, she may reasonably handle 10 such projects (which require her to pick up the analysis)
  • Principle 2. Alice would devote about 40 hours per quarter (five days), or a little under two days a month. Although predicting how long data analysis would take is difficult, this may be reasonable. Many data analyses go through multiple rounds of revision, and so the amount of time devoted may be 10 or 20 times that of the final analysis
  • Principle 3. Alice would be moderately available. One might expect her to respond faster than in the previous case or be available for a quick phone call in a week.

Specialized Analysis
If the study has genomewide genetic data and intends to perform a genomewide association analysis, then Alice may request Bob to consider a greater intensity of involvement—say 10–20 percent, depending on complexity and expectations from Alice.

Alice might agree to 15 percent effort with the understanding that she will follow the statistical literature closely and, if needed, implement customized solutions to complete the statistical analysis. It seems reasonable she could handle between 5–10 such projects (Principle 1), be available about three days a month for the project (Principle 2), and be expected to respond to queries within about a week.

Leading a Statistical Core
If Alice is leading the statistical core of a center led by Bob with three related projects, then a starting point of discussion might be about 30 percent effort. Can one lead five such cores and remain sane (Principle 1)? Alice would have to lead a small team of programmers and be available to answer questions within a day or two (Principle 3). Alice may have office hours at Bob’s center once a week. This could take between one or two days a week.

Leading Projects
If Alice is leading her own projects, either as an independent principal investigator or as part of Bob’s lab, 20–50 percent effort would be considered reasonable. Less than 20 percent effort may be seen as insufficient commitment; to lead a research project, one should be prepared to spend at least one day a week on average. Junior investigators would lean toward the higher end of the scale, while senior investigators would likely be at the lower end.


Statisticians on soft money appointments are supported by multiple projects from different funding sources. I think the three principles laid out would help transparently negotiate percent effort. By agreeing to ground rules, statisticians can reduce potential conflict and build trusting long-term relationships with collaborators.

There is considerable flexibility in how the principles may be interpreted. For example, the FTE estimate is dependent on the individual statistician’s circumstances. Both parties may use the process of negotiating percent effort to also negotiate mutual expectations.

I also recognize that estimating percent effort before working on a project amounts to extrapolation with attendant risks. If anything, many of us have a tendency to overestimate what we can accomplish in a given amount of time. From this perspective, it is helpful to be conservative in the estimates. It is often helpful to start out for a specified period (say six months or a year) and then adjust effort depending on realized complexity.

Finally, while I have considered statisticians in soft money environments, the essential principles are applicable to nonstatisticians and academics in “hard money” environments with appropriate modifications. Hard money appointees are not usually required to adhere strictly to how they are paid, but that might change as universities cut back on guaranteed funding for faculty.

Questions or comments? Follow me on Twitter at @saunaksen or visit my blog.