Household, Business and National Scales

A total can conceal different experiences

A region reports that vegetable supplies increased. A household without a nearby shop may still struggle to obtain them, while a distributor sees fuller warehouses. Each observation can be accurate. The region's total, the firm's stock and the household's access answer different questions.

Choose the unit before calculating: person, household, order, enterprise or territory. State the population covered and the period. Changing scale often changes the available evidence and the meaning of the outcome, not merely the number of decimal places shown. Shared Foundations on systems and measurement help keep that boundary explicit.

A household is not a single person

Household records can describe purchased food, expenditure or experienced difficulty obtaining food. They do not automatically establish each member's intake. Household sizes and the distribution of food within a home matter. Dividing a household total by its members creates an average, not observations of equal shares.

Experience-based food-access tools distinguish individual and household questions. Keep the respondent, the reference period and the intended unit attached to the result. A defined local observation can be useful without being represented as a calibrated national food-insecurity estimate. Detailed nutritional assessment belongs in its own subject.

Production records describe another boundary

A processor can report saleable output, material losses and electricity within a stated operating period. Those records help assess that enterprise. They need upstream inputs, distribution and people's access to support a wider food-system claim. A low facility cost per kilogram does not directly establish an affordable diet.

National food-supply accounts combine production, trade, stocks and uses. Their per-person supply averages do not reveal every group's diet or seasonal experience. Household surveys and flow accounts can complement one another; they should not be treated as interchangeable measurements of what people ate.

Weight the groups on the intended basis

Consider an invented complete record of 1000 eligible households during one month. In group A, 40 of 100 report a defined difficulty obtaining the planned vegetables. In group B, 90 of 900 report it. The group proportions are 40% and 10%. For these 1000 households, the combined proportion is(40+90)/(100+900)=13%.

Simply averaging 40% and 10% gives 25%, which answers an equal-group weighting question. It does not give the overall household proportion here, because the groups differ in size. The 13% total also hides the much higher proportion in A. Report the group values and counts alongside the combined result. These are fictional direct counts, not a validated food-insecurity scale.

Keep absent observations visible

In a separate invented survey, 100 eligible households are invited, 50 respond and 10 respondents report the same defined event. Among respondents, the proportion is 10/50=20%. If the event is binary and the 50 missing households have unknown answers, the complete invited-group proportion could range from 10/100=10% to 60/100=60%.

This is a logical bound from missing answers, not a statistical confidence interval or an estimate for a larger population. Assuming that nonrespondents match respondents is an additional assumption needing justification. Report response coverage and investigate whether some groups were harder to reach. More digits in 20% would not resolve the missing information.

Use several scales for a better question

A useful review might connect a distributor's vegetable dispatches with shop availability and household access observations during comparable periods. Preserve identities and units at each link. Avoid adding kilograms of products that serve different functions, or interpreting the same shipment as new production each time it changes hands.

Disaggregate where it serves the decision: location, income, household type or season may reveal a constraint that a total hides. Protect participant information and report small-group uncertainty honestly. The aim is useful knowledge about who experiences the system, not a large dashboard that replaces their circumstances with one average.

Worked examples

Check your understanding

1. Does national food supply per person show that each person ate that amount?

   Show answer and reasoning

   No.

   A supply aggregate and individual intake have different measurement boundaries.

2. Can 40% and 10% be averaged unweighted to obtain the overall household proportion when group sizes are 100 and 900?

   Show answer and reasoning

   No.

   The household total is 130/1000=13%; equal-group weighting answers another question.

3. Should a household with no response be coded as experiencing no difficulty?

   Show answer and reasoning

   No.

   Its outcome is unknown. Missing observations need separate coverage and assumptions.

Key ideas

• State whether the unit is a person, household, enterprise or territory.
• Combine counts on the intended denominator before interpreting group rates.
• Report variation and missing groups beside an average.
• Link scales through evidence while retaining their different definitions.