Party Identification and the Leaner Problem

The test

The measure is party identification from the ANES Time Series Cumulative Data File, 1952–2024, obtained from the American National Election Studies. The question has been asked in the same form in every study since 1952 — 32 studies, every presidential year and most midterms. The ANES chapter introduces the file and how to read it against its codebook; this brief pulls one question out of it.

This source fits the question because no other kind of source can hold it. Partisanship is not an administrative fact: nobody registers as a “weak Democrat,” and in most states nobody registers as anything at all. It exists only as something people say about themselves, which means it exists only in surveys. And unlike a poll about a vote, no election ever comes along to grade the answer, so the survey’s design decisions are not a layer over the measurement — they are the measurement. A seventy-two-year series on one question exists only because ANES staff kept asking the same words, decade after decade, and reconciled the coding so that one column means one thing throughout.

The test comes in two parts. First, count the independents under each defensible reading of the question and see how far apart the counts land. Second, use the fact that the same file records how each person voted, so the disputed group can be watched rather than argued about.

The data

The cumulative file is one table: one row per interview, with every question that survived reconciliation carried as a column labelled with a VCF number, the file’s own name for a question. This brief reads three of those columns, plus one it deliberately leaves alone. VCF0301 is the seven-point party identification. VCF0303 is the file’s own three-category collapse of it. VCF0704 is the two-party presidential vote, coded 1 for the Democrat and 2 for the Republican. VCF0009z is the weight, which counts some answers more than others to make the sample look like the country, and it sits in the file unused here.

VCF0301 has seven values and looks like a seven-point scale. It is not a scale. It is the record of a branching interview.

VCF0301 looks like one column. It is three questions.

  Q1  Generally speaking, do you usually think of yourself as a
      Republican, a Democrat, an Independent, or what?

  Q2  (if Republican or Democrat) Would you call yourself a STRONG
      [R/D] or a NOT VERY STRONG [R/D]?

  Q3  (if Independent or other) Do you think of yourself as CLOSER
      to the Republican or Democratic party?

The seven points are the answers combined:

  1 strong Dem      2 weak Dem      3 independent-Dem   (leaner)
  4 independent-independent
  5 independent-Rep (leaner)        6 weak Rep          7 strong Rep

Five random rows as they arrive. VCF0303 is the file's own collapse.

year  VCF0301  means  VCF0303
1972  5  Independent-Republican  3
1974  4  Independent-Independent  2
1986  2  Weak Democrat  1
1996  6  Weak Republican  3
2016  1  Strong Democrat  1

Nothing in the seven numbers says that 3 and 5 answered 'Independent'
to the first question. The codebook says it. The data does not.

Follow two interviews through it. One respondent says Democrat to the first question and strong to the second, and is coded 1. Another says Independent to the first question, so is never asked the second; asked the third, she says she feels closer to the Democrats, and is coded 3. The two sit in the same column two numbers apart, and nothing in the column records that the second never called herself a Democrat.

The people in categories 3 and 5 answered “Independent” to the first question. They named a party only when asked a second time whether they felt closer to one. They are called leaners, and whether they belong in the independent column is the entire subject of this brief.

Notice what the data can and cannot tell you. The seven numbers are just numbers; nothing in them records that 3 and 5 came from a different branch of the interview than 2 and 6. That fact lives in the codebook. A file that contained only the column would be unusable for the question the column is mostly used to answer.

The neighbouring column has already taken a side. The codebook note for VCF0303, the tidy three-category version, reads in full:

Collapsed from VCF0301, 1-3=1, 4=2, 5-7=3
  1. Democrats (including leaners)
  2. Independents
  3. Republicans (including leaners)

In the convenience column, leaners are partisans. That decision travels silently into every analysis that reaches for the tidier column.

Four small tables are built from these columns, and the cleaning is minimal: code 0 in VCF0301, meaning don’t know, not asked, or a minor party, is dropped. sevenpoint.csv gives the share of each study year in each of the seven categories. independents.csv counts the independents three defensible ways per year, with the spread between the widest and narrowest. leaners.csv gives each category’s presidential vote, pooled across every year both questions were asked. collapse.csv records where VCF0303 files each category. Every percentage describes the people ANES interviewed, not the country, because the weight is left unused — a choice this brief returns to below.

Before you read on, commit to two numbers. In the 2024 study, what share of respondents do you think said the word independent at the first question — and what share claimed no party at all, even when pressed? Write both down.

What it says about democracy

Here is the share of independents in every study year, counted first as only those who claimed no party at all, and then as everyone who said the word independent.

Figure 1. The share of independents, 1952–2024, under two definitions. Two lines on a time axis, because the claim is a claim about change and the argument is the distance between the lines. The shaded band is that disagreement: every point inside it is a person whose classification depends on a decision nobody states. Hovering breaks the band into leaning Democrats, pure independents and leaning Republicans. In 2024 the two definitions differ by 26.1 points.

Year	Respondents	No party at all	Anyone who said independent	Gap
1952	1,689	4.9%	22.7%	17.8 pts
1972	2,695	14.8%	36.4%	21.6 pts
1992	2,470	12.8%	39.4%	26.6 pts
2012	5,890	13.4%	36.5%	23.1 pts
2024	5,483	6.9%	33.0%	26.1 pts

The two lines tell different stories. The upper one shows a country that became markedly less partisan between the 1950s and the 1970s and has stayed there. The lower one shows independents as a small group that grew, peaked, and has recently shrunk — in 2024 it is 6.9%, lower than at any point since the 1960s. Neither line is wrong. They are answers to different questions that are almost always asked in the same words.

The decision between them is not arbitrary, because leaners can be watched rather than argued about. Here is how each of the seven groups voted for president, across every year in which both questions were asked.

Category	Respondents	Voted Democratic	Loyalty to nearer party
Strong Democrat	8,556	95.3%	95.3%
Weak Democrat	5,531	78.4%	78.4%
Independent-Democrat	3,815	87.4%	87.4%
Independent-Independent	2,368	47.7%	52.3%
Independent-Republican	3,586	13.7%	86.3%
Weak Republican	4,240	16.1%	83.9%
Strong Republican	6,198	2.9%	97.1%

Figure 2. Two measures of the same seven groups. The leaners are the third and fifth bars; if they were independents they would look like the middle one. The first button shows the direction of the vote, which makes a staircase. The second folds both sides together into loyalty to whichever party the person leans toward, and on that measure the leaners sit above the weak partisans on both sides. Hovering gives both numbers at once.

Leaning Democrats vote Democratic 87.4% of the time. Weak Democrats vote Democratic 78.4% of the time. The same holds on the other side: leaning Republicans are loyal 86.3% of the time against 83.9% for weak Republicans. People who decline to call themselves partisans are more reliable partisan voters than people who accept the label. Whatever the leaners are, they are not undecided, and they are not up for grabs.

And look at the middle bar. Respondents who claimed no party at all split 47.7% to 52.3% — the only group in the file that behaves the way the word independent is supposed to describe. That group is 6.9% of 2024 respondents, not a third.

Now put the convenience column beside a do-it-yourself count.

Seven point category	Share	The file files them under
Strong Democrat	20.8%	Democrat
Weak Democrat	18.5%	Democrat
Independent-Democrat	11.9%	Democrat
Independent-Independent	11.6%	Independent
Independent-Republican	10.4%	Republican
Weak Republican	13.0%	Republican
Strong Republican	13.8%	Republican

Reach for VCF0303 because it is tidier, and you will report an independent share of 11.6%. Do the collapse yourself from the seven-point column, keeping everyone who said the word, and you may report 33.9%. The difference is not an error in either direction. It is a definition, made by someone else, in a note.

That matters for how the country gets described. The rise of the independent voter is one of the standard stories about American democracy, and most of it is a story about a label.¹ The electorate behaves far more partisan than it sounds: the genuinely unattached are a small and shrinking group, and everyone quoting one independent share has taken a side in a definition fight without saying so. The number is not uncertain and the data is not dirty. The gap opens because an ordinary English word, independent, has no ordinary-English answer in the file — and the file supplies one anyway, to anybody who does not look.

What this brief cannot tell you

The vote comparison pools every study year, so it shows that leaners have behaved like partisans on average since 1952, not that they have always done so or do so equally today. Party loyalty has risen across all seven groups; a year-by-year version would show the leaners’ loyalty rising with everyone else’s.

Nothing here is weighted. VCF0009z is in the file and deliberately unused, so every percentage describes the people ANES interviewed rather than the country. For comparisons between groups inside one survey this matters less than it would for a level, but it is a choice, and it was made.

The brief takes no position on what partisanship is — whether it is a running tally of evaluations or a social identity acquired early and rarely revised.² That argument is real and long, and none of it can be settled by noticing that a column has three questions in it.

Finally: leaners voting like partisans is evidence that they are partisans, not proof. A person can vote for the same party every time and still refuse the label, and refusing the label may be the more interesting fact about them. The data can tell you what they did. It cannot tell you what to call them.

What you should have learned

• The share of independents in the 2024 ANES is 6.9% or 33.0%, depending on one coding decision about leaners that headline numbers never state.
• VCF0301 is not a scale but the record of a branching interview, and the fact that categories 3 and 5 came from a different branch lives in the codebook, not the data.
• Leaners vote like partisans: leaning Democrats are loyal 87.4% of the time against 78.4% for weak Democrats, and leaning Republicans 86.3% against 83.9%.
• The only group that behaves like the word independent — splitting 47.7% to 52.3% — claimed no party at all, and it is 6.9% of the 2024 study.
• The file’s own convenience column, VCF0303, resolves the leaner question silently; using it instead of your own collapse moves the reported share from 33.9% to 11.6%.

Extensions

None of these is answered above, and every one can be answered from the tables the Sources section hands over.

• independents.csv reports the share of independents three ways for every year, plus the spread between them. Sort by spread and find when the definitions disagreed most. Pick the coding you would use, then argue for it.
• leaners.csv gives each category a pct_voted_democratic and a loyalty. Compare the leaners against the weak partisans on both columns. Do leaners behave like independents or like partisans?
• sevenpoint.csv keeps all seven categories by year. Follow pure_ind across the whole series. Is the rise of the independent voter visible in this column at all?
• collapse.csv shows where the file itself puts each category, with each one’s share. Find the decision you disagree with, and work out how much the headline number moves if you change it.

Two stretch ideas point past the spreadsheet. Gallup publishes its own party-affiliation series (https://news.gallup.com/poll/15370/party-affiliation.aspx), asked of adults rather than of ANES respondents; find its latest independent share and decide which of Figure 1’s two lines it is nearer to, and why. And in the states that register voters by party, registration files count the unaffiliated directly — compare a state’s unaffiliated share against both lines and ask which definition registration is closest to.

Sources

Data

• American National Election Studies. ANES Time Series Cumulative Data File (1948–2024), version of 5 February 2026. Ann Arbor, MI and Palo Alto, CA: University of Michigan and Stanford University. https://electionstudies.org/data-center/anes-time-series-cumulative-data-file/
• ANES. Time Series Cumulative Data File Codebook, same release. The branching wording of VCF0301 and the collapse note for VCF0303 are quoted from it directly.

Academic literature

• Campbell, Angus, Philip E. Converse, Warren E. Miller and Donald E. Stokes (1960). The American Voter. John Wiley & Sons. The seven-point measure begins here.
• Keith, Bruce E., David B. Magleby, Candice J. Nelson, Elizabeth Orr, Mark C. Westlye and Raymond E. Wolfinger (1992). The Myth of the Independent Voter. University of California Press. The book-length version of Figure 2. https://www.ucpress.edu/book/9780520077201/the-myth-of-the-independent-voter
• Petrocik, John R. (2009). “Measuring party support: Leaners are not independents.” Electoral Studies 28(4): 562–572. https://doi.org/10.1016/j.electstud.2009.05.022
• Fiorina, Morris P. (1981). Retrospective Voting in American National Elections. Yale University Press. Partisanship as a running tally. https://yalebooks.yale.edu/book/9780300025576/retrospective-voting-in-american-national-elections/
• Green, Donald, Bradley Palmquist and Eric Schickler (2002). Partisan Hearts and Minds. Yale University Press. Partisanship as a social identity. https://yalebooks.yale.edu/book/9780300101560/partisan-hearts-and-minds/
• Gallup. Party Affiliation, trend since 2004. https://news.gallup.com/poll/15370/party-affiliation.aspx

The data itself

Every figure above is drawn from these tables. They are plain CSV, they open in a spreadsheet, and they are yours to take further.

collapse.csv, independents.csv, leaners.csv, sevenpoint.csv