Cold Climate Heat Pump COP: What Backup Heat Costs

Until July of last year, the federal heat pump test at 5 °F was optional.

Appendix M1 used to print the schedule in a single column. H1 at 47 °F, required. H2 at 35 °F, required. H3 at 17 °F, required. H4 at 5 °F — "optional, steady" (Table 11, 10 CFR part 430, subpart B, appendix M1, the edition superseded on 7 July 2025, pulled through the eCFR versioner API at issue date 5 June 2024). Skip H4 and the appendix supplied equations that extrapolated capacity and power at 5 °F out of the 47, 35 and 17 °F results. The rating then went out as though the machine had been measured down there.

That table is no longer in the CFR. The appendix that took effect on 7 July 2025 prints no heating test conditions of its own at all — it incorporates AHRI 210/240-2024 and defers to that document's tables (current text, amended at 90 FR 1283, 7 January 2025, read 14 September 2026). AHRI sells that standard rather than publishing it, so this article does not claim to know whether H4 is required inside it. That is the first practical finding: the test schedule behind the number on the label is no longer something a buyer can read for free.

That gap is why a second list exists. The people who wrote NEEP's cold climate specification said why they needed one — the federal metric "does not include low temperature testing points below 17°F, assumes the use of electric resistance elements, and tests in steady-state operation (as opposed to allowing modulation)" (NEEP, ccASHP specification and product list, read 14 September 2026).

Three clauses, three separate reasons the number on the box will not tell you what January costs.

Region IV has a design temperature of 5 °F, and that is where the label came from

The metric is a bin calculation. Six generalised climatic regions, each with its own distribution of heating hours, and the efficiency figure is whatever falls out when the unit's performance is run against one of those distributions.

Which one? Here the CFR is still explicit, and this is the piece of the machinery that survived the 2025 rewrite intact. For certifying compliance, use "the fractional heating bin hours specified for Region IV in table 16 of AHRI 210/240-2024." Representations using the hours "specified for Regions other than Region IV" are optional (appendix M1 sections 5.1(b) and 5.2, current text, read 14 September 2026).

Region IV's outdoor design temperature is 5 °F. Region V's is -10 °F.

The bin fractions have moved into the AHRI document with everything else, but the superseded appendix printed them in its table 20, and that version is still retrievable at the issue date above. So I put a real winter next to them. The station is Minneapolis-St. Paul International, October 2024 through April 2025, from NOAA's global-hourly archive through the NCEI access data service (station 72658014922, dataset global-hourly, TMP only, retrieved 14 September 2026), keeping the FM-15 hourly reports and collapsing them onto the nearest clock hour: 5,056 usable of a possible 5,088.

Two things before the table. Table 20's fractions do not sum to one, so each column is normalised to its own total and describes the shape of a heating season rather than its length. And the rows are bins, not thresholds — the appendix's bins are 5 °F wide and centred on 62, 57, 52 and downward, so "below 32 °F" means the bin centred on 27 and everything colder, which is observations under 29.5 °F. Bin the Minneapolis hours the same way or the columns are not comparable.

Region IV spends about one hour in eighty below 5 °F. Minneapolis spent one hour in twelve. Region V, the region nobody is obliged to publish, lands within a tenth of a percentage point of the measurement.

The same equipment carries both numbers when the manufacturer chooses to report them. The Lennox EL21KLV-036 matched to a CK40[C,U]T-36B coil is rated HSPF2 8.5 in Region IV and 7.0 in Region V. Divide by 3.412 and those are seasonal coefficients of performance of 2.49 and 2.05. Same box, same coil, same certificate, 18% apart, and only one of the two is printed where a buyer will see it.

HSPF2 already has a resistance heater inside it

This is where I had the arithmetic wrong for a while, and it flattered the heat pump every time.

HSPF2 is not the compressor's efficiency. The superseded appendix said so in one sentence — for all heat pumps the metric "accounts for the heating delivered and the energy consumed by auxiliary resistive elements when operating below the balance point" — and its summation carried a dedicated term for it: RH(Tj), the electrical energy "used for resistive space heating," modelled as meeting "that portion of the building load that the heat pump does not meet because of insufficient capacity or because the heat pump automatically turns off at the lowest outdoor temperatures." The 2025 rewrite did not change that arithmetic. It moved the description of it into the standard you have to buy, which is why the free text is worth quoting while it is still there.

So HSPF2 divided by 3.412 is a system seasonal COP, already blended with a strip heater, computed against a modelled building load rather than yours. The therms-to-kWh worksheet on this site uses exactly that conversion, and it is the right one for a first pass — but only because it stops there. Divide your heating load by HSPF2 over 3.412 and then add a backup line underneath, and you have bought the resistance heat twice.

Do it one way or the other. Either take the label figure as the whole answer for a Region IV climate, or set the label aside and build the season yourself out of capacity rows and hours. The rest of this article does the second, because in Minneapolis the first one is answering a question about somewhere else.

Four rows off one certificate, and the 70% that decides the badge

The NEEP entry for a matched system reports capacity and power at each tested outdoor temperature, and it carries the AHRI certificate number so the pairing can be checked. AHRI certificate 217121564, Lennox EL21KLV-036-230A outdoor unit with a CK40[C,U]T-36B coil, variable speed, R-454B (retrieved from the NEEP cold climate air source heat pump list product API, 14 September 2026):

Read the capacity column first and the COP column second. Efficiency falling from 3.28 to 2.04 raises the bill. Capacity falling from 30,800 to 21,800 decides whether there is a bill for backup heat at all.

The ENERGY STAR Cold Climate criteria are written around both. To carry the mark a unit must show "COP at 5 °F >= 1.75" and "Percent of Heating Capacity at 5 °F >= 70% of that at 47 °F," and must pass a controls verification procedure confirming those numbers "are achieved by the native controls operating as they would in a customer's home" (ENERGY STAR program requirements for heat pump equipment, version 6.2, revised February 2026, eligibility pages 4 and 5, read 14 September 2026). This pairing makes 21,800 / 30,800 = 70.8%, which rounds to the 70 in NEEP's column and clears the bar with nothing to spare.

None of it belongs to the outdoor unit on its own. Every figure in the table above is a property of the pairing, and the pairing is the thing that quietly changes underneath a quote. In the first 1,200 certificates the NEEP API returned on 14 September 2026, this one outdoor unit appears under 52 certificate numbers across seven coil designations — and the repeats are not all different coils. One designation, CK40[C,U]T-36C, accounts for thirty-nine of the fifty-two, each under its own certificate number, with capacity at 5 °F ranging from 21,200 to 22,400 Btu/h. Five of the eight paired with a B-series coil:

Four HSPF2 figures appear across the whole fifty-two: 8.5, 9.0, 9.6 and 10.4. The highest of them is not cold climate certified, and it is not the best performer at 5 °F either.

The looseness holds at scale, though not as a reversal. Over the full 1,200 certificates the Spearman rank correlation between HSPF2 in Region IV and COP at 5 °F is 0.23 — positive, and far too weak to choose on. The 220 certificates rated HSPF2 10.0 or better have a median COP at 5 °F of 2.00; the other 980 come in at 1.94. Six hundredths of a COP point is what the top fifth of the label buys you at 5 °F. HSPF2 and low-temperature performance are close to two different orderings of the same catalogue, and the label only publishes one of them.

So here is how to read one of these rather than taking a row on trust. The list is manufacturer-submitted. Inside those fifty-two certificates for a single outdoor unit, fifty report a COP at 17 °F somewhere between 2.16 and 2.50. Two report 3.16 and 3.24.

Certificate 217443220 is the higher one: 27,000 Btu/h at 2.44 kW at 17 °F, against 27,800 Btu/h at 3.57 kW on certificate 217443259 for the same outdoor unit. That is a 46% difference in compressor input, which swapping the indoor coil will not explain on its own.

The tell is not the kilowatts. Read down the COP column instead. Certificate 217443220 goes 3.36 at 47 °F, 3.24 at 17 °F, 2.26 at 5 °F: thirty degrees of outdoor air costing twelve hundredths of a COP point, then the next twelve degrees costing ninety-eight. The certificate this article works through goes 3.28, 2.20, 2.04. A heat pump does not hold its mild-weather efficiency all the way down to 17 °F and then fall off a cliff.

So read the column top to bottom before trusting any single number in it, and treat a 17 °F COP sitting within a rounding error of the 47 °F one as a reason to open the certificate. Both anomalous rows here happen to sit on the pairing that carries the two highest HSPF2 figures on the page. Coincidence is entirely possible; it is also the direction in which an error flatters the product, which is reason enough to check rather than assume.

Where the two lines cross, on a house with a load number in front of it

The balance point is not a specification. It is an intersection, and both lines come from documents you can hold.

The load line comes from the Manual J. Suppose it says 36,000 Btu/h at an outdoor design temperature of -11 °F, and take the load to fall to zero at 65 °F, where internal gains and sun cover the rest:

UA = 36,000 / (65 - (-11)) = 474 Btu/h per °F
Load(T) = 474 x (65 - T)

The capacity line comes from the table above, interpolated straight between the tested points. Between 5 and 17 °F it rises 500 Btu/h per degree:

Capacity(T) = 21,800 + 500 x (T - 5)

474 x (65 - T) = 21,800 + 500 x (T - 5)
30,810 - 474T   = 19,300 + 500T
974T            = 11,510
T               = 11.8 °F

Balance point 11.8 °F, at 25,200 Btu/h. Above it the compressor covers the house on its own. Below it, every Btu of the shortfall comes from somewhere else.

That number is far more sensitive to the house than to the equipment. Hold the unit fixed and move only the Manual J figure:

Which is the arithmetic case for doing the envelope work before the equipment. Air sealing does not only cut the load; it drags the balance point down the temperature axis, and the hours below it disappear faster than the load does.

One caveat on the design temperature itself. A single season is not a climatology. In that Minneapolis data the 88th coldest hour — the 1% mark against a full year of 8,760 — was -7.1 °F, so 2024-25 never reached the -11 °F the load calculation was built on. Use the ASHRAE value your Manual J used, and treat any single winter as one draw from a distribution.

Eight percent of the heat, eighteen percent of the kilowatt-hours

Now run every hour of that Minneapolis season through both lines. For each hour: load from the load line, capacity and compressor power interpolated straight between the four certificate rows, heat pump delivers the lesser of load and capacity, resistance takes the remainder at a coefficient of performance of exactly 1.0, which is 3,412 Btu per kWh and no argument about it.

Backup supplied one Btu in twelve and one kilowatt-hour in six. That is the whole point of the exercise: at COP 1.0 against a compressor averaging 2.48, every unit of heat the backup delivers costs about two and a half times as much as the unit before it. The shortfall showed up in 661 hours out of 4,834 heating hours — 13.7% of the season, concentrated in the coldest fortnight of it.

Rebuild it and you should land within about a percent of these figures. The wobble that remains is in how you collapse duplicate reports onto the hour and how you interpolate power between the tested points, and neither moves the conclusion.

Two sanity checks on that arithmetic. The system seasonal COP comes out at 2.22, against 2.05 for the same unit's Region V HSPF2 — close, which is what you would expect given that Region V's bin shape matched this station almost exactly and that HSPF2 has its own resistance term inside it. And heating this house on resistance alone would have taken 21,456 kWh, so the heat pump and its backup together used 45% of that.

The thermostat parameter that costs more than a better compressor

Everything above assumes the compressor runs whenever it can. It very often does not, because a lockout temperature was typed into the thermostat during commissioning and nobody wrote it on the invoice.

Same house, same unit, same weather. The only thing changing is the outdoor temperature below which the compressor is disabled and the strip heat takes the whole load:

A 25 °F lockout is not exotic; it is a common default on dual-fuel and auxiliary-heat configurations, and on this house it turns 1,736 backup kilowatt-hours into 11,101. Six times over, from a setting typed in once. No equipment upgrade in the catalogue moves the season by that much. The rating procedure knows it, too: the appendix's heating equations carry a low-temperature cut-out factor, δ(Tj), precisely because the calculation has to account for units that switch themselves off.

Ask what the lockout is set to on the day the crew leaves, in writing, rather than what the manual's default is.

Which rate the backup lands on

The kilowatt-hour count above is rate-independent. The bill is not, and heating load is one of the few things large enough to change which tariff you belong on.

Xcel Energy's Minnesota Residential Service sheet prints two columns side by side. Standard service is $0.09241 per kWh in the non-summer months. Electric space heating service is $0.06287 for the same months, against an identical $0.10815 in June through September (Minnesota Electric Rate Book, MPUC No. 2, section 5, Residential Service, rate codes A00, A01 and A03, 30th revised sheet 5-1, read 14 September 2026). On 1,736 backup kilowatt-hours the gap between those two columns is $51 a year on the base energy charge alone.

Whether you get it turns on one sentence thirteen sheets later: "Electric space heating charges are applicable only when customer's electric space heating equipment is used as customer's primary heating source" (Rules for Application of Residential Rates, 7th revised sheet 5-13). A heat pump carrying 92% of the season is plainly primary. A heat pump sitting on top of a gas furnace in a dual-fuel arrangement is a question for the utility, in writing, before the equipment is ordered.

Neither column is the price. Bills under that schedule are also subject to the Fuel Clause Rider, and that sheet prints a formula rather than a number — the fuel adjustment factor is set monthly and prorated across billing days, with a residential class cost ratio of 1.0177 (sheet 5-91). The resource adjustment stacks six more riders on top. The only place the total exists is your own bill, which is what the four numbers exercise is for.

For scale while you go and find yours: at a marginal 15 cents the season above is $1,453, of which the backup portion is $260. EIA puts Minnesota's average residential price at 17.52 cents per kWh for June 2026 (Electric Power Monthly, table 5.6.A, released 26 August 2026, read 14 September 2026), but that is revenue divided by sales — it has the customer charge baked into it and it is a June number. A cross-check on the order of magnitude, not an input.

A switchover temperature is a price read backwards

If a gas furnace is staying in the basement, the backup question changes shape. A dual-fuel system does not top the heat pump up; it hands the whole load over at a switchover temperature.

The temperature that minimises cost is not the balance point. It is wherever the heat pump's COP falls to the breakeven COP set by your two fuel rates, and the breakeven formula gives that in one line: 29.31 x AFUE x (your $/kWh) / (your $/therm). Take the answer back to the COP column and read it in reverse:

On this house those settings are worth 114 therms of gas and 6,985 heat pump kWh at a 3 °F switchover, against 412 therms and 3,557 kWh at 25 °F, taking a 92% AFUE furnace. Both are correct answers. Which one is cheaper depends entirely on two numbers off two bills, and it can change sign in a year when gas moves and electricity does not.

The watts nobody quoted you

Three loads sit outside every table above, and only one of them is already accounted for.

Defrost is inside the rating. The H2 test at 35 °F is a frost accumulation test rather than a steady-state one, which is why 35 °F performance can sit below a straight line drawn between 47 and 17. That one is handled. The other two are not.

The base pan heater keeps meltwater from refreezing in the outdoor unit's base, and NEEP records it as a field on the product entry. Of the same 1,200 certificates, 633 report one, at input powers from 39 to 300 watts, cycling rather than continuous, with operation described model by model: one Lennox entry states it "turns on when the unit is in heating mode and temperature is below 23 degree F," a GE entry gives 300 W below 32 °F, a GREE entry 160 W below 33.8 °F, and all of them also run through defrost. Minneapolis spent 2,001 hours below 32 °F in that season. Even at a 25% duty cycle and 150 W, that is 75 kWh nobody's spreadsheet contains.

And the air handler. For a ducted system the rating procedure counts indoor fan energy inside HSPF2, so it is in the label but not necessarily in the certificate row you are reading — worth confirming the AHRI certificate covers the coil and air handler being installed rather than a coil-only rating.

Five rows to take to the site visit

Ask the contractor for the AHRI certificate number of the matched system rather than the outdoor model number, and look it up on the NEEP list yourself. Then fill in five blanks:

1. Design load and design temperature from the Manual J. Two numbers; divide them to get Btu/h per °F.
2. Capacity at 47, 17 and 5 °F for the matched pairing, plus -15 °F if it is published.
3. Where those two lines cross is the balance point. The division takes a minute, and it is the step nobody else in the transaction has any reason to do for you.
4. The compressor lockout setting the thermostat will be commissioned with, in writing.
5. Then find the marginal winter rate on your own bill, and ask whether your utility runs an electric space heating tariff and whether this system would qualify for it.

Rows 1 to 3 say how much backup heat the house will ask for. Row 4 says whether the equipment will be allowed to prevent it. Row 5 turns the answer into money. Nothing on that list requires trusting a brochure, and every source here was read on 14 September 2026 — rate sheets and specifications both move, so check the revision dates on yours.