Why do we need record linkage?

In a perfect world

In a perfect world, everyone (and everything) would have a single, unique identifier. If this were the case, linking any datasets would be a simple left join.

Example

Consider 2 tables of people A and B with no duplicates and each person has a unique id UID. Joining these tables in SQL would simple be:

SELECT *
FROM A
LEFT JOIN B
ON A.UID = B.UID

In reality

Real datasets often lack truly unique identifiers (both inside and across datasets).

The overall aim of record linkage is to generate a unique identifier to be used like UID to our "perfect world" scenario.

Record linkage the process of using the information within records to assess whether records refer to the same entity. For example, if records refer to people, factors such as names, date of birth, location etc can be used to link records together.

Record linkage can be done within datasets (deduplication) or between datasets (linkage), or both.