Explanations and the Underclass Argument

The argument about why people are poor has been running, in recognisably the same form, since before sociology existed.

On one side: poverty is produced by the characteristics of poor people — their choices, habits, family arrangements, or the culture they have developed. On the other: poverty is produced by the arrangements they live under — wages, jobs, housing, discrimination, the design of the benefit system.

It is usually posed as an either-or and it is not one. But some parts of it have been settled by evidence, and knowing which parts is worth more than a position.

Three visitors, one neighbourhood, three explanations — and the test that distinguishes them.

A district of a large industrial city. In 1970 it had four large manufacturing employers within walking distance. By 1990 it had none.

The first visitor, arriving in 1990, describes what he sees: high unemployment, many households headed by one parent, low school completion, and a good deal of informal and illegal economic activity. His explanation is behavioural. People here are not working, are forming families in ways that entrench disadvantage, and are making choices that perpetuate their position.

The second visitor describes the same things and explains them culturally. A way of life has developed here, transmitted between generations — different expectations, different time horizons, a different relationship to work and to institutions. The neighbourhood, on this account, now reproduces its own disadvantage independently of the economy.

The third visitor asks what changed, and when.

The jobs left first. The behaviours the first two visitors describe increased after the plants closed, not before. The people were largely the same people. Their parents had worked in those plants.

And she asks a second question: what else happened when the jobs went?

The households that could leave, left — disproportionately those with the most resources and the most stable employment. Which removed from the neighbourhood not only income but the people who ran the church committees, sat on the school board, knew about jobs elsewhere, and constituted a marriage market with stable earnings. The shops that depended on their spending closed. The institutions that depended on their participation weakened.

So the neighbourhood in 1990 differed from the neighbourhood in 1970 not only in employment but in the density of connections through which employment is found (see 8.3.3 on networks), in the institutions that transmit expectations, and in the presence of adults in work.

Her explanation is that the concentration itself became a cause, on top of the original economic shock.

Now notice the structure of the disagreement. The first two visitors are describing real things. They are describing the state of the neighbourhood after the change and attributing it to the people. The third asked about sequence — which is 7.5.4's process tracing, and it is the question that separates the accounts.

Why does the same evidence get read in opposite directions for two hundred years?

Because the observable facts are compatible with both accounts, and distinguishing them requires exactly the things this course's Part 7 was about: sequence, comparison, and a counterfactual.

Cross-sectional observation cannot settle it. That poor people differ behaviourally from non-poor people is not in dispute and establishes nothing about direction (see 7.6.2). The behaviours could produce the poverty, the poverty could produce the behaviours, or both could follow from something else.

Four kinds of evidence can bear on it, and all four now exist.

Sequence — did the behaviour or the structural change come first? Comparison across societies — do countries with different arrangements produce different outcomes for similar people? Natural experiments — what happens when money or jobs arrive or leave for reasons unrelated to the recipients? And experiments — what do people actually do when given resources?

This lesson is organised around what those four have shown.

The recurring structure of the argument

Deserving and undeserving, and the first time it was tested.

English poor relief had operated for centuries on a distinction between those unable to work and those unwilling. The 1834 reform made it a governing principle: relief was to be made less eligible — deliberately less attractive than the worst available employment — on the reasoning that any relief above that level would create dependency.

The reasoning is the ancestor of every subsequent version of the behavioural argument, and it was a theory about incentives applied without evidence about who the poor actually were.

The first decisive empirical intervention came at the end of the century, from the surveys discussed in 7.6.1. Booth in London and Rowntree in York went and counted, household by household, why each poor household was poor.

What they found reorganised the argument. The largest causes were low wages in regular employment, the presence of several young children, illness, and old age. Idleness and drink accounted for a small minority of cases.

And Rowntree found something more interesting still: the poverty cycle.

Following the life course of an ordinary labourer, he showed that the same person passes into and out of poverty repeatedly — poor as a child in a household with several children, above the line as a young worker before marriage, poor again once children arrive, above it once the children earn, and poor again in old age.

This is the finding that reframed everything. Poverty was not the permanent condition of a distinct class of persons. It was a phase of an ordinary working life, which most of the working class would pass through more than once — and 8.6.1's evidence on persistent versus transient poverty is the same finding, restated with panel data a century later.

The three families of explanation

One — behavioural.

Poverty is produced or sustained by individual choices: about work, education, family formation, spending, and substance use.

What is right about it. Behaviour differs, behaviour matters, and pretending otherwise is not respectful of anybody. A decision to leave school, to have a child at a particular time, or to leave a job has consequences, and an account that treats people as having no agency is both false and condescending (see 1.5.3 on sociologism).

What is wrong about it as an explanation. The causal direction is assumed rather than shown. And the strongest single body of evidence against the assumed direction comes from what happens when the constraint is relaxed — below.

Two — cultural, and it is worth stating what its author actually said.

The culture of poverty argument, in its original form, described a set of adaptations — fatalism, present-orientation, distrust of institutions, particular family patterns — developed by people living in poverty within a stratified market society, and transmitted to children who then found them difficult to escape.

Three things about the original formulation are routinely omitted.

It was explicitly a reaction to conditions. Its author described the culture as an adaptation to marginality and hopelessness, arising where people are economically excluded from a society whose values emphasise accumulation.

He estimated it applied to a minority of poor people — around a fifth, in the settings he studied — and was clear that most poor people did not exhibit it.

And he wrote that it would disappear if the conditions producing it were removed.

The trouble is in the fourth claim, which he did make: that once formed, the pattern is self-perpetuating and does not simply vanish when circumstances improve. That claim is what made the argument usable by people whose conclusions he did not share, and it is the empirically contested part.

The subsequent history is a case study in how an argument gets used. A prominent policy report of the mid-1960s made a substantially structural argument — attributing family disruption to unemployment, low wages and the historic legacy of slavery and discrimination — and framed the result with a phrase about a tangle of pathology. The framing was what was remembered. The report became, in reception, an example of blaming families for their circumstances, which its structural chapters do not support. Both the report's argument and its reception are worth knowing, and neither is well represented by the way it is usually cited on either side.

And culture has since returned to poverty research in a more defensible form. Not as a set of deficient values, but as repertoires, frames and narratives: the culturally available ways of interpreting a situation and deciding what to do, which differ between groups, are learned, and shape action without being preferences and without being deficits. In that form culture is fully compatible with structural explanation, and it is what an account of the second visitor's observations should look like.

Three — structural.

Poverty is produced by the availability and quality of jobs, the level of wages, the cost of housing, the design of transfers, discrimination, and geography.

Its strongest version adds concentration effects: that the removal of employment from a place, followed by the departure of those able to leave, damages the institutions, networks and role models through which employment is found and expectations are formed — so that the concentration becomes an independent cause on top of the original shock.

And a serious internal argument runs alongside it about whether concentrated disadvantage is primarily the consequence of economic restructuring or primarily the consequence of residential segregation, which determines who is exposed to a downturn and prevents the poor from dispersing. On the segregation account, high segregation converts a rise in inequality into a concentration of poverty automatically — and the empirical case is that concentration effects appear where segregation is high and much less where it is not.

Both mechanisms are documented and they are not exclusive, which is the honest position.

What the four kinds of evidence show

Sequence. In the deindustrialising cases studied in most detail, the employment collapse preceded the behavioural changes attributed to culture. That does not prove the direction for every case, and it is exactly the wrong way round for the accounts that treat the behaviours as the cause.

Comparison across societies. This is the test the behavioural account handles worst.

The claim that generous transfers produce dependency and family breakdown predicts that societies with more generous systems should have more of both. They do not. Countries with substantially more generous and less conditional systems have lower poverty, lower child poverty, and no systematic excess of the outcomes predicted. The cross-national pattern runs the opposite way to the prediction, and this is the single most damaging piece of evidence against the strong version.

A widely influential version of the argument held that the expansion of American welfare provision from the 1960s had caused rising non-employment, family breakdown and dependency. The timing is against it in a specific and checkable way: the real value of the main cash benefit fell substantially through the 1970s and 1980s while the outcomes attributed to it worsened. A cause that was declining cannot explain an effect that was rising, and subsequent research found the behavioural responses to benefit generosity to be small.

Natural experiments. The cleanest available comes from an unanticipated income stream arriving for reasons unrelated to the recipients.

In a long-running study of children's mental health, a subset of families began receiving substantial regular payments partway through, as a result of a new revenue source distributed to members of a particular community. The payments were not means-tested, not conditional, and not related to anything about the recipients.

Children in families lifted out of poverty by the payments showed reductions in behavioural and conduct symptoms, converging towards the levels of children who had never been poor, with the effects concentrated among the previously poor and absent for those who had not been. Parental supervision improved.

Experiments. And the most direct answer to the behavioural claim comes from cash transfer programmes, of which there are now a great many, evaluated experimentally.

The recurring public objection is that money given without conditions will be wasted on alcohol and tobacco. A systematic review of the evidence across many programmes and countries found no systematic increase in spending on those goods — with the great majority of studies finding no effect or a negative one.

And the measured effects run the other way: increased schooling, better nutrition, investment in productive assets and small enterprise, and improvements in health.

This is as close to a decisive answer as social science produces on a behavioural question, and it should be weighed as such.

And the finding that reconciles more of the argument than any other, with its caveats.

A body of work on scarcity argues that being short of a resource — money, time, food — changes how people think, in ways that are functional in the short run and costly in the long run.

The core mechanism: scarcity captures attention. The pressing shortage becomes the focus, which improves performance on the immediate problem and reduces the mental capacity available for everything else — planning, deferring, remembering the appointment, evaluating a long-term trade-off.

Why this matters for this lesson. The behaviours attributed to a culture of poverty — short time horizons, neglect of the non-urgent, apparently poor decisions about long-term commitments — are what this mechanism predicts as consequences of scarcity, not as causes of it. The direction is reversed, and the account is compatible with people being ordinarily competent.

And the caveats must be stated. The most widely cited specific demonstrations of cognitive effects have had replication difficulties, and some later studies using different designs have failed to find the predicted effects around periods of financial pressure. The general framework is better supported than the headline effect sizes, and this is precisely the situation 7.8.2 described.

So: an illuminating and probably partly correct account, whose specific quantitative claims should be held loosely — which is the proportionate reading, and more useful than either citing it as established or dismissing it because one study did not replicate.

Because the either-or framing is the mistake, and it survives because both sides find it useful.

The defensible synthesis has three parts, and each is supported.

Behaviour matters and is substantially produced by circumstances. People make decisions; the decisions are made under constraints; the constraints affect both the options and the capacity to evaluate them. Neither the account with no agency nor the account with no structure describes what the evidence shows.

Culture is real and is not a deficit. It is a repertoire of available interpretations and strategies, learned, adapted to circumstances, and changing when circumstances do. In that form it explains local variation that structure alone does not, and it does not license the conclusion that changing people's attitudes would change their position.

And structure is where the variation is. The strongest single argument in this whole Topic is comparative: countries with similar technology, similar demography and similar attitudes have very different poverty rates, and the differences track wages, transfers, housing costs and public services (see 8.1.2, 8.5.1, 8.4.2 — the same conclusion has now been reached from four directions).

Three questions to carry.

What is the sequence? Which changed first — the circumstances or the behaviour?

What is the comparison? If an explanation appeals to something universal, it must explain why the outcome differs so much between comparable societies.

And what happens when the constraint is relaxed? This is the question the cash transfer evidence answers, and the answer has been consistent across dozens of programmes.

One last thing about the argument's persistence. It has run for two centuries in the same form. That is itself a sociological fact requiring explanation, and the explanation is not that the evidence is unavailable. It is that the two accounts imply very different obligations — one requires changing poor people and one requires changing arrangements — and the question of who bears a cost is settled by which account prevails (see 7.1.2). Which is why it keeps being reopened regardless of what has been shown.

A neighbourhood observed in 1990 supports three explanations, and only the question of sequence separates them. In the best-studied cases the employment collapse came first, and the departure of those able to leave removed the networks, institutions and role models through which work is found — so the concentration became a cause on top of the original shock.

The argument's structure is two centuries old. The 1834 principle of less eligibility applied an incentive theory without evidence about who the poor were; Booth and Rowntree then counted, and found low wages, several young children, illness and old age accounted for most poverty. And Rowntree's poverty cycle showed the same person passing into and out of poverty across a life course — the finding that panel data reproduced a century later.

Three families of explanation. Behavioural — right that behaviour matters, wrong to assume the causal direction. Cultural — and the original formulation described adaptations to marginality, estimated they applied to a minority, and said they would disappear if conditions changed; the contested claim is that they persist once formed. The subsequent history shows how a structural argument framed with one memorable phrase was received as its opposite. Culture has returned in a defensible form as repertoires and frames rather than deficient values, compatible with structural accounts. And structural — jobs, wages, housing, transfers, discrimination and geography, with an internal argument about whether concentration follows from economic restructuring or from segregation, which converts rising inequality into concentrated poverty automatically.

Four kinds of evidence. Sequence favours the structural account in the cases studied most closely. Cross-national comparison is decisive against the strong behavioural claim: more generous systems produce less poverty, not more dependency — and the influential argument that welfare expansion caused American family breakdown fails on timing, since the real value of the main benefit fell while the outcomes worsened. A natural experiment in which unconditional payments arrived for reasons unrelated to recipients found children's behavioural symptoms improving, concentrated among the previously poor. And experimental evidence across many cash transfer programmes finds no systematic increase in spending on alcohol and tobacco, with gains in schooling, nutrition and productive investment.

The scarcity framework reverses the direction: short horizons and neglect of the non-urgent are what shortage produces, not what produces it. Its general account is better supported than its headline effect sizes, several of which have had replication difficulties.

And the argument persists because the two accounts imply different obligations, not because the evidence is unavailable.

Deserving / undeserving poor — the distinction underlying poor relief and its modern successors.

Less eligibility — the principle that relief must be less attractive than the worst available work.

Poverty cycle — Rowntree's finding that a life course passes into and out of poverty repeatedly.

Culture of poverty — adaptations to marginality, described by its author as applying to a minority and as removable by changing conditions; contested in its self-perpetuation claim.

Cultural repertoire / frame — the available interpretations and strategies through which situations are understood; not preferences and not deficits.

Concentration effects — the additional damage caused by the co-location of disadvantage after selective out-migration.

Spatial mismatch — the separation of workers from the locations where jobs are.

Residential segregation — the mechanism converting rising inequality into concentrated poverty.

Dependency thesis — the claim that transfers produce non-employment and family breakdown; contradicted by cross-national comparison and by timing.

Bandwidth tax — the reduction in mental capacity available for non-urgent matters under scarcity.

Temptation goods evidence — the systematic finding that cash transfers do not increase alcohol and tobacco spending.

One — ask about sequence. Take any account of a disadvantaged place and establish what changed first: the economy or the behaviour. The information is usually available and rarely presented.

Two — run the comparative test. Take a claim that a benefit or a family pattern causes poverty. Find a country with more of the alleged cause and check whether it has more of the effect.

Three — read the original. Find the actual text of the culture of poverty argument and note the three qualifications its author placed on it. Then note how it is usually cited.

Four — check the timing. For any claim that a policy caused an outcome, plot both over time. A cause that was shrinking cannot explain an effect that was growing.

Five — apply the scarcity test to yourself. Recall a period when you were short of time or money. List the non-urgent things you neglected, and ask whether an observer would have described those as a cultural characteristic.