Reading Society
Sociology Fundamentals
Course 1 · Chapter 10 · Reading Society: Putting the Imagination to Work
Chapter 9 ended by promising that this chapter would ask how to check a claim against the figures and would draw the course together. Every chapter so far has carried a habit: facts that came from a page were attributed to it, and anything from my own general knowledge or arithmetic was marked "(check". This chapter turns that habit into a method for reading the claims you meet every day, in headlines, reports and conversation. It first borrows five ideas about evidence from Wikipedia pages on statistics and reasoning, then applies them, together with ideas from Chapters 1 to 9, to an invented news story.
Five Traps in Reading Evidence
Chapter 1 described the sociological imagination as the awareness of the relationship between personal experience and the wider society. Using it on a claim means asking what the claim is really based on. The five Wikipedia pages below name the commonest ways a claim goes wrong. They are general pages on statistics and reasoning, not sociology pages, so the links to sociology in this section are my own (check).
| Trap | What the page says | Source page |
|---|---|---|
| Correlation taken as cause | The page describes "the inability to legitimately deduce a cause-and-effect relationship between two events or variables" solely from an observed association. It names reverse causation, a hidden third variable affecting both, and coincidence. | Correlation does not imply causation |
| Group figures applied to individuals | The ecological fallacy, "a formal fallacy in the interpretation of statistical data", occurs when inferences about individuals are deduced from inferences about the group they belong to. | Ecological fallacy |
| Trends that reverse when groups combine | Simpson's paradox is a phenomenon in which a trend appears in several groups of data and "disappears or reverses when the groups are combined". | Simpson's paradox |
| Who got counted | Selection bias is "the bias introduced by the selection of individuals, groups, or data for analysis". The page lists sampling bias, volunteer bias (self-selection), attrition and nonresponse bias among its forms. | Selection bias |
| Seeing only what fits | Confirmation bias is the tendency to search for, interpret, favour and recall information in ways that support existing beliefs. The page names biased search, biased interpretation and biased recall. | Confirmation bias |
Evidence from the pages themselves
Each page gives at least one real case, which is better than an invented one for seeing the trap clearly.
- Ecological fallacy. The page says the term was coined in 1958 by Selvin, after W. S. Robinson's seminal 1950 paper. Robinson used 1930 census data, with a state-level correlation between illiteracy and the proportion of people born outside the US that the page first gives as -0.53, and then, following a 2011 correction, as -0.46. The individual-level correlation was +0.12. The correction changes the figure by 0.07 (arithmetic done in Python). The important point is the sign: the group-level relationship was negative and the individual-level one positive, so the group figure pointed the wrong way for individuals. The page's own correction is also a lesson: even a famous number can be revised.
- Simpson's paradox. The page reports the 1973 University of California, Berkeley admissions figures: 8,442 men applying with about 44% admitted and 4,321 women with about 35% admitted, 12,763 applicants in all, with about 41% admitted. Applying the page's rounded percentages gives roughly 3,714 men and 1,512 women admitted, and 41.0% overall (computed in Python from the page's approximate percentages, so the counts are estimates) (check). The page attributes the 1951 paper to Edward H. Simpson, and the name "Simpson's paradox" to Colin R. Blyth in 1972; the gap between those two dates is 21 years (computed in Python). The excerpt I received did not give the department-level breakdown, so I do not state what explains the Berkeley gap.
- Correlation and cause. The page's examples include a 1999 study published in Nature linking childhood sleep with the light on to later short-sightedness, and a later Ohio State study that found parental short-sightedness was the common cause. It also describes the "Redskins Rule", which predicted every US presidential election from 1936 to 2000 inclusive, and then stopped working. 2000 minus 1936 is 64 years, so even a long run of success can be coincidence (computed in Python; the interpretation is the page's, as I read it).
- Selection bias. The page's examples are qualitative: in a dieting trial, those who drop out are often the ones for whom the programme was not working, so the people who remain look more successful than the whole group was. It gives no figures, so none are quoted.
- Confirmation bias. The page credits Peter Wason with the concept, in the sense of a preference for information consistent with a hypothesis, and mentions his 1960 experiment. It gives no year for when the term itself was coined.
A Worked Example: An Invented Headline
Everything in this section is invented for practice. The town, the school and all the numbers are made up, and none comes from a source. Suppose a local paper prints this headline: "Exclusions up 25% as pupils glued to their phones." The article reports that a secondary school excluded 40 pupils last year and 50 this year, that the school roll was 1,250 last year and 1,300 this year, and that a head teacher says phones are the cause. Here is how the course lets you read it.
| Question | Where it comes from | Applied to the headline |
|---|---|---|
| Is the number a rate or a total? | The statistics pages above; the dark figure in Chapter 9 | 40 to 50 is a rise of 25% in the total, but the roll also rose by 4%. Per 1,000 pupils the figure goes from 32.0 to 38.5, a rise of about 20% (computed in Python). The rise is real but smaller than the headline says. |
| Is it cause or correlation? | Correlation does not imply causation | The article shows no data on phone use at all. Even if both rose, a third variable or reverse causation could be at work. |
| Who decides what counts? | Labelling and social control, Chapter 9 | An exclusion is a formal sanction. A change in the school's rules, or a new head teacher, could raise exclusions without any change in pupil behaviour (my own suggestion) (check). |
| What is one school's story worth? | Selection bias; the ecological fallacy | One school over one year is a small and possibly unrepresentative sample, and a school-level figure says nothing about any individual pupil. |
| Whose voice is quoted? | Confirmation bias; Chapter 9 on moral panics | The head teacher is a single source with a view. Chapter 9 described how media can amplify a threat; whether this is a case of that needs more than one story (my own comparison) (check). |
Applying the Whole Course
Real claims rarely stay inside one chapter. The habit that makes this course's pages trustworthy is the same habit you can use on others' claims: say where each piece comes from, mark the parts that you added yourself, and say how much you actually read.
Where is the source?
Chapter 1's sociological imagination asks you to separate a personal trouble from a public issue. Ask whether a claim rests on a named source or only on "experts say".
How much was seen?
Chapters 4 to 9 said how much of each page was read. The Confirmation bias page, for instance, was read to 100,000 of 266,612 characters, which is 37.5 percent, leaving 166,612 unread (computed in Python). A claim about a long document should say how much of it was read, too.
Is it your own addition?
Mark opinions and arithmetic as yours. That is the job of the "(check" label used since Chapter 1.
Do sources disagree?
Chapter 4 found two pages giving Mead different dates, and Chapter 7 found two pages phrasing Weber differently. When sources differ, say so; do not pick quietly.
- Who says it, and what is their stake? (Chapter 2: whose interests; Chapter 9: who labels.)
- Is the figure a rate, a total or a share, and compared with what? (Chapters 7 and 8: the pay-gap and poverty-line examples.)
- Does it claim a cause where it only shows a link?
- Who was counted, and who was left out? (Chapter 9: the dark figure of crime.)
- Does a group figure get applied to individuals?
- Is a category being treated as fixed when it is constructed? (Chapter 8.)
- What would change my mind, and have I looked for it? (The confirmation bias trap.)
Where to Read More
- Pages used. This chapter uses six Wikipedia pages: Correlation does not imply causation, Ecological fallacy, Simpson's paradox, Selection bias, Confirmation bias and Misuse of statistics. The Confirmation bias page was cut off by the fetch tool: only the first 100,000 of 266,612 characters were read (37.5 percent). For the other five the tool reported no length, but each appeared to run to its references and footer, so they seem complete; I cannot tell the exact length of any of them.
- Summaries. The fetch tool returns a summary made by a small model, not the raw page, so quotations are only as exact as that summary.
- Not covered. I did not read the Berkeley department-level data, Robinson's own paper or Selvin's. I also did not read anything on base rates, since the Misuse of statistics page did not mention them.
- Beyond this course. Read the original works of the founders named in the earlier chapters if you want more than a page-based account, and consider a general introductory university text on sociology and one on statistics, which cover the same ground in more detail (general suggestions of mine, not drawn from the pages) (check).
Exercises
A town council reports (invented figures) that a youth club cost 90,000 last year and that 450 young people used it, while reported youth offences in the town fell from 120 to 96 over the same year. A councillor says the club cut offending by 20%. First, check the arithmetic: the cost per young person and the percentage fall. Then apply ideas from at least three earlier chapters (for example Chapter 9's differential association, labelling and the dark figure of crime, and Chapter 5's groups) to say what the councillor's claim would need to show before you accepted it.
Critique this invented claim: "Regions with more graduates have higher average incomes, so a degree makes each person richer." Name which of the five traps it falls into, using the Ecological fallacy and Correlation pages as the chapter reports them. Give one alternative explanation, then say what evidence at the level of individuals would be needed.
Choose one earlier chapter and summarise it in your own words in about 100 words. Then state your confidence in the summary as high, medium or low, and give the reasons using the course's habits: how much of its pages were read (look at the chapter's "Where to Read More"), which statements were marked "(check", and whether any sources disagreed. Finish with one question the chapter leaves open.
Quick Reference
- Correlation and cause: an association alone does not show cause; look for reverse causation, a third variable and coincidence
- Ecological fallacy: deducing facts about individuals from facts about their group (term from Selvin, 1958; Robinson's 1950 case)
- Simpson's paradox: a trend in each group that reverses when the groups are combined
- Selection bias: who was selected for analysis changes the answer; sampling, volunteer and attrition bias are forms
- Confirmation bias: searching for, interpreting and recalling information that fits what you already think
- Rate versus total: in the invented headline, 25% in the total became about 20% per 1,000 pupils
- Reading limits: the Confirmation bias page was read to 37.5 percent
- Check: the links from statistics pages to sociology, the Berkeley counts, the exclusion-rules suggestion, the lens applications and the suggested further reading
You have finished Sociology Fundamentals. What each chapter taught:
- Chapter 1, What Sociology Is: the study of human society, the word's origin, Durkheim's social facts, Mills's sociological imagination and the methods used.
- Chapter 2, Theories of Society: functionalism, conflict theory and symbolic interactionism as three lenses, with Marx, Weber and Durkheim behind them.
- Chapter 3, Culture: norms, folkways and mores, values, symbols, subcultures, culture shock and cultural relativism.
- Chapter 4, Socialisation and Identity: how people learn a culture, the looking-glass self, Mead's "me" and "I", roles, Goffman and feral-child evidence.
- Chapter 5, Groups and Organisations: primary and secondary groups, community and society, bureaucracy, the iron law of oligarchy and weak ties.
- Chapter 6, Structure and Institutions: social structure, status, roles, institutions and competing accounts of social order.
- Chapter 7, Inequality and Class: stratification, three accounts of class, mobility, absolute and relative poverty and the Gini coefficient.
- Chapter 8, Race, Ethnicity and Gender: social construction, ethnicity, discrimination, the pay gap and intersectionality.
- Chapter 9, Deviance and Control: strain, differential association, labelling, moral panics, crime as a legal category and social control.
Where to go next: reread any chapter's "Where to Read More" and follow one source to its original text; pick one statistic from this week's news and run the seven-question checklist above on it; and keep marking what is yours and what is borrowed. An introductory university text in sociology, and one in statistics, are sensible general next steps.