Agriculture & Future Food Systems · Article

Questions, Indicators and Causal Claims

Choose a food-system measure that answers a clear question, and distinguish a recorded improvement from evidence about what caused it.

Preparation and related reading

Shared Foundations

This article helps you

Match an indicator to a defined outcome, group and period.
Keep implementation records separate from changes in people's food access.
Calculate an observed change without turning it into an unsupported causal effect.

Start with a question people can answer

A vegetable delivery service adds an extra vehicle. Its report says more orders arrived on time the following week. That is useful information, but several questions remain: did the same customers receive the service, did the weather improve, and did the change make food easier to obtain? A vehicle count, a delivery record and a household experience describe different parts of the system.

An indicator is a defined measure used to track a question. Begin with the intended result, then state the people or products covered, the period and the measurement rule. 'Better distribution' is too broad to count consistently. 'Orders arriving inside the agreed delivery window during this week' can be recorded, provided the window and eligible orders are specified.

Work completed and outcomes achieved differ

Inputs are the resources used, such as staff time, vehicles and funds. Activities are the work performed. An output is the resulting service or product: scheduled deliveries or meals provided. Outcomes concern what happens when people use those outputs. More scheduled trips need not mean reliable delivery, affordable food or an adequate diet.

Keep records along that chain. A project can miss its goal because the service was never delivered, because eligible people could not use it, or because another constraint remained. Counting only the final outcome can hide these differences. The links are proposed mechanisms to investigate, rather than guarantees built into the names.

Describe the change on the same basis

For an invented example, exactly 100 eligible orders are observed in each of two weeks using the same arrival-window rule. On-time orders rise from 90 to 96. The recorded proportions are 90% and 96%, a difference of 6 percentage points. Relative to the original 90%, the increase is 6/90 × 100, approximately 6.7%. These are two descriptions of one recorded contrast.

The counts do not measure a six-point improvement in food security. Nor do they establish the vehicle's effect. Retain the denominator and the definition: a different number of orders or a relaxed arrival window could change the apparent result. Report cancellations, excluded routes and missing observations rather than silently treating them as successful deliveries.

Invented delivery records with one consistent rule
Record	Before	After
Eligible observed orders	100	100
Inside the agreed window	90	96
Observed on-time share	90%	96%

A recorded contrast leaves the cause open

Original hypothetical record; arrows express reasoning,not causal proof or delivery flows. The same denominator and measurement rule are stipulated.

Read the diagram as text

With 100 observed orders in each invented week, 90 and 96 on-time deliveries give a 6-percentage-point contrast. Identifying the cause needs further comparison.

Parts

Before: 90 of 100: 90% inside one agreed arrival window
After: 96 of 100: 96% under the same recorded rule
Observed change: +6 points: A descriptive contrast, not the vehicle's proven effect
Causal contribution unresolved: Weather and other changes need an appropriate comparison

Connections

Before: 90 of 100 → After: 96 of 100: observe later
After: 96 of 100 → Observed change: +6 points: 96 − 90
Observed change: +6 points → Causal contribution unresolved: ask why

A cause requires a comparison

The causal question asks what would have happened to the same relevant outcome without the intervention. That unobserved alternative is called a counterfactual. A before-and-after record supplies a baseline, but the earlier week is not automatically a valid estimate of the later week without the vehicle.

Weather, order mix, road repairs and customer selection might also change. A suitable evaluation could use a planned comparison or another justified design, with assumptions appropriate to its setting. Similar-looking groups are not automatically interchangeable. A precise calculation or a statistically clear contrast cannot repair an inappropriate comparison. Detailed evaluation methods belong in their shared statistical homes.

Keep each measure within its scope

Food-access questions need evidence about access. A national supply account describes available food flows; it does not observe what every person eats. An experience-based survey can investigate difficulty obtaining food, while a process record can show whether deliveries occurred. These sources complement one another when their different units and definitions remain visible.

Select a small set that addresses the actual purpose. Specify the collector, timing and source before the change. Record uncertainty and coverage limitations alongside the value. A missing response is not evidence of success, and no one food-access measure establishes safety, nutritional adequacy, affordability and long-term resilience together.

Write the conclusion the evidence supports

A defensible delivery conclusion is: 'The observed on-time share increased from 90% to 96% under the stated rule; other changes and the added vehicle's causal contribution remain unresolved.' That preserves the improvement while identifying the next useful question. It does not require hiding results until a perfect study exists.

Use the record to improve the next investigation: compare comparable periods, retain service failures and ask intended users about barriers. Review unexpected burdens, including staff time or higher prices. Shared Foundations explain system boundaries and uncertainty; neighbouring topics develop food access, data sources and comparison across scales.

Worked examples

Percentage points or relative change?

What do the invented 90-of-100 and 96-of-100 delivery records show?

Compute 90/100 = 90% and 96/100 = 96%.
Subtract: 96 − 90 = 6 percentage points.
For relative change, use the original share: 6/90 × 100 ≈ 6.7%.

The observed share rose by 6 percentage points, approximately 6.7% relative to its original value. Neither calculation establishes why it rose.

Check a proposed explanation

The vehicle was added during a week when severe road disruption ended. Does the later record isolate the vehicle's effect?

Identify the intervention and the other change.
Ask how delivery performance would have developed without the vehicle.
Seek an appropriate comparison and document its assumptions.

No. Ending disruption is a plausible competing explanation; the before-and-after record alone cannot separate the contributions.

Check your understanding

Does recording ten additional scheduled trips prove that ten additional households obtained affordable food?
Show answer and reasoning
No.
A schedule is an output record. Actual receipt,people served and affordability need their own observations.
If the delivery window is relaxed between weeks, is the on-time proportion directly comparable under the original rule?
Show answer and reasoning
No.
The outcome definition changed. Retain both definitions or recalculate comparable records where possible.
Should an unknown arrival time be entered as an on-time delivery?
Show answer and reasoning
No.
It is missing evidence; report its coverage separately rather than inventing success.

Key ideas

Define the question,unit,people,period and indicator before interpreting a result.
Implementation outputs and later outcomes describe different stages.
An observed change is not automatically a causal effect.
Keep denominators,definitions,missing records and comparison assumptions visible.

References

Sources supporting the explanations and examples in this article.