EDORA
Skip to content
∑

EDORA • Learn

Justice Index · Advocacy Lab · Field Guide

EDORA Learn — Methods

← Back to Learning Center

Series Breaks and Definition Changes

Longitudinal data are rarely seamless. Definitions evolve, record systems migrate, and coding rules shift. These changes create “series breaks” — points where the numbers stop being directly comparable.

What Causes a Series Break?

  • Definition updates: A dataset redefines a variable, such as expanding “custody” to include short-term detention or redefining “referral” to include screenings.
  • System transitions: A new case management or data warehouse platform replaces the old one, changing variable names or record linkage logic.
  • Counting rules: A jurisdiction begins counting petitions instead of adjudications, or adds a new population subgroup.
  • Policy changes: Legislation alters reporting requirements, shifting the mix of included cases.

Recognizing the Signs

Series breaks often appear as sudden jumps or drops that do not match any real-world event. Analysts should cross-reference metadata, policy memos, or archived documentation to confirm whether a definitional change explains the discontinuity.

  • Large, one-year shifts with no underlying program change.
  • Missing or duplicated categories following a new reporting template.
  • Sudden flattening or volatility coinciding with a database migration date.

How to Handle Series Breaks

  1. Annotate clearly: Mark the break year or quarter on every chart and table with a note such as “Definition revised in FY2019.”
  2. Separate series: If definitions diverge too far, start a new series rather than stitching incompatible data together.
  3. Recalculate retrospectively: Where old raw data are available, rebuild earlier series using the new definitions to restore comparability.
  4. Document context: Include a short narrative of what changed, when, and why. Readers should never guess.

Data & Methods

The research text highlights numerous examples where states revised key indicators without updating documentation, making long-term trend analysis unreliable. The preferred practice is to maintain a “change log” — a running record of variable definitions, schema revisions, and date ranges affected. Analysts comparing across states or years should always consult that log before interpreting differences.

Related

Transparency note: Every variable revision should have a date-stamped record and annotation. Apparent trends without accompanying definition notes risk confusing reporting artifacts for genuine change.