Imagine needing to scale a workforce practically overnight, while guaranteeing that every new hire can perform under intense pressure. That was the reality for wartime American industry in 1939. Ability tests were rapidly adopted by the military and manufacturers to assess candidates, yet decision-makers lacked a consistent framework to prove that these tests were actually delivering a high rate of capable workers. It was in this historical context that H.C. Taylor and J.T. Russell published a method allowing organizations to accurately forecast the hit rate of their selection assessments[1].
The underlying model, now known as the Taylor-Russell model, is still used today to demonstrate the practical effectiveness of selection tools.
Why Every Selection Tool Requires a Compromise
Consider how people make decisions under uncertainty. Whether a physician is interpreting a diagnostic test, a quality-control engineer inspecting manufactured parts, or a hiring manager reviewing candidate assessments, they must act on imperfect information. Most measurement tools, be it a blood test, a pressure gauge, or a job interview, produce results that can be wrong.
Unless an instrument is perfectly accurate (and none ever is), these errors take two distinct forms. A false positive (Type I error) occurs when a measure detects a signal that does not exist, such as a positive pregnancy test for someone who is not pregnant. A false negative (Type II error) occurs when a measure fails to detect a signal that is present, such as antivirus software missing a computer virus.
As Signal Detection Theory[2] formalized, this inherent uncertainty means every selection system is ultimately an exercise in risk allocation. For any given diagnostic tool in hiring, keep this in mind:
While you cannot escape the dilemma for a specific selection method, you can choose which error your organization is more willing to absorb. Optimizing for sensitivity (true positive rate) means lowering the threshold to ensure you catch every qualified candidate. Prioritizing specificity (true negative rate) means raising the threshold to maximize the certainty that selected candidates are suitable.
Forecasting Accuracy with the Taylor-Russell Framework
The Taylor-Russell model was designed to help organizations evaluate the quality of their selection tools. Using just a few key quantities, it estimates probabilities for making correct and incorrect hiring decisions, provided one accepts the foundational assumption that employee performance can be cleanly split into “successful” and “unsuccessful” hires. We’ll return to this assumption later.
To apply the Taylor-Russell model, you need the following information, or at least a reasonable estimate for each:
| Parameter | Explanation |
|---|---|
| Base Rate | The proportion of people in your applicant pool who would perform well on the job if you hired them all without using any selection tool. |
| Selection Ratio | How selective you are, i.e., the number of people you intend to hire divided by the total number of applicants. |
| Criterion Validity | The accuracy of your selection tool, i.e., the correlation between the instrument and actual job performance. |
In practice, these parameters are not always known with sufficient certainty. Base rates are seldom measured directly and typically rely on an “educated guess” based on experience. Similarly, unless your organization has conducted a local validation study or leverages published validity data from a vendor’s technical manual, criterion validity must be estimated using benchmark ranges from the scientific literature.
How Typical Instruments Compare in Practice
If you lack concrete evidence for the accuracy of your own instrument, your second-best option is to rely on benchmark data. Industrial and organizational psychology has studied the predictive power of common hiring tools for nearly a century. The table below summarizes the latest meta-analytic findings. It shows the criterion-related validity you can expect from each method.
These figures represent operational validities under real-world conditions, i.e., the strength of the relationship between test scores and job performance. In our consulting experience, many organizations still do not utilize more accurate selection methods and overemphasize the importance of marginal predictors such as unstructured interviews and job experience.
Note that the table shows the simple, isolated effect of each selection method. Combining multiple hiring tools often yields incremental gains. For instance, an evidence-based, structured employment interview on its own averages a validity of .42. Pair it with cognitive ability, biodata, and integrity testing, and studies show the combined predictive validity averages .61[5].
Try an Interactive Example
Use the following app to build an intuition for the Taylor-Russell model using simulated data. Every data point represents an applicant in the hiring process. The horizontal axis indicates an applicant’s score on a given selection tool (e.g., performance in a job interview). The vertical axis indicates how well that applicant would perform on the job if hired (e.g., supervisor’s rating months after being hired).
Manipulate the three model parameters to see how they influence the overall hiring accuracy (i.e., proportion of correct decisions).
Several insights from this simulation are straightforward: adjusting the base rate (i.e., the proportion of qualified candidates in the applicant pool) shifts the horizontal boundary separating successful from unsuccessful hires. Naturally, selecting hiring tools with higher validity remains the most effective way to maximize accurate hiring decisions.
Nevertheless, as discussed in the next section, the Taylor-Russell model provides a valuable framework for reflecting on and optimizing your organization’s hiring strategy.
How the Taylor-Russell Model Informs Selection Strategy
The key assumption the Taylor-Russell model operates on is that job performance can be meaningfully split into a strict binary outcome (i.e., an employee is either satisfactory or unsatisfactory). In many occupational roles, this assumption is a grave simplification. However, it holds to varying degrees. Organizations that hire for roles with fixed threshold decisions[6] (i.e., all applicants who meet a certain criterion get hired) and that use hard performance goals (e.g., sales revenue) usually benefit more from this framework.
Let’s look at a few cases when the Taylor-Russell model can be useful.
Calculate the Return on Investment of Better Hiring Tools
Especially for growing companies, deciding how much to invest in evidence-based selection tools can be challenging. Here, the Taylor-Russell model provides a framework for a quick, back-of-the-envelope ROI calculation. All you need to know is the cost of a bad hire, the price difference between your current process and a more sophisticated assessment tool, and the incremental validity that new tool provides. In most cases, avoiding even a handful of mis-hires means evidence-based assessments easily pay for themselves many times over. Use our assessment utility calculator to estimate the net financial return for your organization.
Assessment Utility Calculator Enter parameters for your specific hiring situation and use the Taylor-Russell to calculate the return on investment of your hiring tools. Open the calculatorWeigh Error Trade-Offs Based on Role-Specific Risks
In high-consequence roles (e.g., an airline pilot), a single mis-hire (false positive) carries catastrophic financial, operational, or safety costs, making high specificity and strict cutoffs paramount. Conversely, in specialized fields where the war for talent is fierce, wrongfully rejecting a great candidate (false negative) means losing rare talent directly to a competitor. In these cases, optimizing for sensitivity with more forgiving thresholds ensures you do not pass on top performers. By comparing the predicted share of successful hires across different selection ratios, you can see how much success rate a stricter cutoff buys you and how many qualified applicants it costs.
Compensate for Unfavorable Base Rates via Sourcing Strategy
If the baseline quality of your applicant pool is low (e.g., a low base rate where only 20% of applicants naturally possess the necessary skills), relying solely on selection tools puts immense pressure on your cutoff scores, leading to high rejection rates and potential candidate pipeline bottlenecks. Your strategy should split focus between selection accuracy and targeted sourcing/employer branding to naturally elevate the initial candidate pool quality before screening even begins.
Limitations of the Taylor-Russell Model
- The model assumes that every candidate you select completes the process and accepts your offer, so it overstates the yield of long, multi-stage funnels in which strong candidates drop out or accept a competitor’s offer along the way[7].
- Splitting performance into “successful” and “unsuccessful” treats a hire who barely clears the bar the same as a top performer, so the model undervalues tools that are especially good at identifying exceptional talent[8].
- Where you draw the line between success and failure is a judgment call, and moving that line changes the base rate and, with it, every forecast the model makes.
- The forecasts assume a straight-line relationship between test scores and job performance, so treat them with caution if performance levels off beyond a certain score (e.g., when a minimum level of a skill is all the job requires).
Conclusion
The Taylor-Russell model turns three quantities you can estimate today (base rate, selection ratio, and criterion validity) into a concrete forecast of how many of your hires will succeed. It makes the trade-off between bad hires and missed talent explicit and shows that the validity of your selection tools is the strongest lever you control. Treat its output as a well-grounded estimate rather than a guarantee, and use it to compare selection strategies.
Takeaways
- Every selection tool makes errors, so decide deliberately whether your organization can better afford a bad hire or a missed great candidate.
- Optimize for sensitivity or specificity based on your talent target group: favor strict cutoffs in high-consequence roles and more forgiving ones where qualified talent is scarce.
- Invest in validity first, replacing marginal predictors such as unstructured interviews with structured interviews and cognitive ability tests, and combine tools for incremental gains.
- If you lack local validation data, benchmark your instruments against meta-analytic estimates.
- Treat a low base rate as a sourcing problem as much as a selection problem.
- Use the model for quick ROI estimates and scenario comparisons (e.g., with our assessment utility calculator), and report ranges whenever your inputs are uncertain.
