top of page

Proxy variables: what’s left out of the data can still show up in the results

Writer: Kat Greenbrook
Kat Greenbrook
Nov 9
4 min read

Updated: 11 hours ago

"I don’t see colour." People usually say this with good intentions. They mean they treat everyone the same, regardless of their race or ethnicity.


But even if one person treats everyone the same, the systems around them, like housing and the law, may not. Many of today's gaps between ethnic groups come from past decisions that are still at work without anyone lifting a finger. Their effects are everywhere, including in data.


If you leave variables like ethnicity out of your analysis, those effects still show up in the results. You just lose the ability to see them.


This happens through proxy variables.



What are proxy variables?


A proxy variable stands in for something you can't measure directly. For example, where someone lives is often used as a proxy for income. 


When it comes to ethnicity, proxies work in two directions.



  1. Ethnicity as a proxy


Ethnicity is about the groups you belong to, through shared culture and ancestry. In New Zealand, ethnicity is meant to be self-reported, and many of us identify with more than one.


When data shows a gap between ethnic groups, like in school results or life expectancy, it's easy to assume the reason sits with the people in each group. But apart from a few health conditions linked to ancestry, that interpretation is usually wrong. Most of these differences come from how systems, like education and health care, have treated each group over generations.


That makes ethnicity a proxy for those systems. The gaps show who the systems work well for, and who they don't.


When these gaps are shared without that context, people commonly put them down to choices or culture. Explaining them this way, while leaving the systems out, is part of what sociologists call colour-blind racism. It lets inequality carry on without anyone being openly racist.



  1. Proxies for ethnicity


The systems that ethnicity is a proxy for also affect other things, like where people live and how much they earn. Because of this, these variables often line up with ethnicity. So they can stand in for it, even when ethnicity isn't in your data.


New Zealand's school deciles are a good example. Until 2023, each school was given a decile based on census data. This data included the incomes, jobs, and qualifications of families in its area. Deciles were used to target extra funding to schools in communities facing more disadvantage.


Ethnicity was part of the decile calculation until 2005, when the government removed it to make sure funding was based on "need, not race". But the remaining variables still gave lower deciles to schools with more Māori and Pacific students. More than a decade later, a report for the Ministry of Education found that the "low decile" label was acting as a proxy for ethnicity.



What proxy variables mean for policy


Nearly twenty years later, the government of the day told public agencies to prioritise services "on the basis of need, not race". It was concerned that agencies might use ethnicity as a proxy for need, and asked them to consider other variables first.


But if other variables, such as deprivation, already stand in for ethnicity, does removing it make a difference?


A 2026 New Zealand study tested this. It looked at deaths from conditions that health care could have treated. Even after accounting for age, rurality, deprivation, income, and existing health conditions, Māori were about 46% more likely to die from them than Europeans, and Pacific peoples about 18% more likely.


In this analysis, ethnicity picked up risk that the other variables missed. So the data doesn't support treating need and ethnicity as separate things. Here, ethnicity was part of how need showed up.


Like deprivation data, ethnicity data tells you about groups. Not everyone in a group shares the same risk, but that's true of every variable used to target services.


Even so, ethnicity doesn't belong in every decision. It depends on what it's a proxy for. Used as a proxy for how systems have treated a group, it can direct support where it's needed and help narrow gaps. Used as a stand-in for stereotyped risk or suspicion, it can widen them. In the 1970s, police and immigration officials targeted Pacific communities in their search for people who had overstayed their visas. Pacific peoples made up about a third of overstayers, but 86% of prosecutions.



What to do about it


  1. Keep ethnicity in your data, even if you don't use it to make decisions. If you don't record it, you can't check whether a decision affects ethnic groups differently. Be careful, though. It can be misused, so be clear about who uses it and why. For Māori data, that includes Māori having a say in how it's used.


  2. Know what ethnicity is standing in for. It might be in your analysis directly, or carried in through variables like where people live. Either way, check how much it's driving your results, and for which groups. Where you can, also measure how systems affect people directly, such as distance to services or experiences of discrimination.


  3. Check whether decisions narrow or widen gaps. Look at how outcomes change for each ethnic group. This applies whether ethnicity is used in the decision, left out of it, or replaced with another variable. As long as ethnicity is still in your data, you can check.


  4. Read and report gaps in context. A gap shows where a system isn't working as well for some groups. Before explaining or sharing it, look at the systems behind it, including how well they fit different cultures, and include them in what you report. Also report what’s working well for each group.



Wanting to stay neutral is a good instinct, and leaving ethnicity out of your analysis can feel like the way to do it. But the effects of how systems have treated different groups are still there. Seeing ethnicity and its proxies in your data is how you check whether those systems treat people fairly.




Published 9 November 2026

I write about data and power, so if you'd like to follow along, the best place to do that is my newsletter. This article is part of the Things We Tell Ourselves series. Each article looks at an everyday saying, what the data shows, and what it means if you tell stories with data. If you think I’ve missed something important, please let me know at kat@roguepenguin.co.nz.


Kat Greenbrook is a data storytelling consultant, author, and workshop facilitator based in Wellington, New Zealand. She is the founder of Rogue Penguin and the author of The Data Storyteller's Handbook.

bottom of page