Which is harder: obtaining a relatively high-paying job or running a business with a good income? Comparing only the final amounts quickly becomes an exchange of examples. Someone achieved one; someone else achieved the other.
What I want to ask next is whether the conclusion still applies to a person with an ordinary starting point and limited resources.
This essay examines the comparison method. It does not estimate actual income distributions or recommend employment or business ownership. My concern is that a question apparently answered by data may not yet specify whom it intends to describe.
Unpack what ordinary means
Ordinary background, ability, and resources are intuitive descriptions, not ready-made analytical variables. Education, skills, location, available time, family responsibilities, capacity to absorb losses, and relationships can affect available paths.
These conditions do not necessarily move together. Someone may have little capital but extensive industry experience. Another may have savings but cannot tolerate interrupted income. Calling both ordinary does not give them identical opportunities.
I would identify the conditions most likely to change the result and state how others are treated. That narrows the question but brings it closer to actual circumstances.
There is an opposite danger: specifying so many conditions that the comparison describes one invented person for whom no evidence exists. The purpose is to reveal consequential differences. Missing data should remain uncertain rather than become fabricated probabilities.
Current outcomes are not chances from a starting point
The proportion of current business owners achieving an income uses current observed owners as its denominator. The proportion of people who reach it within a period after entering uses a different group.
Looking only at ongoing businesses may exclude people who left. Employment figures also need an identified denominator: the whole population, employed people, or a particular occupational sample.
Successful participants may differ from new entrants in experience, resources, and selection conditions. Assigning the characteristics of observed successes to prospective entrants does not establish the difficulty they face.
Hernán and Robins explain selection bias and conditions for causal comparisons. Those ideas help examine how a sample was formed; they do not provide real success probabilities for these two paths. Causal Inference: What If
The first questions are who became visible in the data, who did not, and whether the difference matters for the question being asked.
Income figures may measure different things
Wages, business revenue, operating profit, and disposable income are not interchangeable. Labor hours, capital, volatility, and continuity may be hidden behind a yearly amount.
A comparison should state its accounting period, whether the level is sustained, which costs are included, and how personal labor is treated. Avoiding incompatible quantities comes before elaborate modeling.
Difficulty itself may mean a high entry barrier, a low achievement rate, a long wait, or unacceptable consequences of failure. Two paths can rank differently on these dimensions.
More success stories will not resolve a disagreement about which difficulty is under discussion.
Include time and paths that do not succeed
The same final goal can involve different preparation periods and income variation along the way. Conditions during the wait may determine whether a path is feasible for someone with family responsibilities.
Failure is also more than missing a target. Can investment be recovered? Are acquired capabilities transferable? What options remain after exit? These are information requirements, not advance claims that one path is safer.
I would use conditional comparisons. Which factor becomes decisive when time is tight? What changes with existing expertise? Does the earlier ranking survive if income cannot be interrupted?
That does not produce personal financial advice. It identifies missing evidence and the conditions that deserve examination first.
Investigate gaps without filling them by assumption
Some activities or income forms may be difficult for public surveys to cover. That warrants checking the target population, sampling, questions, and nonresponse.
Possible omission does not establish many overlooked easy opportunities. Omitted cases might perform better or worse, or use quantities that cannot be combined with the original figures.
I would make the gap specific: which group is absent, whether supplementary evidence exists, whether definitions align, and whether a larger or smaller gap could change the judgment.
If an essential denominator is unknown, an exact probability is unjustified. Precision in presentation cannot repair an undefined comparison.
The habit transfers to technical choices
“Which model is better?” needs a task, input distribution, and evaluation. “Which study method works?” needs a population, setting, and decision. “This automation saves time” needs an account of who saves time and who acquires maintenance work.
Across these questions, the conditions of a conclusion should travel with it.
I want a comparison to leave a discussable statement: under these assumptions, this choice has these advantages; if a specified condition changes, revisit the judgment. It may feel less decisive than a universal ranking, but it can serve an actual decision.
Alongside the decisions behind systems, it raises another question: beyond a person’s starting point, who can change the opportunities and rules they face?