Solar Payback Calculation: Degradation and Escalator

Page four of the proposal, below the savings chart, in eight-point type: Assumes 5.0% annual utility rate escalation and 0.25% annual module degradation. Savings shown in future dollars.

Those two sentences carry three assumptions. On the worked system below, defensible settings for the same three dials produce paybacks anywhere from 8.3 to 11.3 years, and 25-year savings totals from $56,567 to $190,250. Same roof, same hardware, same price. The chart was never the finding. The fine print was.

This is the sensitivity pass on item 7 and item 8 of the twelve a proposal must show. That post listed the dials; this one measures each against a published source and turns each one by hand. The stakes are 2026 stakes: the 30% Residential Clean Energy Credit under 26 U.S.C. §25D applied only to expenditures made through 31 December 2025, so there is no federal credit left to absorb a third of an optimistic assumption.

One formula, three dials

Every row in this post comes out of the same geometric series. The inputs:

C   = cash price                      $24,000  (8.0 kW DC at $3.00/W)
S1  = year-one savings                 $2,016  (11,200 kWh × $0.18)
e   = utility rate escalator           dial 1
d   = annual degradation               dial 2
r   = (1 + e) × (1 − d)                one year's compounding

The $0.18 is deliberately not a national average — it is a placeholder for the marginal rate off your own tariff sheet, which is the second of the four numbers on your bill. Substitute yours before any output below means anything. Cumulative savings after N years, and the payback year solved directly:

total(N) = S1 × (r^N − 1) / (r − 1)

payback  = ln(1 + C × (r − 1) / S1) / ln(r)

Two lines of arithmetic any spreadsheet reproduces. Dial three, inflation, never appears in the formula — which turns out to be the whole problem with it.

Dial 1: degradation — what is signed versus what is measured

Degradation has an unusual property among proposal inputs: somebody underwrites it. The module manufacturer publishes a warranted output floor, and that document is the only degradation figure backed by an obligation.

The floors are specific. JinkoSolar's current limited warranty, revision dated 19 September 2025, prints a table per model: Tiger Neo N-type modules are warranted to lose at most 1.00% in year one and 0.40% per year after, ending at 87.40% of nameplate in year 30; its P-type Tiger Pro line gets 2.00% and 0.45%, ending at 84.95% (JinkoSolar Limited Warranty, REV.EN20250919, read 31 August 2026). REC warrants a minimum of 92% in year 25 on its Alpha series (read 31 August 2026). Two things follow. A proposal modelling 0.25% flat on a module warranted at 1% + 0.40% is more optimistic than the manufacturer's own signature. And the flat rate hides the first-year step that most warranties price in separately.

Measurement runs worse than warranties, which is what you would expect — warranties are floors a manufacturer expects almost never to touch, not forecasts. NREL's 2012 analytical review assembled close to 2,000 measured degradation rates from the literature and found a median of 0.5% per year (Jordan & Kurtz, Photovoltaic Degradation Rates — An Analytical Review). A decade later the PV Fleet Performance Data Initiative analysed 1,700 commercial and utility-scale sites — 7.2 GW, roughly 6–7% of the entire U.S. market — and found a median performance loss of 0.75% per year, ranging from 0.48% in cooler climates to 0.88% in hotter ones (Jordan et al. 2022, Progress in Photovoltaics, read 31 August 2026). Those are professionally maintained plants, not rooftops; the figure includes system-level effects a module warranty never covers, like soiling trends and inverter behaviour. There is no equivalent public dataset for houses, so 0.75% is the honest stress value, not a prediction for your roof.

Turn the dial on the worked system, escalator held at 3.63%: at 0.25% degradation payback is 10.18 years; at the Jinko N-type warranty slope of 0.40% it is 10.24; at the fleet median of 0.75% it is 10.39. Eleven weeks of movement across the whole span, from the proposal's setting to the measured median. Degradation is the dial that gets argued about the most and moves payback the least — its real effect lands in the 25-year total, where 0.25% versus 0.75% is a $5,091 difference.

Dial 2: the escalator — pick a window, get the answer you wanted

The escalator claims to know what your utility will do for 25 years. Nobody knows that, so the honest move is to measure what it did. EIA's Electric Power Monthly, Table 5.3, gives the U.S. average residential price by year: 12.55 cents per kWh in 2016, 13.15 in 2020, 13.66 in 2021, 17.30 in 2025 (EIA Table 5.3, data through June 2026, read 31 August 2026). Three compound annual rates hide inside those four numbers:

2016 → 2025:  (17.30 / 12.55)^(1/9) − 1  =  3.63% / yr
2016 → 2021:  (13.66 / 12.55)^(1/5) − 1  =  1.71% / yr
2020 → 2025:  (17.30 / 13.15)^(1/5) − 1  =  5.64% / yr

Same table, same agency, and a factor of 3.3 between the lowest window and the highest. Fit the escalator to 2020–2025 — the rows nearest the top of the table, and the ones a seller will reach for — and the proposal's 5% looks conservative against 5.64%. Widen the window to the full decade and the same 5% sits a third above the measured 3.63%. Nothing about the utility changed between those two readings. The window was doing the arguing, and whoever picks which five years count as history has already picked the conclusion.

Regulation does not settle it either. California's Solar Consumer Protection Guide allows providers to assume at most 10% annual escalation as of 2025 (read 31 August 2026) — a ceiling on exaggeration, not an endorsement of it. On the worked system, 10% produces an 8.29-year payback and a $190,250 lifetime headline. Both are arithmetically correct and neither describes anything except the assumption. National averages have the same defect in milder form: your escalator should come from your own utility's granted rate cases and your state's row of that EIA table, not from the U.S. line.

Dial 3: inflation — the one that is already inside dial 2

Here is the part proposals skip. The 3.63% measured escalator is a nominal rate; general inflation ran almost as fast. The CPI-U annual average was 240.007 in 2016 and 321.943 in 2025 — the 2025 figure averages eleven months, since BLS never published an October 2025 index (BLS series CUUR0000SA0, read 31 August 2026). That is 3.32% per year. Divide one by the other:

real escalation = 1.0363 / 1.0332 − 1  ≈  0.3% / yr

Electricity, on the national average, got about a third of a percent more expensive per year in real terms across that decade. Nearly the whole escalator was inflation wearing a utility bill as a costume.

Which reframes the fine print's last sentence: savings shown in future dollars. Under the proposal's own dials, the worked system saves $6,123 in year 25. At 3.32% inflation, that is $2,798 of 2026 purchasing power — the document prints the big number and lets you read it as today's money. Redo the whole 25-year sum in constant 2026 dollars (real escalator 0.3%, fleet-median degradation) and the $92,808 headline becomes roughly $47,800. The payback year in dollar-count terms doesn't change, but if you ask when the system has returned $24,000 of purchasing power, the answer stretches to about 12.2 years. Neither figure is wrong. Only one of them is on the page.

Six settings, one system

Every row: 8.0 kW DC, $24,000 cash, 11,200 kWh year one at $0.18. Nominal dollars throughout, so the last column overstates by design — see dial 3.

Read the columns separately. Payback spans three years across the table — material, but survivable. The lifetime total spans 3.4×, and it is the number the chart on page one is drawn from. The dials were never really aimed at the break-even year. They were aimed at the poster.

Your proposal states two of the three settings somewhere, or is obliged to tell you if asked in writing. Find the escalator and the degradation rate, drop them into the two-line formula with your own marginal rate, and then run the fleet-median row beside them. The distance between those two paybacks is the size of the argument the fine print was hoping to have without you.