A historical dataset is a set of figures gathered at a particular date from a particular population. Using it well means saying what is measured, when, and for whom, before you draw any conclusion. This skill appears whenever a depth-study source includes census figures, survey results, court records or production totals.
It follows defining the scope of a question and belongs to the United States depth-study pathway.
What do you ask of any dataset?
Use five questions in order.
- What is measured? Households owning something, arrests, wages, votes.
- When? The year or period of the count.
- Who or what was counted? The population, and how large it was.
- How big is the effect? Use a percentage or a difference, and keep the raw numbers.
- What can it not show? Reasons, other places, other years and groups not counted.
Worked example
The figures below are invented for practice. The town and surveys do not exist.
Dataset A (invented). Survey of 250 households in one invented town, 1926: 50 households owned a radio.
Dataset B (invented). Survey of 400 households in the same invented town, 1930: 140 households owned a radio.
Question: What can you claim about radio ownership in this town?
Step 1, what is measured: households owning a radio.
Step 2, dates: 1926 and 1930.
Step 3, population: households in one invented town. The surveys have different sizes: 250 and 400.
Step 4, size of the change: because the groups differ in size, use percentages.
- 1926: 50 ÷ 250 = 0.20, which is 20%.
- 1930: 140 ÷ 400 = 0.35, which is 35%.
The share of surveyed households with a radio rose by 15 percentage points. The raw count rose by 90 households, but that comparison alone would mislead because the survey sizes differ.
Step 5, limits: the data does not show why ownership rose, whether the same households were surveyed, or what happened in other towns.
Answer: In this invented town, the share of surveyed households owning a radio rose from 20% in 1926 to 35% in 1930. The data cannot show reasons or national patterns.
The mistake to watch for
A student wrote the following sentence about the dataset.
“By 1930, 35% of Americans owned radios.”
The sentence changes the population and the thing counted. The data covers households in one invented town, not Americans, and it counts ownership by household, not by person. It also uses the word “Americans” as if the town stood for everybody.
Correction: “In one invented town in 1930, 35% of the surveyed households owned a radio.”
Check yourself
1. An invented survey of 300 households in 1928 found that 90 owned a radio. What percentage is that?
Show answer
90 ÷ 300 = 0.30, so 30%. Check: 30% of 300 is 90.
2. In a different invented town, 60 of 240 households owned a radio in 1928. Can you say this town had more radio ownership than the town in question 1?
Show answer
60 ÷ 240 = 0.25, so 25%, which is lower than 30%. In these two invented surveys the first town had the higher share. You still cannot say why, or claim anything about other towns.
3. Name the population and date in this sentence: “A city count of 1,200 court cases in 1925 shows crime was out of control.”
Show answer
Population: court cases in one city, with no comparison year given. Date: 1925. The phrase “out of control” is an opinion, not something the number proves. A second date or a comparison group would be needed before saying anything about change.
Where this leads next
Numbers sit alongside views, so the next lesson is comparing contemporary perspectives with attribution. The integrated practice set mixes data and views.
If tables and percentages make source questions feel harder than they are, our teachers can work through them with you in online one-to-one History tuition.