Terrence May

Dear Professor Swanson,

Having met you for the first time over 40 years ago, when you were hired as the State of Alaska’s first State Demographer, and now as a retired county government administrator, I found Melissa’s question and your solid answer gave me much to think about. 

For a demographer, this can be a Catch-22: you lose either way. Produce single-number forecasts that miss, and you get blamed.

Produce probabilistic forecasts, and decision-makers may ignore the upper and lower bounds and fixate on the middle value—then still blame you when it is wrong. Or they use the wide range as an excuse to do nothing: “Let’s wait and see.”

The Communication Trap

When you give decision-makers a single-number forecast:

· They quickly understand it.

· They believe it (often too much) and treat it as gospel.

· They can act on it.

· If it is wrong, they can blame you clearly: “We relied on the expert, and the expert got it wrong.”

When you give them a probabilistic forecast:

· Many decision-makers do not understand it. It is too complicated, and they have other pressing issues competing for their time. It forces them—often part-time elected officials—to think about multiple possible outcomes. Or they conclude you are waffling and do not want to be pinned down. They want quick, actionable, black-and-white answers.

· Those who do understand it may interpret it differently: some treat the bounds as hard limits; others treat them as meaningless.

· They cannot act on it directly. It creates paralysis by analysis. They must translate it into a decision, which is much easier with a single number.

· If it is wrong, they blame you for not being clear enough.

The demographer is in a double bind. Make it simple by implying false precision, and you risk being dishonest. Make it honest, and it becomes unusable because decision-makers have a low tolerance for uncertainty.

Clearly, probabilistic forecasting is technically superior and intellectually honest. The problem, as I see it, is twofold: our institutions and decision-making systems penalize it, and many decision-makers do not understand it. Demographers are therefore perversely incentivized to produce single-number forecasts. Those make for easier politics. Probabilistic forecasting makes for better science and better policy decisions.

So, I believe the demographer’s professional and ethical responsibility is to provide rigorous forecasts as a range of numbers based on probability, taking into account all relevant factors. Demographers should insist on doing so when asked to provide population projections.

Here is where education comes in. Decision-makers cannot put the onus on demographers to give them cover with single numbers that imply false precision. They must learn to think probabilistically and take full accountability for making informed decisions that require a tolerance for uncertainty. They must model different scenarios based on the demographer’s low, middle, and high projections, weigh the risks and uncertainties along with a cost-benefit analysis, and choose one—while remaining ready to reallocate resources as needed. 

Terrence May

You gave an excellent explanation of why demography matters beyond the academic world, and why it’s more than counting people. Applied demography has practical real-world implications.

One thing you said particularly got my attention and is worth repeating here. “While population aging is widely discussed, declining fertility rates may have even deeper consequences because they gradually reduce the number of people entering the workforce, forming families, and supporting social institutions.”

That’s the blind spot. We’re planning for an older society but not for a smaller one—and that’s the actual challenge. Many countries are (or soon will be) experiencing population decline. It has compounding effects. You can manage aging if the base is stable. You can’t manage anything if the base is collapsing.

Grateful for the clarity on what actually matters.