§ Methodology
The research behind Should I Quit?
This tool integrates three validated psychometric instruments with Monte Carlo simulation of career and financial trajectories. Everything runs in your browser. Below is a full accounting of the instruments, scoring, stochastic models, and limitations.
The three psychometric instruments
Copenhagen Burnout Inventory (CBI). Developed by Kristensen, Borritz, Villadsen, and Christensen (2005) at the Danish National Institute of Occupational Health. Validated in the PUMA study (n = 1,914 at baseline) and subsequently across more than 40 countries. This tool uses the Personal Burnout (6 items) and Work-related Burnout (7 items) subscales — 13 items total. Response scales are frequency (Always / Often / Sometimes / Seldom / Never) for items 1-10 and intensity (To a very high degree / high / somewhat / low / very low) for items 11-13. Item 10 ("Do you have enough energy for family and friends?") is reverse-coded. Each subscale is scored as the mean of its items on a 0-100 scale.
Work and Meaning Inventory (WAMI). Developed by Steger, Dik, and Duffy (2012) and published in the Journal of Career Assessment. The instrument captures three dimensions of meaningful work through 10 items on a 5-point Likert agreement scale (rescaled to 0–100 for dashboard consistency). Subscales per Steger's authoritative scoring sheet: Positive Meaning (items 1, 4, 5, 8 — 4 items), Meaning Making through Work (items 2, 7, 9 — 3 items), and Greater Good Motivations (items 3, 6, 10 — 3 items). Item 3 (“My work really makes no difference to the world”) is reverse-coded, marked with (R) in the interface. Cronbach's α = 0.93 for the total scale in the original validation sample (n = 370). Earlier versions of this tool used a 6-item short form; the current version restores the full validated 10-item scale.
Turnover Intention. Three items adapted from the measurement tradition established by Mobley (1977) in the Journal of Applied Psychology, which identified quit cognitions as the most proximal — and strongest — antecedent of actual turnover. This measurement approach was empirically validated in the Griffeth, Hom, and Gaertner (2000) meta-analysis in the Journal of Management, which established turnover intention as having the single largest effect size among all measured predictors of actual turnover. The items are public-domain and used here with standard five-point agreement scaling.
The Monte Carlo engine
For each of 10,000 iterations and each path (stay vs quit), the engine simulates a full year-over-year trajectory across burnout, meaning, and compensation. The v2 rebuild uses user-driven parameters wherever possible, replacing the population-average defaults of earlier versions. Where parameters cannot be elicited from the user, the tool uses literature-grounded shape (e.g., the qualitative pattern of the honeymoon curve) combined with documented author choices for specific numerical values.
| Parameter | Distribution / formula | Source |
|---|
| Annual salary growth (stay) |
Normal(μ = user input C5, σ = 2%) |
User-driven mean; σ is an author choice reflecting that known-job raise variance is typically small |
| Annual salary growth (new job, year 2+) |
Normal(μ = user input C6, σ = 3%) |
User-driven mean; σ is an author choice higher than stay path because new-employer trajectories are less predictable |
| Job-search duration (easy) |
LogNormal(median = 2mo, σ = 0.4) |
Median is an author choice informed by BLS CPS unemployment-duration medians |
| Job-search duration (moderate) |
LogNormal(median = 4mo, σ = 0.5) |
Author choice, BLS-informed |
| Job-search duration (difficult) |
LogNormal(median = 8mo, σ = 0.6) |
Author choice, BLS-informed |
| New-job starting salary |
Triangular(low, mid, high) |
All three points from user input C2 |
| Stay-path burnout trajectory |
by+1 = by + 0.05 × (50 − by) + Normal(0, 4) |
Mild regression-to-mean (0.05/yr) plus noise; v1.1's “user-selects-trajectory” parameterization removed because current CBI score already encodes present trajectory |
| Stay-path meaning trajectory |
my+1 = my + 0.05 × (50 − my) + Normal(0, 4) |
Same form as burnout; gentle regression toward population mean |
| Quit-path burnout honeymoon |
by = b0 + (−drop) × exp(−y / τ) + Normal(0, 3) |
Drop magnitude from user input C7; τ = 1.44 is author choice: at this value, 50% of the user's expected relief remains at year 1, ~12% at year 3 — matches Boswell, Boudreau & Tichy (2005) qualitative hangover-over-years pattern |
| Quit-path meaning honeymoon |
my = m0 + boost × exp(−y / τ) + Normal(0, 3) |
Boost magnitude from user input C8; same τ as burnout |
Note on trajectories decaying toward the user's pre-quit baseline: the v2 honeymoon model assumes that the positive effects of a new role fade back toward the person's existing baseline rather than settling at an idealized “new baseline.” This is more conservative than the v1.1 model, which assumed a permanent baseline improvement on the quit path. If your new role produces a sustained improvement beyond the honeymoon period, the simulation will under-estimate the quit path. The trade-off: assuming permanent improvement is not a claim the tool can support without information about the specific new role, which the user doesn't yet have when running this simulation.
Composite Life-Quality Index
The composite LQI collapses three axes into a single 0–100 number: LQI = (1/3) × (100 − burnout) + (1/3) × meaning + (1/3) × financial-stress-score. Equal weights are the simplest defensible choice — no literature consensus exists on work-specific weights (Diener, Ryff, and VanderWeele's flourishing frameworks all use equal-weighting conventions, or explicitly avoid prescribing weights). Earlier versions of this tool used 40/40/20 weighting; the equal-weighting is documented as an author choice rather than a literature-derived claim.
Financial-stress-score uses a path-comparable formula: both the stay and quit paths are scored against the same baseline of “cumulative compensation needed to cover monthly expenses over the horizon” (user input C4 × 12 × years). Score = 100 when cumulative covers the full need; score = 0 when it covers ≤40% of need; linear interpolation in between. This replaces the v1.1 approach which compared each quit-path simulation against its paired stay-path simulation, producing a circular self-comparison that always yielded 50 for the stay path.
The archetype matching logic
The six archetypes are matched in order of specificity on four signals: burnout composite, meaning composite, turnover intent, and runway months. Thresholds use the published CBI band cutpoints (Creedy et al. 2017; also used in Borritz et al. 2006 and widely across the subsequent CBI literature): burnout <50 is low, 50–74 is moderate, 75–99 is high, and 100 is severe. Ordering matters because multiple rules can match — the first match wins, so stricter archetypes (Committed, which requires low burnout AND meaningful work AND low turnover) are tested before broader ones (Bridge-Builder, which requires moderate burnout with low turnover intent). When a user has not entered their runway, the tool surfaces a Runway Unknown prompt for severe-burnout cases rather than silently defaulting to zero savings (which was a misclassification bug in v1.1).
The regret minimization lens
Jeff Bezos popularized the regret-minimization framework as a personal decision heuristic, but the empirical foundation is Gilovich and Medvec's 1995 Psychological Review paper. Their core finding: in the short term (within a year), people regret actions they took; in the long term (5+ years), they regret inactions they didn't take. The regret lens in this tool complements the Monte Carlo because it captures emotional projections that financial and burnout trajectories can't — identity, relationships, the counterfactual life. When the two lenses disagree, the disagreement itself is useful signal.
Limitations
The tool cannot model what matters most about your specific decision. Family dynamics, the specifics of the role you'd move to, industry volatility, your health, your identity — none of this enters the simulation. What it provides is a structured outside view: a way to see your situation as one instance in a class of similar situations, with distributions of plausible outcomes. Self-report is also subject to mood, recent events, and survey fatigue — your results are most useful when you run the tool a second time after a few weeks have passed.
Documented author choices. Three numerical parameters in the simulation are not directly user-driven and cannot be derived from published literature. We document them here rather than hide them:
- Compensation variance (σ = 2% stay, σ = 3% quit) around the user's stated mean raise. Values chosen to reflect that known-job raises are more predictable than new-employer raise trajectories. No specific citation supports these exact values; they are conservative choices that introduce visible variance in violin plots without dominating the user's mean input.
- Honeymoon decay rate τ = 1.44 years. This value produces a curve where ~50% of the user's expected honeymoon remains at year 1 and ~3% remains by year 5. The qualitative shape (peak early, decay over 1–3 years toward baseline) is grounded in Boswell, Boudreau & Tichy (2005); the specific τ value is an author choice chosen to match that qualitative timeline. Different users may experience faster or slower decay depending on their transition circumstances.
- Financial stress mapping maps cumulative-to-needed ratio linearly: ratio ≥ 1.0 produces score 100, ratio ≤ 0.4 produces score 0, linear interpolation in between. The specific 0.4 floor is an author choice with no direct literature support; it represents a forgiving threshold where a user earning 40% of needed compensation receives the minimum score rather than being considered entirely without financial buffer. Users whose subjective sense of financial stress rises at higher coverage ratios may find this model too lenient.
The honeymoon model is symmetric and may be optimistic or pessimistic for any given user. The simulation assumes the user's expected burnout relief and meaning boost both decay toward their pre-quit baselines. Users whose new role produces a sustained improvement beyond the honeymoon period will see the quit path under-represented. Users whose expectations are more aspirational than realistic will see the quit path over-represented in the early years.
Confidence degrades with horizon length. 3-year projections are tightly anchored to your current scores. 5-year projections compound modest uncertainty. 10-year projections are interpretive — compound uncertainty across burnout trajectory, meaning trajectory, compensation variance, and life events that cannot be modeled here makes them useful as qualitative thought experiments rather than point estimates.
This is not a clinical instrument. The Copenhagen Burnout Inventory is a research measure, not a diagnostic tool. If your burnout scores are severe, please speak to a qualified professional before making major life decisions based on this simulation alone.