Reading and Literature Studies · Reading Informational and Graphic TextsReading Complex Data Graphics: Rates, Margins of Error, Log Scales and Maps
2.2 1.6 1.7About 70 minutes
Learning goals- I can explain why totals and rates can tell opposite stories, and calculate a rate per 1,000.
- I can interpret a poll's margin of error and decide whether a difference is meaningful.
- I can read a logarithmic axis correctly.
- I can analyse how titles, annotations, colour and map design guide a reader's interpretation of a data graphic.
- I can evaluate whether a data graphic supports the claim made about it.
Words to know
rate
A number expressed per unit of population or time, such as thefts per 1,000 residents, allowing fair comparison between groups of different sizes.
per capita
Per person (Latin: by heads).
margin of error
The range around a poll result within which the true value probably lies, at a stated confidence level (often 19 times out of 20).
logarithmic scale
An axis on which equal distances represent equal multiplications (for example ×10) rather than equal additions.
choropleth map
A map that shades regions by a value; large, empty regions can visually dominate.
annotation
Text added to a graphic to point out a feature or tell the reader how to interpret it.
Hook: the city with the most thefts is the safest
A headline says City A has more than seven times as many bike thefts as Town B. True. It also has fifteen times as many people. Per person, a bike in Town B is twice as likely to be stolen. The same numbers support opposite headlines, depending on whether the graphic shows totals or rates.
Data graphics are text features with their own grammar. In university, in the news, and in reports you will read for work, you need to read that grammar fluently and critically.
Remember from Grade 11
In ENG3U you checked whether a bar chart's axis starts at zero, whether icons are scaled honestly, whether the sample is large enough, whether a time range was cherry-picked, and whether correlation was being presented as causation. Keep using all of these. Grade 12 adds rates, margins of error, log scales and maps.
1. Totals versus rates
To compare groups of different sizes, divide by size. Rate per 1,000 = total ÷ population × 1,000.
- City A: 900 ÷ 300,000 × 1,000 = 3.0 thefts per 1,000 residents.
- Town B: 120 ÷ 20,000 × 1,000 = 6.0 thefts per 1,000 residents.
Totals answer where is the biggest workload? (useful for a police budget). Rates answer where is the risk to each person highest? Neither is wrong; the question decides which one is honest.
2. Margins of error in polls
A poll of about 1,000 people usually reports a margin of error of about ±3 percentage points, 19 times out of 20. That means: if the poll were repeated many times, about 95% of the results would land within 3 points of the true value.
So a result of Yes 46%, No 42% really means Yes somewhere around 43 to 49, No somewhere around 39 to 45. The ranges overlap, so the poll cannot clearly say which side leads. A headline saying Majority backs library would also be wrong twice over: 46% is not a majority.
3. Logarithmic scales
When values range from 1 to 10,000, a normal axis squashes the small values flat. A log scale spaces 1, 10, 100, 1,000 and 10,000 equally, so each step means ×10. Two things to remember:
- A straight line on a log axis means steady percentage growth (for example, doubling every week), not steady adding.
- Halfway between 10 and 100 on a log axis is about 32, not 55 (because 10 × 3.16 ≈ 32 and 3.16 × 3.16 ≈ 10).
Log scales are common in science and in reporting on populations, earthquakes (the moment magnitude scale is logarithmic) and the spread of illness.
4. Maps and design choices
A choropleth map shades regions by a value. Large regions with few people (much of northern Canada, for instance) can dominate the picture even when most people live in small southern regions. Ask: Is the map showing land or people? Also read the legend's breaks: shading 0 to 49% one colour and 50 to 100% another can hide big differences inside each band.
Finally, read the words on a graphic as carefully as the data. The title states the claim the designer wants you to take away; annotations point your eye (Thefts doubled!); the source note tells you whom to trust; colour signals good and bad. These are the text features that turn numbers into an argument.
Worked exampleLet's evaluate a graphic and its headline
A news site posts the left-hand chart above (totals) with the title City A: Bike Theft Capital. Evaluate the graphic and title.
- Read the features first: title, axis label (number of thefts), source note (none given), annotation (none).
- Check the measure: it shows totals for places of very different size. A capital title implies risk, which needs a rate.
- Recalculate: City A 3.0 per 1,000; Town B 6.0 per 1,000. By risk, Town B is higher.
- Consider purpose: if the article is about the City A police workload, totals are relevant; if it is about where your bike is safest, the graphic misleads.
- Verdict and fix: the title overclaims. A fairer version: show both charts, titled More thefts in City A, higher risk per resident in Town B, with population and a source note.
Evaluate in order: features → measure → recalculation → purpose → verdict. Always ask whether the graphic's measure fits the claim in its title.
Worked exampleLet's read a log-scale claim
A science article's chart uses a log y-axis. A line rises from 100 to 1,000 between week 1 and week 4, then from 1,000 to 10,000 between week 4 and week 7. A reader says, It's growing at the same speed, so nothing has changed. Respond.
- Read the axis: each equal step is ×10.
- Interpret the slope: the line has the same slope in both periods, so the growth factor is the same: ×10 every three weeks.
- Compare amounts: but the absolute increase grows hugely: +900 in the first period, +9,000 in the second.
- Response: The percentage growth rate is unchanged, but the number added each period is ten times larger. A straight line on a log scale means steady multiplication, not steady addition.
Log scales show rates of growth clearly but can make large absolute increases look calm. Name which one you are describing.
Just for funAn old statistics joke: when a billionaire walks into a coffee shop, the average wealth of the customers jumps into the millions. The median customer still pays for their own latte.
Hands-on activityRecalculate a real graphic
You need: a news article or report with a data graphic about places or groups of different sizes (Statistics Canada's website and Canadian newspapers have many), a calculator, and about 30 minutes.
- Note the graphic's title, axis labels, legend, annotations and source.
- Decide whether it shows totals or rates. If totals, find the population figures (StatCan census profiles give them) and calculate rates per 1,000.
- If it is a poll, find the margin of error and check whether differences are larger than it.
- If it is a map, ask whether it shows land or people.
- Write a fairer title and a two-sentence evaluation.
What to look for: correct calculations shown; the measure matched to the claim; specific features named.
Your turn
1Calculate the rate per 1,000 for a town with 45 thefts and 15,000 residents.
45 ÷ 15,000 × 1,000 = 3.0 per 1,000.
2Why can't you say Yes leads in a poll of Yes 46%, No 42% with a ±3 margin of error?
The ranges (43 to 49 and 39 to 45) overlap, so the true values could be equal or reversed.
3What does 19 times out of 20 mean in a poll report?
If the poll were repeated many times, about 95% of results would fall within the margin of error of the true value.
4On a log axis marked 1, 10, 100, 1,000, what does each equal step represent?
5What does a straight rising line on a log axis show?
Steady percentage (multiplicative) growth, such as doubling every fixed period.
6What question should you ask of every choropleth map?
Is it showing land or people? (Large, sparsely populated regions can dominate the picture.)
7Name three text features of a data graphic that guide interpretation.
Any three: title, axis labels, legend, annotations, source note, colour.
More practice: warm-up
1What does per capita mean?
2How many times more thefts does City A have than Town B in total?
3A poll reports 52% ± 4. What is the range?
4Halfway between 10 and 100 on a log axis is about what value?
5What is an annotation on a chart?
Text added to point out a feature or tell readers how to interpret it.
More practice: on your own
1Region X has 2,400 cases and 800,000 people; Region Y has 300 cases and 60,000 people. Which has the higher rate per 1,000?
X: 2,400 ÷ 800,000 × 1,000 = 3.0. Y: 300 ÷ 60,000 × 1,000 = 5.0. Region Y has the higher rate, despite fewer cases.
2A poll shows Candidate P 38% and Candidate Q 31%, ±3. Is the lead meaningful? Explain.
P's range is 35 to 41, Q's is 28 to 34. The ranges do not overlap, so P's lead is larger than the margin of error and is likely real (at the poll's confidence level).
3A chart on a log scale shows a line flattening. Does that mean the numbers stopped growing?
No. Flattening on a log scale means the growth rate is slowing; the numbers may still be increasing, just by a smaller percentage each period.
4A map of Canada shades each province and territory by the share of residents who voted for an option. Why might it mislead?
The territories and northern provinces cover huge areas with small populations, so their colours dominate the map, while a small, populous region's colour barely shows. The map suggests the vote by land, not by people.
5Explain how a chart title can make a neutral graphic persuasive.
The title tells readers what to see before they read the data (Crisis! versus Steady numbers), so it frames the interpretation; annotations and colour do the same.
Challenge
1Design a graphic that fairly answers both where is the workload? and where is the risk? for the theft data. Describe it.
Sample: a two-panel graphic (small multiples) with matching style: left panel total thefts (900 and 120), right panel thefts per 1,000 (3.0 and 6.0), both axes starting at zero, a subtitle giving populations (300,000 and 20,000), a neutral title such as Bike thefts: totals and rates, City A and Town B, and a source note.
2A poll of 400 people has a margin of error of about ±5. Why is it larger than a poll of 1,000, and what does that mean for readers?
Smaller samples vary more from the true population value, so the range of likely values is wider. Readers should treat differences under about 5 points as unclear, and be cautious with headlines about small shifts in small polls.
3Write a short evaluation of this claim: The graph is on a log scale, so it is trying to trick you.
Sample: the claim is unfair. Log scales are a standard, honest tool for data spanning many orders of magnitude, and they show growth rates clearly. They can mislead readers who do not notice the scale, so a responsible graphic labels the axis clearly and may add an annotation. The fault, if any, lies in poor labelling, not in the scale itself.
Quiz
1. A town of 20,000 has 120 bike thefts. The rate per 1,000 residents is...
2. A poll shows Yes 46%, No 42%, ±3 points. The best conclusion is...
3. On a log axis marked 1, 10, 100, 1,000, the distance from 10 to 100 equals the distance from...
4. Why can a choropleth map of Canada mislead?
5. A chart of total thefts titled Bike Theft Capital overclaims because...
6. A straight line rising on a log-scale chart shows...
What this lesson teaches: ENG4U Reading and Literature Studies 2.2 (text features of graphic texts: titles, axes, legends, annotations), 1.6 (analysing how elements of a graphic combine to communicate meaning) and 1.7 (evaluating whether a graphic supports its claim).
How to help at home:
- When the news shows a chart or poll, ask: Totals or rates? What's the margin of error?
- Work through one rate calculation together; it takes a minute and sticks.
- Look at a map of election or survey results and ask whether it shows land or people.
Common difficulties:
- Reading log axes as ordinary axes. Point out the ×10 steps.
- Treating small poll differences as real. Compare with the margin of error.
- Ignoring the words. Titles and annotations are where graphics argue.
Facts used: all theft and poll data are invented. A poll of about 1,000 people typically has a margin of error of about ±3 points, 19 times out of 20; the moment magnitude scale for earthquakes is logarithmic.