Missing data

Missing data is information that is unavailable, unrecorded, or absent from the dataset being used. It can involve one field, an entire record, or a reporting period. A blank or null value does not automatically mean zero, no event, no injury, or no action.

Absence can have different meanings

A value may be unknown to the agency, omitted from an export, withheld for privacy, or not applicable to the event. A technical error or delayed submission can also create a gap. These causes have different implications, even when the public file displays the same empty cell.

A data dictionary may distinguish an explicit “unknown” code from a blank that means the field does not apply. The FBI's NIBRS manual includes field-specific rules of that kind. It also defines a zero report for a month with no reportable activity, separating an affirmative zero from no information received. FBI manual, data elements and section 3.3.

At the dataset level, BJS's LEARCAT documentation shows why an agency's presence does not establish twelve months of records. Missing periods can affect every statistic built from the file. LEARCAT reporting coverage.

Example

Hypothetical example: A response-time file contains 100 dispatched incidents. Ninety have valid dispatch and arrival timestamps; ten have no arrival value. The analyst can calculate durations for 90 records but cannot infer zero-minute responses for the other ten.

Suppose 72 of the 90 valid responses met a stated target. Among measurable responses, the percentage is 72 ÷ 90 × 100 = 80 percent. Calling it “80 percent of all dispatched incidents” would conceal the ten unknown outcomes.

The analyst should report the included count and missing count, then investigate whether the ten records were canceled, never arrived, or simply incomplete. Those explanations could require different eligibility decisions. Deleting them without explanation hides a consequential choice.

Common confusion

Missing versus zero: Zero is a known numerical value under the source's rules. Missing means that the value is not available. Replacing every blank with zero can lower averages or invent an absence of activity.

Missing versus not applicable: A field about transported patients may not apply to an incident with no transport. That differs from a transport record whose destination is unknown.

Missing versus suppressed: Suppression intentionally withholds a value, often for privacy or reliability. The underlying value may exist even though it is not published. A suppressed cell should not be interpreted as low activity unless the source says so.

How missingness changes interpretation

A statistic based only on complete records may describe a different group from all eligible events. If records with longer or more complicated responses are more likely to be incomplete, the available-data result can look better than the full set would.

Do not assume the missing records resemble the observed ones. If an estimate fills gaps, identify the method and distinguish estimated values from reported observations. A source's imputed figure is not an additional event discovered in the records.

For a public table, preserve documented missing-value labels and explain the denominator used in each percentage. For a trend, check whether the amount or cause of missing data changed over time. A more complete release can show a higher total without a corresponding increase in events.

The goal is to make the limits visible while retaining useful information. “Ninety measurable responses; ten missing arrival times” is clearer and more honest than presenting a complete-looking average from an incomplete file.

Related terms

• Not applicable
• Data suppression
• Zero report
• Data completeness

Sources

• FBI — NIBRS User Manual, version 2025.0, June 23, 2025, section 3.3 and field-specific data-element instructions. Distinguishes explicit zero reporting and context-dependent values.
• Bureau of Justice Statistics — LEARCAT Methodology, dated October 2, 2024. Documents reporting-month and population-coverage limitations.