# Lesson 11: Social, Economic, and Political Determinants of Health

*Companion-podcast transcript • Sarah & Kiffer*  
*Office Hours episode to listen to after working through the lesson*

---

**Sarah:** Welcome back to Office Hours. I'm Sarah.

**Kiffer:** And I'm Kiffer. This episode goes with Lesson eleven, Social, Economic, and Political Determinants of Health.

**Sarah:** It's for after you've finished the lesson. We'll add some perspective on how these ideas are used, some critique of the evidence behind them, a few of the questions students tend to find thorny, and some extra worked examples, including one or two that are harder than anything on the lesson page.

**Kiffer:** There are also a few places where we'll ask you to work something out before we give the answer. When that happens, you'll hear a few seconds of quiet. Pause the audio if you'd like more time to think it through.

**Sarah:** We'll take four questions. First, why it matters that the social gradient is a slope across the whole population. Second, what it means when a study says health behaviours explain part of the gradient. Third, how we can tell whether a health gap is closing, and how that connects to the difference between inequality and inequity. And fourth, a question Kiffer and I see a little differently, which is whether policy should be universal or targeted.

**Kiffer:** Question one. Why does it matter so much that the social gradient is a slope across the whole hierarchy, with no single cliff at the bottom?

**Sarah:** I think a lot of students hear the word gradient and nod along, but the picture in their heads is still poverty. There's a group at the bottom that's sick, and everyone else is roughly fine.

**Kiffer:** That's the picture the Whitehall studies challenged. Within one employer, each grade had higher mortality than the grade above it. The people in the middle were less healthy than the people at the top, even though none of them were poor. So let's see what that shape does to the numbers, with a made-up example.

**Sarah:** Go ahead.

**Kiffer:** Picture a hypothetical city of one hundred thousand adults, split into five income groups of twenty thousand people each. In the top group, five people per thousand die each year. Then it climbs one step at a time: six per thousand in the second group, seven in the middle group, eight in the fourth group, and ten per thousand in the bottom group.

**Sarah:** So the bottom group has twice the death rate of the top group.

**Kiffer:** Right. Now suppose every group had the top group's death rate. The deaths above that level are the excess deaths linked to the gradient. Here's the question for everyone. What share of the excess deaths happen in the bottom group? Take a few seconds.

*(Pause)*

**Sarah:** My first instinct is to say most of them, maybe three quarters, because the bottom group has by far the highest rate.

**Kiffer:** That's the instinct the example is built to test. Let's count them. For each group, take its rate minus five per thousand, and multiply by twenty thousand people.

**Sarah:** Okay. The second group is one extra death per thousand, times twenty thousand, which is twenty excess deaths. The middle group is two per thousand, so forty. The fourth group is three per thousand, so sixty. And the bottom group is five per thousand, so one hundred.

**Kiffer:** And what does that come to when you add up the excess deaths across all four groups?

**Sarah:** Twenty plus forty plus sixty plus one hundred is two hundred and twenty excess deaths. So the bottom group has one hundred of the two hundred and twenty, which is about forty-five percent. My guess was well off. More than half of the excess deaths, about fifty-five percent, happen in the three middle groups.

**Kiffer:** And that's the practical point of the gradient. Each middle group has a smaller excess, but there are more people in the middle groups combined. Geoffrey Rose made a related argument in 1985, in a paper called Sick Individuals and Sick Populations. He pointed out that a large number of people at a small risk may give rise to more cases than a small number of people at high risk.

**Sarah:** So a program that reached only the bottom group, and somehow brought its death rate all the way down to the top group's level, would still leave more than half of the excess deaths in place.

**Kiffer:** In this example, yes. The real numbers vary by outcome and by place, but the shape is common, and it's the main reason the lesson ties the gradient to universal policy. We'll come back to that in the last question, because the conclusion is more debatable than the arithmetic.

**Sarah:** Question two. Whitehall II found that the lowest employment grade had higher mortality than the highest, and that health behaviours explained part of that gap. What does explained actually mean here?

**Kiffer:** Let's start with the published numbers. Silvia Stringhini and colleagues analysed 9,590 Whitehall II participants, with deaths tracked to 2009. People in the lowest socioeconomic group had about 1.6 times the risk of death of people in the highest. Smoking, alcohol, diet and physical activity measured at the start of the study explained about forty-two percent of that association. When the behaviours were measured repeatedly over follow-up, the figure rose to about seventy-two percent.

**Sarah:** I'd like to try turning that into a number. If behaviours measured at the start explain forty-two percent, what would the ratio look like after taking them into account?

**Kiffer:** Good, let's give everyone a moment on that one too. Starting from a ratio of 1.6, what's left after removing forty-two percent of the association? Take a few seconds.

*(Pause)*

**Sarah:** Forty-two percent off means fifty-eight percent remains. So 1.6 times 0.58, which is about 0.93.

**Kiffer:** Before we move on, what would a ratio of 0.93 mean for people in the lowest grade?

**Sarah:** It would mean they had slightly lower mortality than the highest grade, after adjustment. That can't be right. Taking behaviours into account would have flipped the gradient upside down.

**Kiffer:** That sanity check is worth doing every time. So where did the calculation go wrong?

**Sarah:** I took the forty-two percent off the whole ratio. But a ratio of one means no difference at all. The association is only the part above one, which is 0.6. So forty-two percent of 0.6 is explained, and fifty-eight percent of 0.6 remains. That's 0.6 times 0.58, which is about 0.35. Add back the one, and the adjusted ratio is about 1.35.

**Kiffer:** That's the usual simple way to read a percentage explained. Researchers sometimes do the calculation on a different scale, so treat this as a rough reading. Now try the same steps with the seventy-two percent from the repeated measures.

**Sarah:** Twenty-eight percent of 0.6 remains, which is about 0.17. So the ratio would fall to about 1.17. Still above one, but much closer to it.

**Kiffer:** Now here's the thorny part. It's tempting to read that seventy-two percent as meaning the gradient is really about personal choices. Why would that be a misreading?

**Sarah:** Because the behaviours themselves follow the gradient. People in lower grades smoke more, for example, and that pattern has social causes of its own.

**Kiffer:** Right. In the language of the WHO Commission's framework from the lesson, behaviours are intermediate determinants. They sit on the pathway between social position and death. When a study adjusts for them and the gradient shrinks, that tells us something about how social position gets into the body. It doesn't tell us that social position stopped mattering.

**Sarah:** This connects to the Black Report, too. It offered four kinds of explanation for the class gradient: artefact, meaning a quirk of measurement; selection, meaning that poor health pushes people down the social ladder; materialist or structural explanations; and cultural or behavioural ones. The report favoured the materialist explanations.

**Kiffer:** And the behavioural findings fit inside a structural account. If smoking is more common in lower grades because of stress, advertising, the cost of quitting supports, and so on, then a behaviour is one of the ways a structural cause does its work. The selection explanation is the one worth keeping in view as a real alternative, because in many data sets illness can cause lower income as well as the other way around. Studies that measure social position before illness begins, and then follow people forward in time, help to separate the two directions.

**Sarah:** Question three. How can we tell whether a health gap is closing? This one matters for the Truth and Reconciliation Commission's Call to Action nineteen, which asks the federal government, in consultation with Indigenous peoples, to set measurable goals to identify and close the gaps in health outcomes, and to report on progress each year. Infant mortality is one of the indicators it names.

**Kiffer:** And it turns out that whether a gap has closed depends on how you measure it. Let's use made-up numbers, for two hypothetical regions. Ten years ago, infant mortality in Region A was twelve per thousand live births, and in Region B it was five per thousand. Today, Region A is at eight per thousand and Region B is at three. Here's the question. Did the gap between the two regions get smaller? Take a few seconds.

*(Pause)*

**Sarah:** I'll start with the difference. Ten years ago, twelve minus five is seven per thousand. Today, eight minus three is five per thousand. So the absolute gap narrowed by two deaths per thousand births.

**Kiffer:** Now do the same comparison with the ratio, dividing one region's rate by the other's.

**Sarah:** Ten years ago, twelve divided by five is 2.4. Today, eight divided by three is about 2.67. So on the relative scale, Region A went from 2.4 times Region B's rate to about 2.7 times. The relative gap widened.

**Kiffer:** Both statements are true at once. You can see why if you work out each region's own improvement. Region A fell from twelve to eight, which is a drop of one third. Region B fell from five to three, which is a drop of forty percent.

**Sarah:** So the region that was already better off improved faster in relative terms, while the worse-off region gained more deaths prevented per thousand births.

**Kiffer:** Exactly. That's why a report that says only the gap is closing, or only the gap is widening, should make you ask which measure it used. A careful report gives both.

**Sarah:** And this is where the inequality and inequity distinction from the lesson comes in. The difference between the two regions is a health inequality, because it's a measured difference. Calling it an inequity is a further judgement, that the difference is unnecessary, avoidable and unfair. Margaret Whitehead set out that distinction in 1992.

**Kiffer:** Here's a way to see that the judgement is doing real work. Suppose the two regions differ because Region A has a much older population, and the outcome is all-cause adult mortality. Part of that difference is age, and almost nobody would call the age part unfair. With infant mortality, age structure isn't an issue, and when the difference traces back to underfunded services, housing and water, it meets all three parts of the definition. The arithmetic gives you the size of the gap. Deciding that it's an inequity takes an argument about causes and fairness.

**Sarah:** And the choice between absolute and relative gaps can carry a judgement too. If a government wanted to report good news here, it would pick the difference. If an advocate wanted to show that things were getting worse, they'd pick the ratio.

**Kiffer:** Which is a good argument for setting the measure in advance, with the communities affected, before the results come in. Call to Action nineteen asks for the goals to be set in consultation with Indigenous peoples, and agreeing on how progress will be measured is part of that.

**Sarah:** Question four, and this is where Kiffer and I see things a little differently. The lesson argues that universal programs flatten the gradient more effectively than targeted ones. I'm more sympathetic to targeting than the lesson is.

**Kiffer:** Go ahead and make the case for targeting, and then I'll make the case for universal programs.

**Sarah:** My first point is about need. If a budget is limited, spending it on everyone means some of it goes to people at the top who are already doing well. Targeting puts more of each dollar where the risk is highest. My second point is that universal programs don't always reach everyone equally. Cesar Victora and colleagues described what's called the inverse equity hypothesis in 2000. New health interventions tend to reach better-off groups first, so inequalities between groups can widen at first, and they narrow only once the intervention reaches the most deprived groups.

**Kiffer:** Those are fair points, and the second one is a real limit on the universal argument. Here's my side. Our city example showed that more than half of the excess deaths can sit above the bottom group, and a program aimed only at the bottom can't reach them. Means-tested programs also carry stigma, people miss out because the paperwork is hard, and programs used only by the poor tend to have fewer political defenders when budgets are cut.

**Sarah:** Let me test that with our made-up city. Suppose one program is aimed only at the bottom group and cuts its death rate by twenty percent, from ten per thousand to eight. That's two fewer deaths per thousand, times twenty thousand people, so forty deaths prevented each year. A second program is universal and cuts every group's death rate by five percent. The city has one hundred plus one hundred and twenty plus one hundred and forty plus one hundred and sixty plus two hundred deaths a year, which is seven hundred and twenty. Five percent of that is thirty-six deaths prevented.

**Kiffer:** So in this example the two programs prevent about the same number of deaths. What happens to the gap between the top and bottom groups under each one?

**Sarah:** With the targeted program, the bottom group goes to eight per thousand, so the gap with the top shrinks from five per thousand to three. With the universal program, the top falls to 4.75 and the bottom to 9.5. The difference only shrinks to 4.75, and the ratio stays at exactly two. So in this example, the targeted program does more to narrow the gap.

**Kiffer:** It does, with the effect sizes we chose, and those were made up. A universal program that worked better in the middle groups, or a targeted one that missed many eligible families, would change the answer. What the example shows is that the choice depends on how well each program reaches each group, which is an empirical question.

**Sarah:** But you'd accept that universal doesn't automatically mean equal benefit.

**Kiffer:** I would. My view is that a universal foundation is the better default for the reasons I gave, and that it works best when the support is scaled up for the people who need it most. The 2010 Marmot Review in England gave that idea a name, proportionate universalism. It describes services delivered to everyone, at a scale and intensity proportionate to the degree of need.

**Sarah:** And Canada has a familiar example. The Canada Child Benefit, which began in July 2016, is a tax-free monthly payment for families with children, and the amount is reduced as family net income rises above a threshold. So it's broad, and it's also scaled to need.

**Kiffer:** That's a good example of the middle ground. I'd still lean toward universal foundations for things like hospital and physician care, where a single public system has been politically durable in Canada. You'd lean further toward targeting when the budget is tight.

**Sarah:** I would. But I think we agree on the practical rule. Start from a program that's open to everyone, scale the support to need, and check who actually takes it up, group by group, so that a universal program doesn't quietly widen the gap.

**Kiffer:** Let's pull it together with three things to take away from this episode.

**Sarah:** First, the gradient runs across the whole population. Because the middle groups hold so many people, a large share of excess deaths can occur above the bottom group, as the city example showed.

**Kiffer:** Second, when a study says behaviours explain part of the gradient, the percentage applies to the excess above a ratio of one, and the behaviours are pathways through which social position acts. They are part of the explanation of how the gradient works.

**Sarah:** And third, whether a gap is closing depends on whether you measure it as a difference or a ratio, and calling a gap an inequity is a judgement about causes and fairness that the numbers alone can't make.

**Kiffer:** If you'd like more practice, rework the two-region infant mortality example with your own made-up numbers, and find a case where the difference narrows while the ratio widens.

**Sarah:** Next time, it's Lesson twelve, Disability, Diversity, and Integrating the Foundations, where we look at disability as part of health and bring together ideas from across the course.

**Kiffer:** Take care, everyone.

**Sarah:** See you in Lesson twelve.
