When Two Analyses Tell Different Stories, Look at the Assumptions

Jun 15
10 min read

A data literacy lesson from Singapore’s housing debate

Two analyses can examine the same issue and reach very different conclusions.

That does not always mean one is dishonest or the other is wrong. More often, it means they are asking slightly different questions, using different definitions, or starting from different assumptions.

Singapore’s housing affordability debate offers a useful example.

In an earlier article, we examined whether HDB resale housing has really become less affordable over time. Using HDB resale transaction data and income data from the Singapore Department of Statistics, our analysis suggested that HDB resale affordability has fluctuated, but has not structurally collapsed when measured against household income.

A contrasting argument has also gained attention: that two salaries today buy less than one salary did a generation ago.

That is a powerful claim. It speaks to a feeling many households recognise. Despite higher incomes, life does not necessarily feel easier. Housing prices feel large. Mortgages feel daunting. Families worry about retirement, children, ageing parents, and whether they are keeping up.

At the family dinner table, the argument feels true.

But this is exactly why data literacy matters.

When a claim feels emotionally persuasive, we need to examine it even more carefully — not to dismiss it, but to understand what it is really saying.

The question is not simply: “Which story is more convincing?”

A better question is: “Why do the conclusions differ?”

That question is at the heart of data literacy.

Data literacy is not just the ability to read charts, calculate ratios, or use analytical tools. It is the ability to examine how a conclusion was reached. What was measured? What was left out? What assumptions were made? What comparison point was chosen? What definitions were used?

This matters because data does not automatically create objectivity.

Data supports objective discourse only when the assumptions behind the analysis are made visible.

Disagreement is not the problem

In public debates, disagreement is often treated as a problem.

One side must be right. The other must be wrong. One narrative must win. The other must be dismissed.

But in good analysis, disagreement can be useful.

When two analyses reach different conclusions, the difference is often traceable. They may be using different measures, different definitions, different time periods, or different baselines.

That is what makes data valuable. It gives us a way to move the conversation from competing opinions to examinable assumptions.

Instead of arguing about who has the better story, we can ask:

Are both analyses measuring the same thing?
Are they using the same definition of affordability?
Are they comparing the same type of household?
Are they looking at the same type of flat?
Are they using a fair historical baseline?
Are the assumptions reasonable?

This is where data literacy becomes a practical skill.

It helps us disagree better.

What our earlier analysis found

In our earlier article, we examined HDB resale affordability using two main data sources: HDB resale transaction data and median monthly household income data from the Singapore Department of Statistics.

We focused on 3-room and 4-room resale flats.

This was a deliberate choice.

These flat types are among the more common public housing types for Singaporean households. They are also closer to the typical HDB living experience than larger or more premium flat types. For a broad discussion about HDB resale affordability, they offer a reasonable anchor.

We then compared median resale prices against median annual household income.

Based on this method, the median resale price of a 3-room flat has generally remained at around 3 to 4 years of median household income over the past two decades. For a 4-room flat, the ratio has generally been around 4 to 5 years.

By this measure, affordability has fluctuated, but it has not collapsed.

In fact, some earlier periods were worse by this measure. For example, 4-room resale flats in 1997 required a higher multiple of household income than in 2025.

This does not mean HDB resale flats feel cheap. They do not.

It also does not mean every household is comfortable. Lower-income households face real pressure. Younger families may feel anxious about large mortgage commitments. The upgrade path from public to private housing has become more difficult. These are valid concerns.

But they are not the same as saying that HDB resale affordability, measured against income, has structurally broken down.

That distinction matters.

Why another analysis may reach a different conclusion

A contrasting analysis can reach a much more worrying conclusion if it makes different methodological choices.

One argument that has gained attention is that two salaries today buy less than one salary did in the past. It is a powerful claim because it speaks to a broader sense of financial pressure.

But the conclusion depends heavily on how the analysis is constructed.

Three choices are especially important.

1. Individual income versus household income

The first choice is the income measure.

If we compare housing prices against individual income, affordability will naturally look more stretched than if we compare prices against household income.

That does not make individual income wrong. It simply answers a different question.

Individual income asks: how affordable is a flat for one worker?

Household income asks: how affordable is a flat for the household that is actually buying the flat?

For HDB affordability, household income is usually the more relevant measure. HDB eligibility, CPF usage, loan assessment, and mortgage servicing are generally considered at the household level rather than purely at the individual level.

This matters because dual-income households are not a new phenomenon. If many households already had more than one working person in earlier decades, then the historical baseline was not necessarily “one salary buys the flat.” It may have been closer to “household income buys the flat.”

This does not remove the emotional force of the dual-income argument.

Many families may indeed feel that the second income is no longer creating comfort, but merely helping the household keep up.

But that is a different claim.

It is a claim about broader household financial pressure, not necessarily a claim that HDB resale affordability itself has deteriorated dramatically.

2. Defined flat type versus unspecified flat type

The second choice is the housing measure.

If an analysis refers to a “standard HDB flat” without specifying the flat type, it becomes difficult to evaluate the conclusion.

Is the analysis referring to a 3-room flat, 4-room flat, 5-room flat, executive flat, or a blended average across flat types?

This matters because different flat types serve different household needs and have different price histories. A shift in the mix of flats being compared can change the result significantly.

For example, comparing smaller flats in one period with larger flats in another period may exaggerate the change in affordability. Similarly, using an average price without understanding the flat type mix may produce a conclusion that is harder to interpret.

This is not a minor technical detail.

It affects what the analysis is actually measuring.

Good data analysis requires us to define the object of comparison clearly. Without that clarity, readers cannot easily verify, replicate, or challenge the conclusion.

3. Stated assumptions versus unstated assumptions

The third choice is the baseline assumption.

Some affordability arguments suggest that a generation ago, households commonly finished paying off their HDB mortgages by their mid-40s, leaving many years of peak earning power for retirement savings

and wealth accumulation.

This may be true for some households. But as a general benchmark, it requires evidence.

A household buying a flat at age 30 on a 25-year loan would finish paying the mortgage around age 55, not the mid-40s. Finishing much earlier would require substantial prepayments, shorter loan tenures, or other financial support.

That may have happened for some families, but it should not be assumed to represent the typical household experience unless the data supports it.

This is where data literacy becomes especially important.

A claim can sound reasonable because it fits a familiar story. But if the baseline is not verified, the conclusion may rest on a memory, impression, or selective experience rather than a representative pattern.

Testing the single-income argument

One fair challenge to our earlier analysis is that household income may make affordability look more comfortable than it feels.

After all, one of the most powerful arguments in the current housing debate is that the second income has shifted from being a source of additional comfort to being structurally necessary.

In other words, households may be earning more, but they may not feel better off.

That is a valid concern. So we tested the argument using a more conservative assumption: what if we measured HDB resale affordability using only median individual income from work?

This does not mean we believe single income is the best measure of HDB affordability. HDB eligibility, CPF usage, loan assessment, and mortgage servicing are typically assessed at the household level.

But testing the single-income assumption is useful because it allows us to examine whether the conclusion changes dramatically under a more demanding benchmark.

The result is instructive.

For 3-room resale flats, the ratio has generally moved between about 5 and 8 years of individual income. For 4-room resale flats, it has generally moved between about 7 and 10 years, with the highest ratios occurring in the late 1990s.
In other words, even when we assume only one income, the data does not show a simple story of affordability getting steadily worse over time.

The ratio was highest in the late 1990s. It then fell in the early 2000s, rose again around 2011 to 2013, and eased after government cooling measures and market adjustments. In recent years, the ratio has risen again, but remains within the broad historical range.

This does not mean households feel comfortable.

A flat that costs five, six, or eight years of individual income is still a major financial commitment. It also does not mean every household can afford every flat, or that lower-income households are not under pressure.

But it does suggest that the claim “two incomes today buy less than one income did before” needs to be examined carefully.

The issue may not be that HDB resale affordability has structurally collapsed.

It may be that households today face a more complex set of financial pressures: higher aspirations, childcare costs, ageing parents, retirement concerns, lifestyle expectations, private housing aspirations, and anxiety about future security.

That is an important distinction.

One is a claim about HDB resale affordability.
The other is a claim about overall household financial pressure.

Both are worth discussing. But they should not be treated as the same claim.

Why both stories can feel true

At this point, it may be tempting to ask: so which analysis is right?

But that may not be the most useful question.

A better answer is that the two analyses may be describing different things.

One analysis is asking whether HDB resale flat prices have risen faster than household income. Based on our method, the answer appears to be: not dramatically, at least for 3-room and 4-room resale flats over the period studied.

The other argument is asking whether households feel more financially stretched today, even with two incomes. The answer may also be yes.

These two statements can both be true.

A stable housing affordability ratio does not mean households feel financially secure. Families may face pressure from childcare costs, ageing parents, retirement adequacy, healthcare concerns, education expenses, lifestyle expectations, and the wider gap between public and private housing.

In other words, the issue may not be that HDB resale affordability alone has collapsed.

It may be that overall household financial pressure has become more complex.

That is a more nuanced discussion.

It is also a more useful one.

But we can only get there if we separate the claims clearly.

Data helps us have better disagreements

This is the broader lesson.

Data does not remove disagreement. It makes disagreement more productive.

Without data, public discourse often becomes a contest of narratives. The most emotionally powerful story wins. The story that feels most familiar gets repeated. The claim that confirms our lived experience becomes persuasive.

Data gives us a different way to engage.

It allows us to ask:

What exactly is being claimed?
What evidence supports the claim?
What assumptions sit underneath the conclusion?
Would the conclusion change if we used another reasonable measure?
Are we discussing the same issue, or have we mixed several issues together?

These questions do not eliminate judgment.

They improve judgment.

That is why data literacy is not just a technical skill. It is a thinking skill.

It helps citizens evaluate public arguments. It helps leaders make better decisions. It helps organisations avoid being swayed by dashboards, charts, or AI-generated outputs that look convincing but may rest on weak assumptions.

In the age of AI, this becomes even more important.

Generative AI can produce fluent arguments very quickly. It can summarise, compare, explain, and persuade. But fluency is not the same as reliability. A well-written answer can still be based on incomplete evidence, unclear definitions, or flawed assumptions.

That is why human judgment remains essential.

The critical skill is not just knowing how to get an answer.

It is knowing how to examine the answer.

The real value of data literacy

The housing affordability debate is not just a debate about housing.

It is an example of how easily different analytical choices can lead to different conclusions.

Use individual income, and the story looks more worrying.

Use household income, and the picture may look more stable.

Leave flat type undefined, and the comparison becomes harder to evaluate.

Define flat type clearly, and the analysis becomes more transparent.

Assume early mortgage payoff was typical, and the past looks much easier.

Question that assumption, and the historical comparison becomes less straightforward.

This is the real value of data literacy.

It does not guarantee that everyone will agree.

But it gives us a better way to disagree.

It moves the conversation from:

“Whose story feels more convincing?”

to:

“Whose assumptions are more defensible?”

That is a healthier form of public discourse.

It is also a skill that every organisation needs.

Because in business, as in public policy, decisions are often shaped by competing narratives. Different teams may interpret the same numbers differently. Leaders may be presented with dashboards that point in different directions. AI tools may generate recommendations that sound confident but require careful validation.

The answer is not to reject data.

Nor is it to accept data blindly.

The answer is to build the literacy to question it well.

At FYT Consulting, we help professionals and organisations build the analytical and critical thinking skills to ask better questions, examine evidence more carefully, and make better decisions.

Not based on the most compelling narrative.

But on the most defensible reasoning.