I'm republishing this blog post that I originally published in June 2016. IRS migration data continues to be misinterpreted and assumptions made that are not entirely accurate as seen in today's Hartford Courant editorial section. This content is still relevant and worth republishing.

Earlier this week, I received an email from The Yankee Institute with the subject: “Leaving Connecticut” and I let out a sigh. The flurry of data around migration patterns in Connecticut, the conflicting messages about whether the state is losing or gaining population is confusing – and I’m someone who looks at data frequently! I thought about having a session at our next conference (June 28th) on the different sources of migration data and what the data actually say but the agenda filled up with requests from our users and my riveting idea didn’t make it to the top.

However, when I saw this email and then the additional claim that $3.8 billion left with the 27,000 people I decided it was time to shed some light on this data. Although the infographic in the newsletter does not mention the data source, when I went to the report it was as I suspected – the IRS Migration files.

I’ve looked at this data and I too was tempted to use it the way they aggregated and summarized the data. However, there are a few problems with the conclusions they drew from this particular dataset.

About the data

These data released by the IRS look at year-to-year address changes reported on individual income tax returns. These data track state and county level in and out flows of tax returns along with providing aggregate Adjusted Gross Income (AGI). A net-migration number for states and counties can be calculated looking at the difference between the inflows and outflows.

However the problem arises when the AGI is used to calculate income losses or gains. These back-of-the-envelope calculations can lead to inaccurate claims about what is happening and what these data actually tell us.

3 Reasons you cannot assume Adjusted Gross Income (AGI) follows a person

  1. Most people cannot take their income with them – when people move due to a change in employment, the job (and wages) remain in the stateand just the person moves. The job then gets filled by another person and the wages remain in the state – therefore not necessarily a loss of overall AGI to the state.

  2. Some income may transfer to other people in the state – for example, if a small business owner leaves Connecticut, another small business owner in the state may gain customers – increasing income within the state that won’t be captured in the IRS migration data.

  3. Not all income is lost if a person moves – some people move their physical location but some people leave a state but continue to work there. Yes it is true that some income does follow a person such as –Social Security, investment earnings, and pensions.

The income that is ‘lost’ when people move typically goes to two groups

  • People moving into the state replace the majority of the income of people moving out
  • People entering the state’s labor force receive most of the rest of the income previously earned by those who left - these people aren’t necessarily migrating

Yes, Connecticut data from the IRS files, show an overall net out-migration of households


An analysis by the Center on Budget and Policy Priorities1 that examined the national data from 1993-2011, show that for every state that experienced net out migration, those states also experienced an increase in the total number of people employed (measured by an increase in new non-farm jobs). For CT, although it lost 135,000 households, it also filled the jobs of those who left2 and generated 94,000 new jobs.

Another Problem: Cannot compare to previous years because the Data Changed

When you go to the IRS website, it explicitly states that AGI data prior to 2011-2012 cannot be compared to earlier years. Beginning with data for 2011–2012, the IRS introduced a number of enhancements to improve the data’s overall quality, as well as provide a new series of information, therefore these data are not comparable to prior years. Under the new methodology, full-year data are now available. Previously only individual tax returns filed before the end of September were included.

As a result of the changes that were made in their methodology, it now includes late filers. And as it turns out, late filers generally have higher incomes thus it is not surprising that the overall reported AGI increased.

Other Limitations of the IRS Migration Data

Assuming all income is ‘lost’ when people move also exaggerates the actual amount

  • Counting all the income of people who are retiring as ‘lost income’ is inaccurate because their wages would be ‘lost’ even if they didn’t move
  • When people leave a state after being laid off, the state economy does not ‘lose income’ because they migrated; they migrated because they and the state lost income.
  • There is no requirement that the address on a federal tax return be that of the taxpayer’s legal sate of residence under state law. However, this dataset counts a change of address as a move.

Census Data does not support the claim of ‘lost’ income

What’s also interesting – when I looked at the distribution of household income as reported by the Census, I actually found that the percent of households in the highest income bracket increased as a percentage of all households in Connecticut – challenging the idea that the wealthy are leaving.

Statewide Income Distribution, 2000 to 2014

Out migration is a concern. However, studies have shown that the vast majority of moves occur due to either family-related reasons or a new job, job transfer, or lost job. Therefore, it behooves policymakers to take a holistic approach when considering the implications of a net outflow and create an economy that provides sufficient opportunities for residents to find employment and support the public goods that people seek out such as -- quality schools, safe roads, parks, mass transit, and other critical services.

  1. Mazerov, Matthew, “State “Income Migration” Claims Are Deeply Flawed,” Center on Budget and Policy Priorities, October 20, 2014.
  2. It is not necessarily true that jobs were filled on a one-to-one basis since economies are dynamic and jobs may be eliminated. But what is true is that Connecticut created a sufficient number of jobs between 1993-2011 to replace the jobs of the people who left and to employ an additional 94,000 people.