Start with what each number measures
Political data tables often put counts and percentages side by side. A count is a number of cases: people surveyed, ballots recorded, proposals introduced, or responses in a category. A percentage expresses a count as a share of a larger group. Neither is inherently more useful. The right measure depends on the question you want to answer.
Imagine a survey table with two districts. In District A, 30 of 120 respondents support a proposal. In District B, 40 of 80 respondents support it. District B has both a higher count of supporters and a higher share: 40 out of 80 is 50%, while 30 out of 120 is 25%. But those two statements answer different questions. The counts describe how many supporters were recorded in each district’s survey responses; the percentages describe how common support was among those respondents.
Find the denominator before comparing percentages
The denominator is the group used as the basis for a percentage. To find it, ask: “Percent of what?” A table might show the percentage of all residents, registered voters, people who voted, or people who answered a particular survey question. Those groups are not interchangeable. A heading such as “percent supporting the proposal” is incomplete unless the table makes clear whose support is being measured.
Using the district example, 30 supporters in District A represent 25% of its 120 respondents. The same 30 people would be 15% of all 200 respondents across both districts. Both calculations are correct, but they describe different groups. If the table does not identify its denominator, look for a note, a column labeled “total,” or a description of how the data were collected. If you still cannot tell, avoid drawing a firm conclusion from the percentage.
Pay attention to missing responses. If 100 people took a survey but only 90 answered a particular question, a percentage for that question might use 90 rather than 100 as its denominator. Some tables exclude “don’t know” responses; others include them. Checking this detail can explain why two tables reporting the same topic appear to disagree.
Match the comparison to the question
Counts are useful when the question concerns volume. An office planning how many printed forms to provide might need to know how many people requested them. Percentages are useful when comparing groups of different sizes. If one district has many more respondents than another, its count may be larger even when the behavior in question is less common there.
Consider a second example: 60 of 300 respondents in one group and 30 of 100 in another choose the same answer. The first group contributes more answers in absolute terms, but its share is 20%, compared with 30% in the second group. A statement that the answer was “more common” needs to specify whether it means more responses or a greater share of respondents.
Also distinguish a change in percentage points from a percent change. If support rises from 20% to 30%, that is an increase of 10 percentage points. Relative to the original 20%, it is a 50% increase. Calling both simply “a 10% increase” would obscure the comparison. When reading a table across time or groups, check which measure the accompanying text uses.
Check whether categories and totals fit together
A row of percentages may add to 100% when every case belongs to exactly one category and all categories are shown. But a survey question that allows multiple answers can produce percentages that total more than 100%. Percentages can also fall slightly short of or exceed 100% because of rounding. Neither result automatically signals an error; the table’s notes should explain how responses were classified.
Totals can hide differences between groups. In the first example, 70 of the 200 respondents across both districts support the proposal, giving an overall share of 35%. That overall figure does not describe either district particularly well: their shares are 25% and 50%. Before treating a total as representative, look for meaningful breakdowns, such as district or survey group, and check the size of each group.
Read the notes before drawing a conclusion
A percentage can be calculated correctly and still support only a narrow claim. A survey percentage describes the people represented by the survey, not necessarily everyone in the community. Selection, nonresponse, small group sizes, and question wording may affect how confidently a result can be generalized. A table may also report weighted percentages, adjusted to reflect a target population, alongside unweighted counts of respondents. In that case, dividing the displayed count by the displayed total may not reproduce the displayed percentage.
A useful final check is to put any conclusion into a full sentence: “Among the people who answered this question, this many chose this option, representing this share of that group.” If the table does not provide enough information to complete that sentence, say what is unknown. Clear denominators and modest conclusions make political data easier to understand without asking a table to prove more than it can.
