Intrinsic Value Per Share IVPS

A reader asked what a car actually earns, and it reorders the whole group

The cheapest of these dealers on reported earnings is third cheapest once the windfall comes out, and the dearest reported is not the dearest normalised. The ranking a screen gives you is not the ranking.

Ten years of gross profit per vehicle, read off the audited income statements, and what the dealers' single-digit multiples become once the pandemic windfall comes out of the denominator.

researchauto-retail

A reader wrote in after the first auto retail post with a suggestion: compare new and used units and gross profit per unit across the dealers, because new-vehicle units also tell you something about market share. It was the right question and I had not asked it. Ten years of it now sit in the auto retail study, pulled off the audited income statements for the five franchised dealers that publish the parts, and it changed my conclusion rather than decorating it. Thank you — the rest of this post is what your question turned up.

The short version: the windfall is measurable, it is not evenly held, and once you take it out the group reorders. Whichever of the two defensible normalisations you prefer, you get the same new ranking, and it is not the one a screen reports.

What the measurement is

For each of the five, gross profit per new vehicle retailed, every year from FY2016 to FY2025. Revenue less cost of sales by vehicle line, divided by that line's units, with each company's own printed total gross profit as the test that the lines add up. Where a filing does not publish cost of sales by line, there is no figure: Penske Automotive does not, in any of the ten years, so it is absent rather than estimated.

All five trace the same arc. About $1,857 a car through FY2016–2019, a peak in 2022 at every one of them, and a slide since that still leaves the group at 1.7× the old figure.

What is still to give back, and who is holding it

Take each company's FY2025 gross profit per new unit, subtract its own FY2016–2019 average — they did not start level, so each is judged against itself — and multiply by the new units it sells today. That is the gross profit at risk if the new-vehicle line returns to where it was. Set it against gross profit less selling costs and you get the share of earnings resting on the unwind.

It runs from 13% to 32%. That is a spread of more than two times inside a price band about a turn wide, and it is the finding the reader's question produced.

AutoNation has already surrendered 81% of its climb from pre-pandemic to peak. Asbury Automotive has surrendered 55%. The market prices them at 8.0× and 7.6× trailing earnings.

The multiples, struck twice

Normalising needs a view of what normal is, and there are two defensible answers that give very different numbers. I carry both rather than choosing.

On the dollar basis, gross profit per new vehicle returns to each company's own FY2016–2019 figure in dollars. On the margin basis it returns to its own FY2016–2019 figure as a share of the selling price, which is a far smaller give-back because average new selling price rose 44% over the window, from $35,819 to $51,630. In the unit a dealer manages, the first asks for $1,230 a car back and the second $508.

Price on 30 October 2026 over diluted earnings per share. Normalised columns deduct new-vehicle gross profit above each company’s own pre-pandemic baseline, taxed at 25%.
Retailer Reported On dollars On margin Exposure
AN 8.0× 10.2× 8.2× 13%
LAD 8.2× 12.1× 8.9× 19%
ABG 7.6× 15.8× 10.8× 30%
SAH 9.7× 20.5× 11.3× 21%
GPI 8.2× 26.4× 13.6× 32%

Rows are in normalised order, cheapest first.

How it changes the relative value

This is the part I did not expect. The dollar basis and the margin basis disagree about the size of the give-back by about 6.4 turns on average. They do not disagree about the order. Both rank the five identically — AN, LAD, ABG, SAH, GPI, cheapest to dearest. So the ranking does not depend on which basis you believe, which is a sturdier conclusion than either column on its own.

The reported ranking is ABG, AN, GPI, LAD, SAH. Compare the two and the reordering is the whole story:

  • Asbury Automotive is the cheapest of the five on reported earnings at 7.6×, and third cheapest once the windfall comes out, at 15.8× on dollars. It has given back 55% of its climb and carries 30% exposure, behind only Group 1 Automotive. The discount is real and so is the reason for it.
  • AutoNation is the cheapest on both normalisations, at 10.2× and 8.2×, while sitting at 8.0× reported — neither the cheapest nor the dearest reported figure in the set. It has taken 81% of the pain already, carries 13% exposure, and spent the windfall retiring stock rather than buying stores. On this measure it is the one that looks mispriced.
  • Group 1 Automotive is the inverse: mid-pack on reported earnings at 8.2×, and the dearest normalised figure of the five at 26.4×. It carries 32% of gross profit less selling costs resting on the unwind, $346 M of gross profit above its own baseline. Nothing in its reported multiple says so.
  • Sonic Automotive carries the dearest reported multiple of the five at 9.7× and is fourth, not fifth, on both normalisations. The name that looks the most expensive in this group is not the one with the most to lose.

The hedge I expected to find, and did not

The intuitive protection is the workshop: a dealer whose gross profit leans on parts and service should care less what a new car earns. This set contains something close to a controlled test, and the test fails. Asbury Automotive and AutoNation draw effectively the same share of gross profit from parts and service — 47.9% against 47.6% — and their exposure differs by more than two times, 30% against 13%.

The reason is that the mix is a mirror of the cycle rather than a defence against it. Parts and service did not shrink in 2021 and 2022; new-vehicle gross profit doubled and made everything else look smaller. As it unwinds the workshop's share climbs back, so a parts share level with FY2019 does not mean the exposure has gone.

One qualification, because it cuts the other way. In dollars rather than as a share of the mix, the annuity has grown at every one of the five — from 1.5× its own pre-pandemic average at Sonic Automotive to 4.3× at Lithia Motors, which bought most of its. A bigger workshop on the same margin does protect earnings. It does not protect them in proportion to its share of gross profit, which is the claim being tested. Anyone underwriting the annuity should underwrite its dollar growth rate, not its percentage of the mix.

Grading the last post

The 3 November post led with the dealers being "every one of them on a single-digit price-earnings ratio, and all of them within about a turn of each other." Accurate as reported, and the wrong thing to lead with.

The multiples were single-digit, on all five of them. I did not ask what the denominator contained. On the dollar basis not one of them keeps a single-digit multiple — the cheapest is 10.2× — and on the gentler margin basis two do.

"Within about a turn of each other" was true on enterprise value to EBITDA, where the five still sit 1.2 of a turn apart, and I let it stand for how alike these companies are. Once the windfall comes out of the earnings they are spread over 16 turns. The tight band I reported as a finding was a property of the measure, not of the businesses.

What I am not claiming

Not that gross profit per new vehicle reverts to its FY2016–2019 dollars. Not that these are bad businesses — the five have compounded earnings per share at 14% to 22% a year over seven years and the franchise annuity under them is real. Not that anyone should sell anything.

The claim is narrower. The reported multiple is struck on an earnings base containing a known, measurable, temporary component; that component differs by more than two times across companies the market prices within a turn of each other; and the ordering you get after removing it is stable across both ways of removing it.

The best argument against all of it is inflation, and it is a good one. A dealer earning the same percentage on a car that costs 44% more earns more dollars without earning a better margin. Gross profit per new vehicle is 1.7× its pre-pandemic level in dollars and only 1.2× as a share of the selling price. That case gets its own section in the study, including the part of it that cuts against itself: through the baseline window the margin was already falling, from 5.29% of the selling price in FY2016 to 4.77% in FY2019.


So: does gross profit per new vehicle settle nearer $1,857 a car, or nearer 5.02% of a $51,630 car? The first puts these five at 10.2× to 26.4× earnings. The second puts them at 8.2× to 13.6×. I lean nearer the second and hold it loosely.

The workings, the charts and the workbook with the baseline years in it are in the auto retail study. Priced at the 30 October 2026 close. Take a look and tell me which one you would underwrite — and where I have this wrong.

← All research notes