Should You Quit Your Job? A Scientific Decision Model for Resignation

1. Personal Utility Function: Quantifying Job Satisfaction

1.1 Multidimensional Utility Model

Employee total utility is expressed as a weighted combination of multiple dimensions: U = \sum_i w_i \cdot u_i(x_i), where w_i is the weight of dimension i, u_i is the utility function for that dimension, and x_i is the actual value of the dimension.

Common utility dimensions include salary, career development, work atmosphere, work‑life balance, job stability, and others.

1.2 Diminishing Marginal Utility of Salary

Economic theory shows that salary utility exhibits diminishing marginal returns. Using a constant relative risk aversion (CRRA) utility function: U(s) = (s / b)^{1-\gamma} / (1-\gamma) where s is salary, b is basic living cost, and \gamma is the risk‑aversion coefficient (typically 0.5‑2).

When salary doubles from 5,000 RMB to 10,000 RMB, the satisfaction increase is the same as from 10,000 RMB to 20,000 RMB, but the absolute increment is larger at low income, illustrating diminishing marginal utility.

Example calculation: Assume a base salary of 5,000 RMB.

Salary 5,000 RMB: utility baseline.

Salary 10,000 RMB: increment ≈ 0.693.

Salary 15,000 RMB: increment ≈ 0.405.

The diminishing marginal utility is evident.

1.3 Time Discounting of Career Development Utility

Career development value must be discounted over time. Let the annual time‑preference rate be r and the discount factor be δ = (1 + r)^{-t}. The discounted benefit of future growth G_t is δ·G_t.

Parameter guide:

Younger workers: emphasize long‑term development.

Mid‑career workers: balance present and future.

Near‑retirement: prioritize present benefits.

2. Cost‑Benefit Model of Resignation

2.1 Total Cost Function

From the employee perspective, total resignation cost includes search cost, conversion cost, and risk cost.

Search cost:

Search time (months), usually 1‑6 months.

Current monthly salary.

Income loss proportion during search (≈0.1‑0.2 while employed, 1.0 after resignation).

Effort and opportunity cost (≈10‑30 % of monthly salary).

Conversion cost:

Relocation expenses and deposit (≈5‑20 k RMB if moving cities).

Learning cost: 3‑6 months to regain normal productivity; productivity loss per month is λ_t.

Social cost: loss of existing network, estimated 5‑15 % of annual salary.

Risk cost (expected loss):

Probability of trial‑period failure.

Probability of high match with new job.

Loss from trial‑period failure (re‑search time and income loss, ≈3‑6 months salary).

Loss from mismatch (continuous low satisfaction).

2.2 Expected Benefit Model

The expected benefit is the present value of the utility difference between the new and old jobs, split into monetary and non‑monetary parts.

Considering uncertainty: New salary is treated as a random variable (e.g., log‑normal distribution) with probability density f(s).

2.3 Decision Criterion

Rational employees should resign when expected benefit exceeds total cost. Risk aversion and decision inertia are incorporated via a threshold θ: Benefit - Cost > θ Certainty Equivalent (CE): For risk‑averse individuals, replace expected benefit with its certainty equivalent.

3. Optimal Resignation Timing Model

Resignation timing is a dynamic optimization problem. Define V(t, s) as the optimal value function at time t and state s (salary, position, skill, market conditions). The Bellman equation compares staying versus leaving:

V(t,s) = max\{ \text{Stay}(t,s), \text{Leave}(t,s) \}

4. Career Path Optimization Model

4.1 Markov Decision Process (MDP)

View a career as a sequence of states; each resignation or stay is a decision point.

State space: position, salary, skill level, age.

Decision space: stay, resign.

Transition probability: probability of moving from one state to another given a decision.

Immediate reward: utility of staying versus utility of leaving at the current moment.

Goal: maximize the discounted sum of lifetime utility.

Bellman equation:

V(s) = max\{ R_{stay}(s) + \gamma \sum_{s'} P(s'|s,\text{stay}) V(s'), \; R_{leave}(s) + \gamma \sum_{s'} P(s'|s,\text{leave}) V(s') \}

4.2 Human Capital Accumulation Model

Human capital evolves according to a differential equation: dH/dt = L \cdot \eta(t) - \delta H, where L is learning input (training time, project experience), \eta is learning efficiency, and \delta is depreciation.

Learning efficiency shows diminishing returns (higher skill → harder to learn) and age effects (learning ability declines with age). Discrete‑time approximation shows that a new job with a better learning curve can be optimal even if salaries are similar.

Total value of human capital: includes skill network (social capital) and a conversion factor; changing jobs reduces part of the network capital.

5. Risk Assessment and Uncertainty Model

5.1 Monte Carlo Simulation

Because the future is uncertain, Monte Carlo simulation is used to generate many scenarios.

Algorithm steps:

Define key random variables and their distributions (new salary, promotion probability, macro‑economic factors).

Sample each variable to create a scenario.

Compute the net present value (NPV) for that scenario.

Collect statistics:

Expected benefit.

Standard deviation (risk).

Probability of successful resignation.

5 % quantile (worst‑case outcome).

5.2 Risk‑Adjusted Return

For risk‑averse decision‑makers, use certainty‑equivalent return (CE) instead of expected benefit. CE can be computed with absolute risk‑aversion coefficient (CARA) or relative risk‑aversion (CRRA) via Monte Carlo integration.

5.3 Sensitivity Analysis

Assess how changes in key parameters affect the decision:

New‑job salary (most sensitive).

Time‑preference rate.

Trial‑period success probability.

Learning efficiency in the new role.

If elasticity > 2, the parameter requires careful evaluation or additional data collection.

5.4 Value of Information

When information is incomplete, compute the Expected Value of Perfect Information (EVPI) to decide how much to invest in acquiring more data (e.g., internal referrals, trial periods).

6. Multi‑Objective Optimization: Work‑Life Balance

6.1 Pareto Optimal Analysis

Identify Pareto‑optimal options where no other alternative improves any dimension without worsening another.

Practical steps:

Plot current and potential jobs in a multidimensional space.

Check whether the new job dominates the current one (no worse dimension, at least one better).

If Pareto‑improved, consider resignation.

If trade‑offs exist, assign personal weights to each dimension.

6.2 Work‑Life Balance Index

Construct a composite index:

WLB = w_1·T_{share} + w_2·Q_{life} + w_3·E_{time} + w_4·S_{stress} + w_5·α_{sensitivity}

where T_{share} is personal discretionary time share, Q_{life} is subjective life‑quality score, E_{time} is elasticity of time value, S_{stress} is work‑stress index, and α_{sensitivity} reflects individual sensitivity to stress.

Example shows a new job with higher WLB despite slightly lower salary, leading to a higher overall utility.

7. Behavioral Economics Adjustments

7.1 Prospect Theory Utility

Kahneman and Tversky’s prospect theory states that gains and losses are evaluated asymmetrically. Empirical parameters indicate risk‑aversion for gains (concave) and risk‑seeking for losses (convex), with a loss‑aversion coefficient around 2.25.

Apply to resignation:

Gain: salary increase.

Loss: adaptation cost, loss of familiar environment.

People tend to overestimate resignation risk (loss aversion) and underestimate the opportunity cost of staying.

7.2 Sunk Cost Fallacy

Sunk costs (years worked, emotional investment, specialized skills) should be ignored; rational utility considers only future outcomes.

Common sunk‑cost traps:

"I’ve been here five years; leaving would be a waste."

"The company invested heavily in my training; I must stay."

"My skills are only useful here."

7.3 Status‑Quo Bias

People have a tendency to maintain the current situation even when change is clearly better. This adds an inertia cost that must be overcome by a positive decision threshold.

8. Practical Application: Personal Decision Matrix

8.1 Scoring System

Salary level – weight 0.25 – current score 70, new opportunity 85 – weighted difference +3.75.

Career development – weight 0.30 – current 60, new 90 – weighted difference +9.00.

Work atmosphere – weight 0.20 – current 80, new 70 – weighted difference –2.00.

Work‑life balance – weight 0.15 – current 75, new 60 – weighted difference –2.25.

Job stability – weight 0.10 – current 85, new 65 – weighted difference –2.00.

Weighted total difference: +6.50. A positive value initially supports resignation, but further analysis of costs, annualized benefits, payback period, and total NPV is required.

8.2 Decision Tree Analysis

┌─ Trial success (0.85) ─ High match (0.65) ─ NPV = +500k
              │                         └─ Medium match (0.35) ─ NPV = +200k
Resign ──────┤
              └─ Trial failure (0.15) ─ NPV = –100k

Stay ──────── Normal development (0.70) ─ NPV = 0
              └─ Layoff/Salary cut (0.30) ─ NPV = –200k

Expected net present value of resignation ≈ 328,800 RMB, versus –60,000 RMB for staying; therefore resignation is financially preferable.

8.3 Scenario Analysis

Optimistic (probability 0.25): salary +40 %, excellent development, low adaptation cost → NPV + 600k.

Baseline (probability 0.50): salary +25 %, good development, medium cost → NPV + 300k.

Pessimistic (probability 0.25): salary +15 %, average development, high cost → NPV + 50k.

Even in the pessimistic case the net benefit remains positive, indicating a robust resignation decision.

8.4 Macro‑Environment Considerations

Economic cycle and industry trends affect the success probability of a new job.

Expansion phase: active job market, lower resignation risk.

Normal phase: cautious evaluation.

Recession: raise the resignation threshold or postpone the move.

Industry growth potential (e.g., AI, renewable energy) should be incorporated into the utility of career development.

Conclusion

Mathematical modeling provides a rational scaffold for resignation decisions: it quantifies vague feelings, evaluates multi‑dimensional trade‑offs, identifies key factors through sensitivity analysis, and mitigates cognitive biases via structured analysis.

Nevertheless, models are tools, not masters. Personal values, aspirations, and intuition remain essential. When rational analysis aligns with inner motivation, confidence in the decision grows; when they conflict, further reflection is warranted.

Final advice: Avoid perfectionism; choose the best current option and commit fully to make it successful.