Loose-fill fibreglass vs cellulose fibre in the cold

Cellulose fibre insulation gained R-value as the roof space got colder, from R-18 with the outside air at 4°C to R-20.3 at −28°C (Energy Design Update, October 1991, pages 4 to 6, reporting the Oak Ridge test).

The loose-fill fibreglass tested started losing R-value once the roof space was about 18°C colder than the room below, and at a 39 to 42°C difference its thermal resistance “was about 35% to 50% less than at small temperature differences” (Wilkes and Childs, Oak Ridge National Laboratory, 1992, Abstract).

The loss was measured in loose-fill fibreglass only. The paper’s own words: “No significant convection was found either with fiberglass batts” or with the one loose-fill cellulose fibre insulation tested (Abstract). The authors also wrote that similar curves had been seen for “some other (but not all)” loose-fill fibreglass products of the day (section “Loose-Fill Fiberglass Insulation Only”), so the result is for the products tested, not every blown-in fibreglass.

Those were 1991 products. The manufacturers say today’s loose-fill fibreglass is redesigned, so the thing to ask for is a cold test on the exact product offered to you, which is covered below.

What the manufacturers’ documents say, with the research behind each line:

• The Insulation Institute (NAIMA) says the tested fibreglass was installed light. The tests it describes were on loose-fill fibreglass “intentionally installed improperly at a lighter density than the manufacturers specify”, the R-value “began to degrade” once the roof space “dropped below 12°F” (−11°C), “NAIMA members now formulate products and installation guidance specifically for colder climates”, and “when loose-fill insulation is installed at the proper density the R-value does not decrease” (“Loose Fill Insulation Performance and Convection”, paragraphs 2 and 3).

  The research behind this line

  • Document: Insulation Institute (NAIMA) web article “Loose Fill Insulation Performance and Convection”, paragraphs 2 and 3, undated. Its one footnote is Energy Design Update, “Good News on Attic Fiberglass Performance”, January 1992. Insulation Institute article (archived copy).
  • Who ran it, who paid: the article says “a series of tests were completed” and “Researchers estimate” without naming the laboratory, the sponsor or the product. NAIMA is the trade association of the North American fibreglass, rock and slag wool manufacturers, so the page is the manufacturers’ own account.
  • Year: “several years ago” before the January 1992 article it cites.
  • Product and conditions: loose-fill fibreglass installed “at a lighter density than the manufacturers specify”, in a roof space that “dropped below 12°F”. No density figure and no brand.
  • Result: the R-value “began to degrade”; at the proper density “the R-value does not decrease (in fact, the R-value increases at lower temperature)”; the extra heating cost at the light density “would only be $20 (1992 dollars) per year” for a home in the northern United States.
  • What it does not show: which tests these were (the page does not name Oak Ridge, and the Oak Ridge paper gives its own installed densities of 6.4 to 9.0 kg/m³ in Table 1, blown in by a local contractor), any chart or measured figure for today’s products, any Australian product, or settling.
• Owens Corning says the cause was the shape of the pieces, and its own chart holds R-value to −40°F. Green Building Advisor (Allison Bailes, 9 January 2019) reports Owens Corning’s technical bulletin: in the early 1990s its loose-fill was made by “cutting it into little cubes”, “The bonded cubes did not nest well”, and “After changing their product, Owens Corning has measured the R-value versus attic temperature”, with “the measured R-value is above the labeled R-value for each thickness and at temperatures all the way down to -40°F”.

  The research behind this line

  • Document: Green Building Advisor, “Does Fiberglass Attic Insulation Really Lose R-Value?”, Allison A. Bailes III, 9 January 2019, sections “Not all fiberglass is the same” and “Fiberglass industry research on R-value”. The same article is on the author’s Energy Vanguard blog, 7 January 2019. Green Building Advisor article (archived copy).
  • Who ran it, who paid: Owens Corning, in its own technical bulletin. The article says “at least two fiberglass insulation manufacturers have written technical bulletins about their product” with data measured “under conditions similar to those used in the Oak Ridge study”. Manufacturer-funded. The bulletin is not published on the Owens Corning site; the author offers it in a shared folder, which we have not read, so the figures here are the article’s reading of the chart.
  • Year: the bulletin is undated in the article; the article is 2019.
  • Product: Owens Corning’s redesigned unbonded loose-fill fibreglass at four thicknesses. The article gives no brand name and no installed density for the chart.
  • Conditions: “conditions similar to those used in the Oak Ridge study”, roof-space temperatures down to −40°F (−40°C).
  • Result: “the measured R-value is above the labeled R-value for each thickness” and “the R-value increases as the temperature drops until you get below 0°F”.
  • What it does not show: independent confirmation. The author writes “They do have a vested interest, after all” and “If I find some independent research corroborating the OC and JM data, I’ll be sure to share it here.” No Australian product and no settling.
• Johns Manville’s chart peaks and eases at low density and keeps rising at 0.9 pcf. The same article reports Johns Manville’s bulletin: for both of its densities the R-value “starts at 100% with an attic temperature in the low 40s Fahrenheit and rises as the temperature drops”; the low-density product “hits a peak of nearly 125% at a temperature of about -15°F and then begins to drop”, “hitting about 107% at -40°F”; and “JM’s high-density (0.9 pcf) loose-fill fiberglass keeps rising all the way down to -40°F” (0.9 pcf is about 14 kg/m³).

  The research behind this line

  • Document: the same Green Building Advisor article, section “Fiberglass industry research on R-value” (the two Johns Manville graphs) and “Not all fiberglass is the same”. Green Building Advisor article (archived copy).
  • Who ran it, who paid: Johns Manville, in its own technical bulletin. Manufacturer-funded. Not read by us; the figures are the article’s reading of the two graphs.
  • Year: undated in the article; the article is 2019.
  • Product: Johns Manville loose-fill fibreglass at a lower and a higher density (the higher given as 0.9 pcf, about 14 kg/m³). The article says the research was used to set “design specifications for all of Johns Manville’s loose-fill fiber glass attic insulations” and led to “maintaining an appropriate nodule or tuft size, which decreased the permeability of the installed insulation”.
  • Conditions: roof-space temperatures from the low 40s Fahrenheit (about 5°C) down to −40°F (−40°C). Results are given as a percentage of the labelled R-value.
  • Result: the low-density product peaks near 125% at about −15°F (−26°C) and is back to about 107% at −40°F; the 0.9 pcf product keeps rising to −40°F.
  • What it does not show: independent confirmation (the author found none), the test method, the lower density’s figure, any Australian product, or settling.
• Johns Manville’s own attic card says the R-value “may start to decline” in extreme cold and adds depth in Minnesota. The Climate Pro attic card (BIC-162, April 2013), on its Minnesota page, says “fibrous insulation performance generally improves as attics get colder”, then “at extreme cold temperatures, the performance (R-value) of fiber glass insulation products may start to decline below the labeled value”, “Blown loose-fill insulation products are more susceptible to convection than batts”, “All loose-fill products perform differently” and “Johns Manville has extensively tested our attic insulation products at extreme cold temperatures”. Its table “Additional Climate Pro Needed to Attain R-value in Extremely Cold Temperatures” adds, for R-38 at a winter design temperature of −14 to −17°F (−26 to −27°C), 1 inch of depth and 1.3 bags per 1,000 square feet.

  The research behind this line

  • Document: Johns Manville, “JM Climate Pro Fiber Glass Blowing Wool” attic card, BIC-162 4/13, © 2013, reverse side “JM Climate Pro Fiber Glass Blowing Wool for Minnesota Only”. Johns Manville attic card (PDF) (archived copy).
  • Who ran it, who paid: Johns Manville’s own card. The cold tests it mentions are not described on the card (no laboratory, method, temperatures or figures).
  • Why the card exists: “FTC Trade Regulation Rule Part 460 specifies that all products be rated at 75˚F mean temperature” (24°C) and “Minnesota law requires an insulation product to be designed to provide its stated R-value at winter design temperatures”, so the card adds depth and bags for Minnesota’s design temperatures (−14°F at Winona to −31°F at Bemidji).
  • Coverage chart: “Minimum installed thickness” and “Settled thickness”, defined as “Expected thickness after long-term settling has occured” [sic], print the same figure in every row (R-38: 13.8 inches and 13.8 inches).
  • What it does not show: the test results, the temperature at which the decline starts, or any Australian product. The extra-depth table is the manufacturer’s allowance, not a measured loss. The card is dated 2013; later cards may differ.
• Owens Corning’s current data sheet says the labelled R-value is a winter average. The AttiCat PINK Blown-In Insulation Product Data Sheet (May 2026), under “Thermal Performance”, says “Convection in fiber glass loosefill insulation installed in open attics can reduce its thermal performance at extreme winter temperatures” and that the product “has been designed to anticipate this phenomenon and the labeled R-value reflects the average performance over the winter heating season”. Its coverage chart carries “(MINNESOTA ONLY*)” columns, 1.5 extra inches and 2.5 extra bags per 1,000 square feet for R-38 because “Minnesota Administrative Code requires insulation products to provide their stated R-value(s) at Winter Design Temperatures”.

  The research behind this line

  • Document: Owens Corning, “AttiCat PINK Blown-In Fiberglas Insulation” Product Data Sheet, May 2026, sections “Standards, Code Compliance”, the coverage chart with its Minnesota footnote, and “Thermal Performance”. Owens Corning AttiCat data sheet (PDF) (archived copy).
  • Who ran it, who paid: Owens Corning’s own sheet. No cold test is named; the sheet says “R-values are determined in accordance with ASTM C687”, the standard laboratory method.
  • Product and conditions: AttiCat PINK Blown-In Insulation, a North American loose-fill fibreglass “designed for use exclusively with the AttiCat machine”. No temperatures or figures are given for the winter dip.
  • Result: none stated beyond the sentences quoted and the Minnesota columns.
  • What it does not show: how much the R-value falls in extreme cold, at what temperature, or how the winter average was measured. It is not an Australian product.
• The coverage charts print the settled thickness equal to the installed thickness. Owens Corning’s L77 attic card lists “Minimum Initial Installed Thickness” and “Minimum Settled Thickness” at the same figure in every row (R-38: 13.25 and 13.25 inches), its AttiCat sheet does the same (R-38: 13.5 and 13.5 inches), and Johns Manville’s Climate Pro card does the same (R-38: 13.8 and 13.8 inches). The L77 card also says loose-fill insulations “vary in thermal performance due to factors such as aging, mean temperature, settlement, convection, moisture absorption and installation variation” and that “Failure by the installer to provide both the required bags and at least the minimum thickness” “will result in lower insulation R-value”.

  The research behind this line

  • Documents: Owens Corning, PROPINK L77 Loosefill Insulation attic card, Pub. No. 10012954-D, August 2026, “Attics” coverage chart and the two notes under the charts, L77 attic card (PDF) (archived copy); Owens Corning AttiCat Product Data Sheet, May 2026, coverage chart, AttiCat data sheet (PDF) (archived copy); Johns Manville Climate Pro attic card, BIC-162 4/13, coverage chart, Johns Manville attic card (PDF) (archived copy).
  • Who ran it, who paid: the manufacturers’ own coverage charts, which set the installer’s minimums. No settling test, method or measured figure is given on any of the three.
  • Product: PROPINK L77 (32 lb bag), AttiCat (27.5 lb bag) and Climate Pro (28.5 lb bag), all North American loose-fill fibreglass.
  • Result: the settled minimum equals the installed minimum on every row of all three charts.
  • What it does not show: whether or how much the product settles in a roof over years; the charts state a minimum, and the L77 note lists settlement among the factors that vary performance.
• Knauf lists Supafil Loft in Australia; its cold test and Australian settling figure were not found. Knauf’s Australian site lists Supafil Loft as “an unbonded, glasswool insulation” for “new build and existing ceilings”. Knauf’s European Technical Assessment for the sister products Supafil Loft Plus and Supafil Loft Pro (ETA-18/0707, DIBt, 5 June 2025) lists “Settling under impact excitation in the case of free placing” at “≤ 9 % at a minimum bulk density of 20 kg/m³ and a maximum thickness of 330 mm” without binders and ≤ 11% with, a free-placing bulk density of “20 kg/m³ to 23 kg/m³”, and for “Settling under defined climatic conditions” the entry “No performance assessed”.

  The research behind this line

  • Documents: Knauf Australia product page for Supafil Loft, knauf.com/en-AU (the page refuses automated reading, so the two phrases are as the page is indexed; its datasheet is listed there but unread). European Technical Assessment ETA-18/0707 of 5 June 2025, Deutsches Institut für Bautechnik (DIBt), English translation, section 3 table rows “Bulk density”, “Settlement” and “Airflow resistance”. ETA-18/0707 (PDF) (archived copy).
  • Who ran it, who paid: the assessment is issued by DIBt “on the basis of agreed data/ information, deposited with Deutsches Institut für Bautechnik” by the manufacturer, so it is Knauf’s data under a European assessment body’s seal.
  • Product: “Supafil Loft Plus” and “Supafil Loft Pro”, unbonded loose glass mineral wool, European products. The Australian Supafil Loft is not named in the assessment.
  • Conditions and result: free placing on a ceiling or between beams at 20 to 23 kg/m³; settling under impact excitation ≤ 9% (no binder) or ≤ 11% (with binder) at a minimum 20 kg/m³ and a maximum thickness of 330 mm; airflow resistance “≥ 10,0 kPa·s/m² at a minimum bulk density of 20 kg/m³”; declared conductivity 0.037 W/(m·K).
  • What it does not show: a cold R-value test for any Supafil product (not found), the Australian Supafil Loft’s own figures, or long-term settling in a roof (the climatic-conditions row is “No performance assessed”).

The cause is convection, the same thing that happens in a pot on the stove. When the top of a loose, open insulation is cold and the bottom is warm, the air inside it circulates in loops, and those loops carry heat up through the fibre far faster than still air would (Energy Design Update, October 1991). Loose-fill fibreglass is glass fibre chopped into small tufts and blown into the roof space through a hose, so it goes in light and open, and that open air is where the loops get going.

Our premium cellulose insulation goes in as a dense mat of fibres that grip each other, so it stays put when air moves across it. The video beside this paragraph is a finished job of ours with a fan running in the roof space, and the insulation does not move around.

Oak Ridge showed the cause by laying batts over the top of the loose fibreglass, which brought the R-value back to about the sum of the two (R38.8 to R41.7, Table 3) by stopping the loops. A Swedish study found the same convection starting in low-density materials only, with the onset at a modified Rayleigh number of about 22 to 25 (Wahlgren, Chalmers University of Technology).

The paper also prints the air permeability of each material. The loose-fill fibreglass it tested let air through at 6.7×10⁻⁷ to 1.0×10⁻⁶ ft², against 2×10⁻⁸ to 7×10⁻⁸ ft² for loose-fill cellulose fibre insulation (Wilkes and Childs 1992, sections “Loose-Fill Fiberglass Insulation Only” and “Loose-Fill Cellulose Insulation”, quoting Yarbrough 1990). Worked out from those figures, that loose-fill fibreglass was roughly 10 to 50 times more open to air than cellulose fibre insulation. A light, open layer lets the air it holds start circulating on the coldest nights, which is when your heating bill is highest. On a summer afternoon the hot side is on top, the air inside the insulation sits still, and the loops don’t form, which is why this is a winter question.

Our premium cellulose insulation goes in at about 35 kg/m³, two to four times as dense as most glasswool batts (CSIRO lists glass fibre batts at 8 to 16 kg/m³). At that density our premium cellulose insulation resists air moving through it, so it doesn’t lose performance to air currents in the roof the way loose fibreglass can, and your home holds its temperature through extreme summers and freezing winters. The Oak Ridge figures put loose-fill cellulose fibre insulation at roughly 10 to 50 times less air-permeable than the loose-fill fibreglass it tested (Wilkes and Childs 1992, permeability figures above), and the cellulose it tested went in at 30 to 38 kg/m³, the same range as ours.

Your R-value comes from the settled depth of our premium cellulose insulation, so that’s what we measure and what you can check. Our standard ceiling is a settled 120 mm from the ceiling, which gives about R3.0 and keeps the roof heat off your rooms in summer and the warmth in through winter. I’ve never come back to a job for settling; the jobs I’ve come back to were pumped in short, and a top-up fixed them. Our guide on existing cellulose insulation shows what a short job looks like and how a top-up fixes it.

Four things to ask any installer, and the answers we give.

• The exact product and its installed density, on the quote. Our quote names our premium cellulose insulation, and it goes in at about 35 kg/m³ settled.
• An independent cold test for that product. Oak Ridge tested loose-fill cellulose fibre insulation at 30 to 38 kg/m³ and it gained R-value as the roof got colder; the papers are linked above.
• The settled depth and how it is measured. We put the settled depth on your certificate, measured from the ceiling, so you can check it with a ruler from the manhole against our depth tables.
• What the certificate records. Ours records the product, the settled depth and the R-value from our depth tables, so the answer is on paper before we start.

If you are on the Downs, the Granite Belt or anywhere the frost settles, get a fixed-price quote from us online and we’ll put the settled depth and the R-value on it before you decide. Our premium cellulose insulation goes in at the density that gained R-value in the cold test, and it carries our Comfort Zone™ Life-of-House Guarantee.