Natural Hazards and Job Choice: How do reasons for final job choices align with job choice characteristics?

Do those who said crime risk mattered in their final job choice take jobs in lower crime risk areas?

To assess whether individuals who prioritized crime risk in their final job choice actually selected jobs in lower-crime areas. Those who said crime was important were more likely to choose jobs in objectively lower-crime areas.

OLS Regression Summary

To evaluate whether participants accepted more or less crime risk in their final job relative to their stated values, we computed a crime mismatch score by subtracting each individual’s stated importance of crime from the actual crime risk level of their chosen job location. Positive scores indicated a mismatch toward higher crime risk than individuals claimed to prioritize, while negative scores reflected choices aligned with lower crime exposure. The average mismatch was -0.39 (SD = 1.86), suggesting that participants tended to choose jobs in slightly lower-crime areas than they reported prioritizing. The interquartile range spanned from -2 to 1, and the distribution included both under- and over-alignment with stated values. These findings suggest that while crime is an important stated factor in job choice, some individuals may accept higher crime risk than preferred, while others choose more cautiously than their stated importance would imply.

Do those who said income mattered in their final job choice take jobs with higher salaries?

We tested whether individuals who rated income as an important reason for choosing their final job actually selected jobs with higher salaries. People who said income mattered when choosing a job did, in fact, choose higher-paying jobs — and the effect is both statistically and practically significant. Each 1-point increase on the importance scale corresponded to a ~$13,940 increase in salary.

OLS Regression Summary

To examine whether individuals who emphasized income as important ultimately selected lower-paying positions, we created an income mismatch score by comparing each respondent’s stated importance of income (0–4 scale) to the percentile rank of their chosen job’s salary. The mean mismatch was -2.33 (SD = 1.36), indicating that many participants selected lower-paying jobs even when they said income was important. Notably, salaries increased across decision stages: while the average salary of all job offers was approximately $73,600, jobs that were chosen in any round averaged $82,100, and the final job selections averaged $101,600. This suggests a general preference for higher-paying jobs overall, but still a gap between stated priorities and actual choices, particularly for those who rated income highly. Other considerations may have overridden salary concerns in final decisions.

Do those who said hazard risk mattered in their final job choice take jobs in areas with lower hazard risk?

We assessed whether individuals who prioritized hazard risk in their job decision actually selected jobs with lower objective hazard risk. The model was not statistically significant with a very low R² of 0.007, indicating that only 0.7% of the variance in job hazard risk was explained by self-reported concern. While negative, suggesting a slight trend toward lower-risk jobs among those who cared about hazard risk, the effect was not statistically reliable.

OLS Regression Summary

Even among those who said hazard risk mattered, people often chose jobs in moderate- to high-risk areas. The highest average risk was for severe weather (4.2/5), suggesting that people may accept severe weather risk even if they say it’s an important consideration in their choice of job, or they may lack realistic alternatives in low-risk regions. Volcano risk was the lowest on average (0.25), likely due to its rarity — not necessarily active avoidance. For many hazards, chosen job locations for those who said hazard risk mattered were just as risky or more so than those of the general population. This reinforces earlier regression findings — stated concern does not always drive behavior, possibly due to competing priorities that override hazard avoidance.

The following table of t-values and p-values for each hazard type helps quantify whether the differences in average risk (between people who said hazard mattered vs. all others) are statistically significant.

Difference in Average Hazard Risk: Hazard-Motivated vs. Full Sample

The following set of t-tests compares hazard risk levels at the location of the final job between people who have experienced high vs. low severity impacts from hazards by hazard type. It helps answer the question “Do prior impacts from a hazard influence whether people avoid that hazard when choosing their final job?” Only wildfire hazards showed a significant relationship: people with prior experience with wildfire impacts were more likely to take jobs in high fire risk areas. This might suggest normalization, resilience, or simply lack of avoidance despite personal experience. For all other hazards, past experience did not appear to influence final job risk exposure.

Difference in Average Job Hazard Risk by Hazard Impact Experience

The following t-test results directly assess whether hazard concern in one’s current location influenced actual hazard exposure in final job locations. People who expressed greater concern about earthquakes or wildfire actually ended up in more hazardous jobs for those specific threats. This suggests either a disconnect between attitudes and behavior, or a possible desensitization or lack of viable alternatives in safer locations. Other hazards showed no significant difference, indicating that concern levels do not reliably predict actual avoidance.

Difference in Average Job Hazard Risk by Hazard Concern

We also examined if experiencing a hazard made people less concerned about it. Thee results suggested that greater hazard experience (impact severity) was significantly and positively associated with greater concern. There was no evidence of risk desensitization; in fact, the lived experience amplified concern — especially for wildfire, floods, and severe weather, where both R² values and slopes were high.

Difference in Impact Severity by Hazard Concern

While experience strongly increases concern, concern does not lead to meaningful avoidance of hazard-risk in final job choices. Even with specific hazards, there is no strong or consistent evidence that people who are more concerned about a specific hazard or have experienced more severe impacts avoid that hazard in job selection. This suggests a disconnect between attitudes (concern) and behavior (job choice) when it comes to hazard risk.  

OLS Regression Summary

Predicting Job Hazard Risk: Concern vs. Impact

We also explored whether concern only mattered when impact severity from hazards was high (i.e., including the interaction term between concern and impact severity). There was no significant interaction between concern and impact severity. Individuals who were both highly concerned and experienced high severity impacts from hazards were not significantly more likely to avoid high-risk jobs than others. So far, the evidence continues to show that attitudes (like concern) and experiences (like past impact) do not meaningfully translate into hazard-avoiding behavior when choosing jobs.

OLS Regression Summary

We used a two-step regression to test whether attitudes (concern) influenced behavior (hazard risk in final job) indirectly via stated reasoning (hazard as a reason for choosing the job). While individuals who were more concerned about hazards were more likely to claim that hazard risk mattered in their job decisions, these concerns and stated motivations did not translate into actual behavior. There was no evidence that concern, even when channeled through reasoning, led to selecting lower-hazard jobs.

Step 1: Predicting Final Job Hazard Reason from Hazard Concern

Step 2: Predicting Job Hazard Risk from Hazard Reason and Concern

The hazard mismatch metric — defined as the difference between the level of hazard risk in the final job and how much the respondent claimed hazard risk mattered in their job decision — reveals important insights about the alignment between stated preferences and actual behavior. For cumulative hazard risk, on average, respondents selected jobs with hazard risk levels about 2.6 points higher than would be expected based on how much they claimed hazard risk mattered to them. The distribution was skewed positively, with 75% of respondents accepting equal or greater hazard risk than they said they were comfortable with. A small number selected jobs with notably lower risk than expected (minimum = -17), but this was relatively rare.

On an individual hazard level, severe weather and slide risk showed the highest average mismatch scores (1.27 and 1.30, respectively), suggesting that individuals often took jobs in areas with more risk than they reported caring about. Flood risk also had a moderate average mismatch (0.81), followed by earthquake risk (0.47). In contrast, fire risk had a slightly negative average mismatch (-0.26), meaning people often avoided areas with higher fire risk. Volcano risk had the largest negative mismatch (-0.95), suggesting a consistent pattern of underexposure compared to stated concern. Note that actual volcano hazard risk was generally low due to the low number of jobs located in areas with volcanoes. These results highlight that job selections were not always aligned with expressed hazard-related values, and that this alignment varied substantially by hazard type. Some hazards—especially those more common or visible (e.g., severe weather)—may be more easily tolerated, while rarer or extreme hazards may be more likely to be avoided, regardless of stated concern.

Do High Salaries Override Concerns About Risk?

In a logistic regression examining job choice, salary emerged as the most powerful predictor of final job selection (p < 0.001), with higher-paying jobs significantly more likely to be chosen. This confirms that compensation plays a dominant role in shaping decision-making.

However, the analysis also found a statistically significant interaction between salary and hazard risk (p = 0.039). While hazard risk on its own was not a significant predictor in this model, the interaction indicates that the influence of hazard risk becomes more pronounced as salary increases. In practical terms, this means that even though individuals are generally more likely to choose higher-paying jobs, some may still avoid high-hazard jobs despite the financial incentive.

Interactions with crime risk and cost of living were not significant, suggesting that salary may more easily compensate for those factors in decision-making.

Taken together, these results show that salary is the primary driver, but hazard risk can still meaningfully influence choice at higher salary levels — implying that concerns about safety persist even in the face of financial rewards.

Logit Regression: Predicting Final Job Choice

To examine the influence of education level, discipline, and cumulative hazard risk on job salary, we conducted an Ordinary Least Squares (OLS) regression analysis. The model included the following predictors: education level, discipline (geoscientists or not), and cumulative hazard risk, along with interaction terms between education level and hazard risk, and discipline and hazard risk. The dependent variable was the salary for the final job choice.

Higher education levels tended to be associated with higher salaries, although the effect was marginally significant. Whether or not the participant was a geoscientist did not have a clear effect on salary after controlling for other variables. There was a marginal positive relationship between higher hazard risk and higher salary; however, this result was marginally significant. The interaction between education level and job hazard risk was not statistically significant, nor was the interaction between discipline and cumulative hazard risk. The three-way interaction term between education level, discipline, and cumulative hazard risk was also not significant.

The results of the regression analysis indicate that education level has a positive relationship with salary, though it is marginally significant. Discipline (i.e., whether the person is a geoscientist) does not appear to significantly influence salary after accounting for other variables. Regarding multicollinearity, VIF values for all predictors were below the threshold of 10, suggesting that there was no significant multicollinearity between the variables. The lack of significant interaction terms implies that the effects of education level, discipline, and hazard risk on salary are independent of each other, and there is no evidence from this model that these factors interact to influence salary in a compounded way.

OLS Regression: Predicting Salary from Education, Geoscience Degree, and Hazard Risk

Variance Inflation Factor (VIF) Table

Understanding salary and cumulative hazard risk in final job choice

The analysis of salary and cumulative hazard risk in relation to final job choice reveals several key insights. Based on the figures and percentiles, we observe that the relationship between salary and cumulative hazard risk is generally non-linear, with salary levels increasing as cumulative hazard risk increases, particularly in the higher quantiles of both salary and hazard risk.

The density plot demonstrate that for individuals making a final job choice, the salary typically increases with cumulative hazard risk. The plot shows the concentration of final job choices near the salary range of $50K to $100K and a cumulative hazard risk range of 10 to 16. The density of job choices aligns within these salary and hazard risk boundaries, reinforcing the idea of a compensatory trade-off between higher hazard risk and salary. This indicates that, while hazard risk levels rise, individuals are likely to accept jobs offering higher salaries to compensate for the increased risk associated with certain job locations.

The decision tree identifies a threshold for salary and hazard risk that influences job acceptance. Notably, job salary plays a key role in influencing whether a job is chosen, especially when salaries fall within certain thresholds. The tree’s structure suggests that individuals are more likely to accept higher salary offers when the cumulative hazard risk is relatively moderate (between 10 to 16), and the salary surpasses thresholds such as $82,500.

The quantile values for job offers, job choices, and final job choices suggest that there is a clear increase in both salary and cumulative hazard risk as we move from lower to higher percentiles. For salary, the 10th percentile offers a salary of about $46,600, while the 90th percentile offers $100,620 for job offers. For final job choices, the salary increases further, with the 90th percentile at $182,200. Similarly, the cumulative hazard risk increases across percentiles, with the 10th percentile at 10 for job offers and the 90th percentile at 19 for job offers. For final job choices, the cumulative hazard risk is slightly lower at the higher percentiles.

The analysis also suggests that individuals with a higher education level and those in more hazardous locations tend to accept jobs at a certain salary threshold. This threshold is found around $50K to $100K salary for lower-risk areas and $150K to $200K for high-risk areas. This salary range appears to be a critical factor for individuals making a final job choice in high-risk locations, with job hazard risk levels influencing this decision.

In conclusion, the data suggests that salary is a key factor in job choice decisions, particularly when paired with higher hazard risks. Higher salaries are generally required to incentivize individuals to accept jobs with higher cumulative hazard risks, and the relationship between salary and hazard risk may follow a compensatory pattern where higher salaries offset greater perceived risks.

The decision tree analysis reveals that salary is the primary driver of final job acceptance, especially in the context of cumulative hazard risk. A clear threshold at $82,650 emerges, and below this value, individuals are generally unwilling to accept jobs with higher hazard risk. Above this threshold, higher hazard risks are more acceptable, especially when salaries exceed $100,000, indicating a strong compensatory effect of salary.

The quantile analysis supports this finding with median salary increases from $67,300 (offers) to $86,300 (final choices), while hazard risk slightly increases at each step. At the 90th percentile, final job salaries reach $182,200, with hazard risk remaining elevated, demonstrating that individuals are willing to trade risk for higher pay.

The Gini Index in the decision tree quantifies the purity of splits based on salary and hazard risk, confirming that most job refusals occur in the lower salary–higher risk segment.

Together, the findings show two behavioral clusters, one favoring low-pay, low-risk jobs, and the other preferring high-pay, high-risk roles, provided the salary exceeds a minimum compensation threshold.

Salary and Cumulative Hazard Risk by Quantile

Continue reading

• About the game
• Methodology
• About the participants
• Hazard exposure and concern
• Factors influencing job choice and relocation
• Reasons for current location, occupation, and final job choice
• Alignment of attitudes and actions
• Predicting job choice outcomes
• Conclusions