Make the Leap Research· Working Paper MTL-WP-2026-01

Deliberation Without Resolution: Self-Doubt Is More Common Among Longer-Term Career-Change Deliberators

Jon Miksis

Make the Leap (JJ Accelerated Inc.), United States

Working paper, version 1.0 · 25 August 2026 · Not peer reviewed

Keywords: career change; occupational transition; intention–behavior gap; deliberation; self-efficacy; imposter phenomenon; midcareer

1. Introduction

Research on occupational transition has largely concentrated on two well-defined moments: the antecedents of turnover intention, and the mechanics of the transition itself. Comparatively little empirical attention has been paid to the interval between them — the period in which an individual has identified a desire to change occupation but has not yet acted. This interval is difficult to observe. It leaves no administrative trace, it precedes the behaviors that appear in labor-market data, and individuals within it are not yet identified as job changers by any conventional instrument.

The intention–behavior gap is well documented in psychology (Sheeran & Webb, 2016), building on models linking intention to behavior (Ajzen, 1991). Meta-analytic work has established that intentions translate into action far less reliably than intuition suggests, and that the gap is asymmetric: people who intend often fail to act, while people who act almost always intended to. Proposed remedies typically target the implementation step rather than the intention itself (Gollwitzer, 1999). Most of this literature, however, examines short horizons — weeks or months — and behaviors with low switching costs. Occupational change differs on both dimensions. Its horizon can extend for years, and its switching costs include income, accumulated expertise, professional identity (Ibarra, 1999), and social standing — conditions under which inaction is itself a well-documented default (Samuelson & Zeckhauser, 1988).

This paper uses an unusual observation point to describe that interval. Individuals voluntarily completing an online career self-assessment are, by construction, in the deliberative state: they have identified a desire to change and are seeking information, but the assessment itself is not an occupational action. At intake they report how long they have been considering the change, alongside the barriers they perceive, their financial position, their self-rated disposition toward risk, and their history of prior attempts. This yields a large cross-section of the deliberative interval that is otherwise hard to sample.

The paper is descriptive in aim. It asks: how long does deliberation last, and what distinguishes those who have been deliberating longest? The results are presented as associations, and the limitations section states plainly what the design cannot support.

2. Data and methods

2.1 Sample

Data comprise 23,684 completed self-assessments submitted to Make the Leap (a trading name of JJ Accelerated Inc.), a commercial career-assessment service, between March 2026 and 25 August 2026, when the dataset was extracted. Respondents arrive through a mix of organic search, paid social advertising, and direct traffic. The instrument is free, requires no account, and is completed by adults who have sought it out. Records were included if the assessment was completed and the deliberation-length item was answered; all such records in the corpus were included, with no further exclusion. The analytic dataset contains no free-text or identifying fields.

The sample is explicitly self-selected. Respondents are people who were sufficiently motivated by a desire to change occupation to seek out and complete a structured assessment. The population of inference is therefore adults actively considering an occupational change, not the general workforce, and the deliberation distribution reported here should not be read as a population base rate.

2.2 Measures

Deliberation length. A single-select item, “How long have you been thinking about a career change?”, with six ordered options: a few weeks; a few months; 1–2 years; 3–5 years; 5+ years; as long as I can remember. The final category is treated as ordinally highest in trend tests and excluded from any duration arithmetic, as it has no defensible numeric value.

Barriers. A multi-select item asking what has held the respondent back, with eight options (financial risk; lacking the right skills; feeling too old; family obligations; not knowing what else to do; fear of failure; imposter feelings; being too comfortable). Respondents may select any number, so proportions do not sum to 100%.

Financial runway. A single-select item on months of savings, dichotomized for analysis as three months or less versus more.

Self-rated boldness. A single-item 1–10 self-rating of willingness to take a risk on a change. Single-item measures of this kind are coarse and are treated accordingly.

Prior attempts. A single-select item on whether the respondent has attempted a major career change before, with six options ranging from “never seriously considered it” to “changed successfully.” Critically, this item references the respondent’s lifetime history, not behavior during the deliberation period measured by the first item. This limits what it can establish and is discussed in §5.

Experience. A single-select item on total years of professional experience. The response categories changed during the collection period.Earlier respondents saw a single “15+ years” band (n = 4,545); later respondents saw it divided into 15–20, 21–30, and 30+ years. Analyses requiring the finer bands therefore exclude the earlier respondents, which is why the granular-band models below report n = 19,139 (23,684 − 4,545). Every such analysis is repeated with a harmonized five-band variable (0–2, 3–5, 6–10, 11–15, 15+) covering all 23,684 respondents, and both results are reported.

2.3 Analysis

Proportions are reported with Wilson 95% confidence intervals (Wilson, 1927). Trends across the six ordered deliberation categories are tested with the Cochran–Armitage test for trend (Armitage, 1955). Two-group contrasts use two-proportion z-tests; continuous contrasts use Welch’s t-test with Cohen’s d. A multivariable logistic regression models the odds of never having attempted a career change. Given the sample size, statistical significance is easily attained; effect sizes are therefore reported alongside every test and weighted more heavily in interpretation. Analyses were conducted in Python (scipy 1.13.1, statsmodels 0.14.6). No analysis was pre-registered.

3. Results

3.1 Deliberation is long

Table 1 reports the deliberation distribution. A majority — 71.2%, 95% CI [70.6, 71.8] — selected a response of one year or longer. The modal response is 1–2 years (35.6%), 35.6% report three years or more, and 11.6% select “as long as I can remember.” Only 5.1% are within their first few weeks. Because responses are banded, the 1–2 year category includes anyone at exactly twelve months; the figure is therefore “one year or longer,” not “more than a year.”

The distribution is stable over time: splitting the observation period into earlier and later three-month windows (n = 9,872 and n = 13,812) yields no meaningful association (χ² = 8.5, p = 0.132, Cramér’s V = 0.019).

3.2 Barriers intensify and change composition

Table 2 reports each barrier by deliberation length with a Cochran–Armitage trend test. Seven of the eight barriers show significant trends, and the directions are not uniform. The four barriers concerning the self — lacking the right skills, being too old, fear of failure, and imposter feelings — carry the strongest positive trends in the table. Only lacking-the-right-skills rises strictly monotonically across all six bands; the other three fluctuate between adjacent bands while trending clearly upward, which is the pattern the trend test is designed to detect. Financial risk, the most frequently cited barrier overall, shows no trend at all (z = −1.13, p = 0.257). Reported complacency (“too comfortable”) declines.

Financial risk was the only barrier without a significant trend. Directional contrast between the shortest and longest deliberation groups: lacking skills +9.0pp (p = 1.3×10⁻⁸), too old +7.4pp (p = 3.5×10⁻⁶), imposter +6.6pp (p = 1.5×10⁻⁵), fear of failure +4.2pp (p = 8.6×10⁻³), too comfortable −5.6pp (p = 3.0×10⁻⁷).

3.3 Deliberation is not accompanied by financial accumulation

If prolonged deliberation reflected a savings strategy, longer deliberation should be associated with greater financial runway. It is not, in any practically meaningful sense. The proportion reporting three months of savings or less is 74.4% [71.9, 76.8] among those deliberating a few weeks, 67.9% [65.7, 70.0] among those at 5+ years, and 72.8% [71.1, 74.4] among the “as long as I can remember” group. The overall trend test is significant but weak and non-monotonic (z = −2.43, p = 0.015); the largest contrast across the entire deliberation range is roughly six percentage points, against a baseline in which roughly seven in ten respondents report three months of runway or less regardless of how long they have deliberated.

3.4 Longer deliberators do not rate themselves as less bold

Mean self-rated boldness rises slightly across deliberation categories, from 5.94 (SD 1.83) among those deliberating weeks to 6.42 (SD 1.97) among the “always” group (r = 0.059, p = 6.0×10⁻²⁰; ANOVA F = 20.66, p = 1.2×10⁻²⁰). The combined longest group exceeds the shortest by 0.41 points (Welch t = 6.81, p = 1.3×10⁻¹¹, Cohen’s d = 0.216). The effect is small and should not be overstated. Its value is directional: an account in which prolonged deliberation is principally a manifestation of low risk tolerance predicts the opposite sign, and does not survive.

3.5 Experience is associated with longer deliberation

The proportion deliberating three or more years rises with professional experience, from 26.2% [23.8, 28.6] among respondents with 3–5 years of experience to 39.7% [37.7, 41.7] among those with 30 or more (Cochran–Armitage z = 9.99, p < 10⁻¹⁵; n = 19,139, granular bands). Respondents with 0–2 years of experience are an exception at 31.9%, interrupting otherwise monotonic growth.

Because the granular bands exclude the 4,545 respondents who saw the earlier instrument (§2.2), the analysis was repeated with the harmonized five-band variable across all 23,684 respondents. The gradient is unchanged and slightly stronger: 26.2% [23.8, 28.6] at 3–5 years of experience rising to 38.2% [37.4, 38.9] at 15+ years (z = 11.17, p < 10⁻¹⁵).

3.6 Prior attempts do not decline with deliberation

Across the sample, 48.7% have never attempted a major career change and 27.2% report having changed successfully at some point. Among those deliberating three or more years, 45.9% [44.9, 47.0] have never attempted one. Notably, this is lower than the 48.6% among respondents deliberating less than a year (z = 3.25, p = 1.2×10⁻³). In a logistic regression adjusting for experience, financial runway, boldness, fear of failure, and imposter feelings, each step up the deliberation scale is associated with slightly lower odds of never having attempted a change (OR 0.949, 95% CI [0.929, 0.970], p = 1.9×10⁻⁶; n = 19,139, pseudo R² = 0.018). Refitting on all 23,684 respondents with the harmonized experience variable leaves the estimate essentially unchanged (OR 0.951, 95% CI [0.933, 0.969], p = 2.2×10⁻⁷).

This result should be read narrowly. The prior-attempt item measures lifetime history, and respondents who have deliberated longer are on average more experienced and therefore have had more opportunity to attempt a change. The finding does not establish that long deliberators act more; it establishes only that the data provide no support for the intuitive claim that they act less.

4. Discussion

The results describe a deliberative state that is long, stable in the aggregate, and not obviously progressing toward resolution. Two features are worth drawing out.

The first is the composition shift in Table 2. Barriers are not uniformly higher among longer deliberators; they are differently distributed. The external constraint most often cited — financial risk — is invariant to deliberation length, while barriers concerning the respondent’s own adequacy rise substantially. Complacency, meanwhile, falls. One reading consistent with this pattern is that the reasons people give for not moving migrate over time from circumstance toward self-assessment. Whether this reflects a within-person process, a selection process by which those with self-doubt remain in deliberation longer, or a reporting effect, cannot be distinguished here.

The second is the disconfirmation of two intuitive accounts. Prolonged deliberation is often explained either as prudent preparation or as timidity. Section 3.3 finds no meaningful accumulation of financial runway across deliberation length, and §3.4 finds self-rated boldness slightly higher, not lower, among the longest deliberators. Neither account fits well. A third account — that deliberation persists because the decision is genuinely complex and its costs are concentrated among the experienced — is at least consistent with the experience gradient in §3.5, though the present design cannot test it directly.

For practitioners, the practical implication is modest but concrete: if reported barriers do not attenuate over years of deliberation, interventions that supply time are unlikely to be sufficient, and interventions that supply evidence — direct contact with the target occupation, explicit calculation of the required income floor, structured translation of existing experience — are the more plausible candidates. That proposition is testable and is not tested here.

5. Limitations

Self-selection. The sample consists of people who sought out a career assessment. Findings describe adults actively considering occupational change and cannot be generalized to the workforce.

Cross-sectional design. All associations are between-person. No respondent is observed at two points in deliberation, so no within-person change is measured. Statements such as “barriers intensify” describe differences across groups, not documented trajectories, and reverse causation is entirely plausible: individuals reporting greater self-doubt may remain in deliberation longer rather than deliberation producing self-doubt.

Self-report and instrument coarseness. All measures are self-reported at a single sitting. Deliberation length is retrospective and subject to recall error. Boldness is a single-item self-rating. Barriers are selected from a fixed list, which constrains the response space and cannot capture barriers the instrument does not name.

Misalignment of the prior-attempt measure. As noted in §3.6, prior attempts are measured over the lifetime while deliberation is measured for the current episode. The two are not commensurate, and no inference about action during the measured deliberation is available from these data.

Statistical power and effect size. At n > 23,000, trivial differences reach significance. Effect sizes are reported throughout and several significant results (notably §3.4) are small.

Provenance. The data are collected by Make the Leap (a trading name of JJ Accelerated Inc.), a commercial service that sells career products to a subset of respondents, and the author is its founder. The instrument was designed for product purposes rather than for research, and no analysis was pre-registered. The dataset and analysis code are released so that the reported figures can be recomputed independently.

6. Ethics and data governance

Responses were provided by adults voluntarily completing a free online career assessment. No compensation was offered and no identifying information is required to complete it. This analysis was conducted on de-identified data, and the released dataset contains no names, email addresses, account identifiers, free-text responses, or timestamps finer than the month of submission — only the fixed-choice items described in §2.2.

There was no institutional review board review. The instrument is a commercial product feature rather than a research protocol, and Make the Leap is not affiliated with an institution that operates an IRB. Readers should weight the findings accordingly. The analysis was not pre-registered, and the hypotheses described in §1 were formed after the data existed.

7. Conclusion

Among 23,684 adults actively considering a career change, deliberation typically spans years rather than weeks. Across deliberation length, self-reported barriers concerning skills, age, failure, and legitimacy are substantially higher, financial risk is unchanged, complacency is lower, financial runway is not meaningfully greater, and self-rated boldness is marginally higher. Taken together, the results are difficult to reconcile with accounts of extended deliberation as either accumulating preparation or expressing timidity, and are more consistent with a state in which the perceived obstacle migrates from circumstance to self without being resolved. Longitudinal measurement of the same individuals across the deliberative interval is the natural next step, and is the design this paper cannot substitute for.

Data availability

The de-identified analytic dataset (n = 23,684; no free-text or identifying fields), a codebook documenting every variable and the mid-collection instrument change, both analysis scripts, pinned package versions, and the frozen analysis output are released with this paper. Every figure reported here is reproducible from that dataset with the included scripts, which read the CSV relative to their own directory and require no modification.

References

1. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211.
2. Sheeran, P., & Webb, T. L. (2016). The intention–behavior gap. Social and Personality Psychology Compass, 10(9), 503–518.
3. Gollwitzer, P. M. (1999). Implementation intentions: Strong effects of simple plans. American Psychologist, 54(7), 493–503.
4. Ibarra, H. (1999). Provisional selves: Experimenting with image and identity in professional adaptation. Administrative Science Quarterly, 44(4), 764–791.
5. Clance, P. R., & Imes, S. A. (1978). The imposter phenomenon in high achieving women. Psychotherapy: Theory, Research & Practice, 15(3), 241–247.
6. Samuelson, W., & Zeckhauser, R. (1988). Status quo bias in decision making. Journal of Risk and Uncertainty, 1(1), 7–59.
7. Armitage, P. (1955). Tests for linear trends in proportions and frequencies. Biometrics, 11(3), 375–386.
8. Wilson, E. B. (1927). Probable inference, the law of succession, and statistical inference. Journal of the American Statistical Association, 22(158), 209–212.

Cite this paper

Related: The Career Change Deliberation Study · Make the Leap Research