Bolt Home Loans Research · Working paper

Leverage in VA Refinancing

How the balance you refinance changes the cost of borrowing more.

Original research by Bolt Home Loans

Analysis: Bolt Home Loans. Underlying data: U.S. Department of Veterans Affairs, Loan Volume by County and Loan Volume by State reports.

· Public edition 1.0

Abstract

Finding: in VA fiscal year 2026, VA guaranteed 98,827 cash-out refinances, 16.20% more than the 85,050 of fiscal year 2025, yet cash-outs fell from 41.59% to 32.27% of VA refinances because streamline refinances grew far faster.

This study uses public loan volume workbooks from the U.S. Department of Veterans Affairs (VA) to describe how two kinds of VA refinance moved between fiscal years 2015 and 2026. It reports counts and shares side by side, separates changes within states from shifts between them, tests whether counties followed the national pattern, and explains why the balance being refinanced shapes the cost of borrowing more. All empirical figures are aggregate counts and dollar totals; no borrower or loan-level data are used.

VA Loan Disclaimer: Bolt Home Loans LLC is not affiliated with or endorsed by the Department of Veterans Affairs or any government agency. VA loan programs are offered by private lenders.

Film briefing

Leverage in VA Refinancing. 1 minute 58 seconds, narrated in English with captions. Produced by Bolt Home Loans from the figures in this study.
Transcript

In fiscal 2026, VA cash-out refinances rose 16 percent. Yet their share of VA refinancing fell, from 41.6 percent to 32.3. Both are true. Here’s the fraction.

On top: cash-outs, which let homeowners borrow against their equity. Fiscal 2025: 85,050. Underneath: every VA refinance. 204,507. Fiscal 2026. The top grows, to 98,827. The bottom grows faster. Streamline refinances, which usually lower the rate on an existing VA loan, jumped 74 percent. The share drops more than nine points.

Share can mislead the other way, too. In fiscal 2023, streamlines nearly vanished. Cash-outs were almost 94 percent of the total, on fewer loans than either recent year.

And it’s not one place. Of 498 counties with at least 100 VA refinances in both years, 373 had more cash-outs and a smaller share. Oakland County, Michigan: 79 more cash-outs, and a share six points lower.

These are counts of VA-guaranteed loans. They don’t show how much cash anyone received, or why.

For one household, the question is the price of the new money. Say you owe $400,000 at 5 percent, and want $50,000 more. A cash-out at 5.5 percent costs as much opening interest as borrowing that $50,000 separately at 9.5. Before fees. Not lifetime savings.

More cash-outs. A smaller share. To read a share, watch the denominator. Original research by Bolt Home Loans, from public VA data.

1What happened to VA refinances in fiscal year 2026?

Between VA fiscal years 2025 and 2026, cash-out refinances rose by 13,777 loans (16.20%) and IRRRLs rose by 87,959 (73.63%), so the cash-out share of the 306,243 VA refinances fell by 9.3170 percentage points (VA Loan Volume by State, Grand Total rows).

A few terms first. A cash-out refinance replaces an existing mortgage with a new VA-guaranteed loan that can be larger than the old balance; the difference can go to the borrower or pay other debts (VA.gov). An IRRRL, the Interest Rate Reduction Refinance Loan, is VA’s streamline refinance. It replaces an existing VA loan and generally does not let the borrower take cash. In this study, a refinance means either of those two types, as VA classifies them.

VA counts loans it guaranteed during its fiscal year. Fiscal year 2026 ran from October 1, 2025 to September 30, 2026. That is not the same as a calendar year, and a guarantee date is not the same as the date a borrower applied or the date the loan closed. Figures here should not be compared directly with calendar-year origination statistics from other sources.

The share is a simple ratio. In words: cash-out loans divided by all refinances. In symbols:

share = C ÷ (C + I)

Here C is the number of cash-out loans and I is the number of IRRRLs. Purchase loans are left out on purpose, because the question is what kind of refinance borrowers took, not how refinancing compared with buying.

Table 1. National VA refinance counts, FY2025 and FY2026
MeasureFY2025FY2026Change% change
Cash-out loans (C)85,05098,827+13,777+16.20%
IRRRL loans (I)119,457207,416+87,959+73.63%
Refinances (C + I)204,507306,243+101,736+49.75%
Cash-out share41.5878%32.2708%−9.3170 pp—

The arithmetic is short. In fiscal 2025, 85,050 ÷ 204,507 = 41.5878%. In fiscal 2026, 98,827 ÷ 306,243 = 32.2708%. The difference is −9.3170 percentage points. A percentage point is the plain difference between two percentages; it is not a percent change.

Figure 1. Out of every 100 VA refinances, how many were cash-outs?

FY2025: about 42 of 100

FY2026: about 32 of 100

Each square is one of 100 refinances. Rounded from 41.5878% and 32.2708%. Source: Bolt calculation from VA Loan Volume by State, FY2025 and FY2026 (national.csv).

Which type drove the growth?

Total refinances grew by 101,736 loans. IRRRLs supplied 87,959 of that growth, or 86.46%. Cash-outs supplied the other 13,777, or 13.54%. Both types grew. The share fell because the denominator grew faster than the numerator, not because cash-outs shrank.

This matters for interpretation. A falling share is often read as a sign that fewer people tapped home equity. In this case the opposite was true in count terms: more cash-out loans were guaranteed in fiscal 2026 than in fiscal 2025. What changed was the mix of refinancing activity around them.

Why VA’s “% of all loans” answers a different question

VA’s state workbooks also print a “Total Cash-Out %” column. That figure divides cash-outs by all VA loans, including purchases. Nationally, it is 16.10% for fiscal 2025 (85,050 of 528,340 loans) and 15.21% for fiscal 2026 (98,827 of 649,769). It moved less than our measure because purchase loans, which barely changed, sit in its denominator.

Neither measure is wrong. VA’s column describes cash-outs as a part of the whole guarantee program. Ours describes the choice among refinance types. Readers comparing published figures should check which denominator each one uses before drawing a conclusion.

Reading counts and shares together

A count answers “how many?” A share answers “what part of the whole?” Each can move without the other. If both cash-outs and IRRRLs double, the share stays the same while the count doubles. If cash-outs hold steady and IRRRLs collapse, the share soars while the count stands still.

For that reason this study reports the numerator, the denominator and the share together wherever it can. A reader who sees only “32 percent” cannot tell whether VA cash-out lending expanded or contracted. A reader who sees 98,827 out of 306,243 can.

2How does fiscal year 2026 compare with the last twelve years?

Across twelve VA fiscal years from 2015 to 2026, the cash-out share of VA refinances ranged from 16.63% in FY2021 to 93.78% in FY2023, and the year with the most cash-out loans (FY2022, 207,777) was not the year with the highest share (VA Loan Volume by State, one file per year).

We assembled the Grand Total row from twelve annual VA state workbooks, one for each fiscal year from 2015 through 2026. Each row in Table 3 links to the exact VA file it came from. Use the toggle to switch the chart between the share and the underlying counts. The table holds the same values for readers who do not use the chart.

Figure 2. Cash-out share of VA refinances and loan counts, FY2015–FY2026
0%25%50%75%100%’15’16’17’18’19’20’21’22’23’24’25’26
Bolt calculation from twelve VA Loan Volume by State workbooks; values in Table 3 and national.csv. Nominal counts of VA-guaranteed loans by fiscal year.
Table 3. National VA refinance counts and cash-out share by fiscal year
Fiscal yearCash-outIRRRLRefinancesCash-out shareVA source file
FY2015114,226194,799309,02536.96%fy2015_loans_by_state.xlsx
FY2016136,888215,553352,44138.84%fy2016_loans_by_state.xlsx
FY2017169,039190,913359,95246.96%fy2017_loans_by_state.xlsx
FY2018160,05167,347227,39870.38%fy2018_loans_by_state.xlsx
FY2019145,18594,860240,04560.48%fy2019_loans_by_state.xlsx
FY2020156,330662,064818,39419.10%FY_2020_loans_by_state.xlsx
FY2021165,871831,824997,69516.63%FY2021_loans_by_state.xlsx
FY2022207,777127,949335,72661.89%fy22-loans-state.xlsm
FY202375,4205,00080,42093.78%fy23-states-total.xlsx
FY202467,22050,826118,04656.94%fy24-states-total.xlsx
FY202585,050119,457204,50741.59%fy25-states-total.xlsx
FY202698,827207,416306,24332.27%fy26-states-total.xlsx

Three years stand out. In fiscal 2021, VA guaranteed 165,871 cash-outs and 831,824 IRRRLs. Cash-outs were numerous, but streamlines were five times as numerous, so the share was only 16.63%. In fiscal 2022, cash-outs reached the window’s highest count, 207,777, while IRRRLs dropped to 127,949; the share rose to 61.89%.

In fiscal 2023, the share peaked at 93.78%: 75,420 ÷ (75,420 + 5,000). Yet that year had far fewer cash-out loans than fiscal 2022 and fewer than fiscal 2026. A very high share sat on top of a very small refinance market of 80,420 loans.

These are different populations of borrowers. A year with few IRRRLs is a year in which few VA borrowers held a loan they could profitably streamline. Those who refinanced anyway were more likely to have another reason, such as needing cash. A share can therefore swing widely while the number of people taking cash changes much less.

The middle of the window shows the same mechanism in reverse. From fiscal 2017 to 2018, IRRRLs fell from 190,913 to 67,347, while cash-outs dipped only slightly, from 169,039 to 160,051. The share jumped from 46.96% to 70.38%. Then, in fiscal 2020, IRRRLs surged to 662,064 and the share dropped to 19.10%, even though cash-outs rose.

Across the twelve years, cash-out counts stayed between 67,220 and 207,777, roughly a threefold range. IRRRL counts ranged from 5,000 to 831,824, more than a hundredfold. Streamline volume, not cash-out volume, is what makes the share so volatile.

We do not attribute these movements to any cause. Interest rates, home prices, program rules and lender behavior all plausibly matter, but the workbooks contain none of those variables, and two-variable patterns over twelve points cannot separate them. The safe conclusion is descriptive: a cash-out share without its count, and the count of IRRRLs beside it, can mislead.

3Did the share fall because refinancing moved between states?

Of the −9.31675 percentage-point change across 54 named VA jurisdictions from FY2025 to FY2026, changes in share within jurisdictions account for −8.95458 points and shifts in where refinances occurred account for −0.36217 points; 53 of the 54 had lower shares (Bolt decomposition of VA Loan Volume by State).

A national share is a weighted average of local shares. It can fall for two reasons. Places could each have lower shares, which we call a change in behavior. Or refinancing could shift toward places that already had low shares, which we call a change in weights. A decomposition splits the national change into those two parts.

For each jurisdiction j, let w be its weight, meaning its share of all refinances in the year, and p its cash-out share. Subscript 0 marks fiscal 2025 and 1 marks fiscal 2026. The within part adds up each place’s change in share, weighted by its average weight across the two years:

within = Σ [(w₁ + w₀) ÷ 2] × (p₁ − p₀)

The weights part adds up each place’s change in weight, multiplied by its average share across the two years:

weights = Σ [(p₁ + p₀) ÷ 2] × (w₁ − w₀)

Because both parts use two-year averages, they add up exactly to the total change, with nothing left over. This is a standard symmetric method. It is an accounting identity, not a causal model: it says where the change shows up, not why it happened.

A simple example helps. Imagine two states with equal refinance volume. If each state’s share falls by 10 points, the whole national change is within. If neither state’s share changes but volume shifts toward the state with the lower share, the whole change is weights.

Figure 3. Decomposition of the change in cash-out share, FY2025 to FY2026 (percentage points)
Figure 3 values
Within jurisdictions−8.95458 pp
Changing weights−0.36217 pp
Total−9.31675 pp
Bolt calculation over the 50 states, DC, Guam, Puerto Rico and the U.S. Virgin Islands; per-jurisdiction terms in state-decomposition.csv. Bars show magnitude; both parts are negative.

Nearly all of the decline came from lower shares inside jurisdictions. Shifts in where refinancing happened contributed about a third of a point. 53 of the 54 named jurisdictions recorded a lower share in fiscal 2026. The exception, the U.S. Virgin Islands, went from 1 cash-out out of 4 refinances to 3 out of 5; such tiny counts carry almost no weight.

The two parts sum to −9.31675 points, slightly different from the national −9.3170. The reason is population. The named jurisdictions held 204,506 refinances in fiscal 2025, one fewer than the national 204,507. VA’s county workbook shows one cash-out loan in a row with no county or state label; the national total includes it, but it cannot be assigned to a jurisdiction. In fiscal 2026 the named total equals the national 306,243.

Why does the split matter? If the decline had been mostly weights, the national figure would say little about any one place; it would reflect where refinancing happened to grow. Because the decline is almost entirely within jurisdictions, the national figure is a fair summary of what most places experienced, at least at the state level.

Small jurisdictions deserve caution. A share computed from a handful of loans can rise or fall sharply by chance. They carry little weight in the decomposition, which is appropriate, but their individual shares should not be read as evidence about local conditions.

The weights here count refinances, so large states matter more. An equal-weight view treats every jurisdiction alike. The simple average of the 54 jurisdiction shares fell from 43.82% to 35.82%. That describes a typical jurisdiction rather than the typical refinance, and both readings point the same way.

What the decomposition does not say

A large within-jurisdiction term does not mean that borrowers in each state changed their preferences. It means that, inside each state, the refinance population tilted toward streamlines. That could happen because more existing VA borrowers became able to benefit from a streamline, because lenders marketed one product more, or for reasons the data cannot see. The method locates the change; it does not explain it.

4Did counties show the same pattern?

Among the 498 counties with at least 100 VA refinances in both FY2025 and FY2026, 373 (74.9%) had both more cash-out loans and a lower cash-out share; these counties held 73.80% and 73.38% of national refinances (VA Loan Volume by County).

County data test whether the national pattern was broad or driven by a few places. VA’s county workbooks label each row “COUNTY, ST”. They do not include FIPS codes, the standard numeric county identifiers. We therefore match counties across years only when the label is identical in both files. We do not guess at renamed or merged areas.

VA lists only counties with at least one VA loan that year. A county missing from a year’s file is treated as missing, not as zero. Inside a listed row, a blank category cell is treated as zero only after we confirm that the row’s cash-out, IRRRL and purchase counts add up to its published total. Every row passed that check.

Small counties produce noisy shares. One loan can move a share by tens of points. We therefore require a minimum number of refinances in both years and report three cutoffs. To decide whether a share fell, we compare exact integer cross products, so no rounding can flip a result: the share fell when C₁ × (C₀ + I₀) is less than C₀ × (C₁ + I₁).

Table 4. Matched counties with both rising cash-out counts and falling cash-out share, by minimum refinances
Minimum refis each yearMatched countiesCount roseShare fellBothBoth, % of matchedRefis covered FY25 / FY26
5081665273557270.10%173,552 / 258,638
10049840246937374.90%150,931 / 224,723
25019715919515779.70%103,917 / 153,947

The main cutoff is 100. At that level, 402 of 498 counties had more cash-outs, 469 had a lower share, and 373 had both. At 50 the proportion with both was 70.10%; at 250 it was 79.70%. The pattern grows stronger as counties get larger and less noisy. As a robustness check, excluding Connecticut from the matched set at the 100 cutoff leaves 373 of 494 counties (75.51%) with both changes, compared with 373 of 498 (74.90%); Connecticut’s mixed fiscal 2026 labels chiefly affect local rankings, and this check makes no causal claim.

Why three cutoffs? Any single threshold is a judgment call. Showing the result at 50, 100 and 250 refinances lets readers see whether the conclusion depends on that choice. Here it does not: the proportion rises steadily with size, from about seven in ten to about eight in ten. A pattern that only appeared at one cutoff would deserve much less confidence.

The counts can be read in another way. At the 100 cutoff, 96 qualifying counties recorded fewer or the same cash-outs, and 29 recorded a share that held or rose. Those exceptions are real. The national pattern was broad, but it was not universal.

These figures describe qualifying counties only. They do not mean three quarters of all U.S. counties followed the pattern. Most counties are too small to qualify at 100 refinances, though the qualifying ones hold most of the volume.

An example: Oakland County, Michigan

Oakland County appears once each year under the label “OAKLAND, MI”. In fiscal 2025 it recorded 183 cash-outs and 220 IRRRLs, so 183 ÷ 403 = 45.41%. In fiscal 2026 it recorded 262 and 404, so 262 ÷ 666 = 39.34%. Cash-outs rose by 79, but streamlines rose by 184, and the share fell. It is an illustration, not a representative county.

Were cash-outs concentrated in a few counties?

Ranked by cash-out count, the same five counties led in both years: Maricopa (AZ), Clark (NV), Riverside (CA), San Diego (CA) and El Paso (CO). We ranked each year separately and found the same five. Together they rose from 4,690 to 5,004 cash-out loans, but their share of national cash-outs fell from 5.51% to 5.06%. Cash-out growth was spread more widely than in the largest counties.

Table 5. Five counties with the most VA cash-out loans, FY2025 and FY2026
County (VA label)FY2025 cash-outFY2026 cash-out
MARICOPA, AZ1,3581,480
CLARK, NV9981,004
RIVERSIDE, CA876906
SAN DIEGO, CA748861
EL PASO, CO710753
Five-county total4,6905,004

County explorer

The explorer covers every county label in either year’s workbook. Filter by state, search by name, or rank by fiscal 2026 share among counties with at least 100 refinances. Connecticut is excluded from the ranking because its fiscal 2026 rows mix legacy county labels with planning-region labels.

83 counties.

VA cash-out and IRRRL loan counts and cash-out share by county, FY2025 and FY2026
County (VA label)StateFY25 cash-outFY25 IRRRLFY25 shareFY26 cash-outFY26 IRRRLFY26 share
ALCONAMI4640.00%91047.37%
ALGERMI50100.00%2433.33%
ALLEGANMI284339.44%356335.71%
ALPENAMI11857.89%101441.67%
ANTRIMMI16672.73%151353.57%
ARENACMI5645.45%3730.00%
BARAGAMI1325.00%6366.67%
BARRYMI242548.98%374246.84%
BAYMI233241.82%264934.67%
BENZIEMI13765.00%131154.17%
BERRIENMI463954.12%509235.21%
BRANCHMI131743.33%152141.67%
CALHOUNMI446640.00%609239.47%
CASSMI141646.67%102033.33%
CHARLEVOIXMI91047.37%7943.75%
CHEBOYGANMI141058.33%111837.93%
CHIPPEWAMI61331.58%141942.42%
CLAREMI131644.83%191654.29%
CLINTONMI202742.55%283842.42%
CRAWFORDMI161257.14%71630.43%
DELTAMI211952.50%161748.48%
DICKINSONMI19870.37%142436.84%
EATONMI474252.81%468235.94%
EMMETMI12666.67%101441.67%
GENESEEMI11014742.80%13823437.10%
GLADWINMI82028.57%202940.82%
GOGEBICMI6842.86%71138.89%
GRAND TRAVERSEMI262650.00%295434.94%
GRATIOTMI111150.00%141156.00%
HILLSDALEMI131448.15%212347.73%
HOUGHTONMI12475.00%8947.06%
HURONMI5645.45%3827.27%
INGHAMMI695356.56%8812042.31%
IONIAMI122631.58%133726.00%
IOSCOMI10855.56%171651.52%
IRONMI6746.15%71236.84%
ISABELLAMI121348.00%122334.29%
JACKSONMI456441.28%5410933.13%
KALAMAZOOMI617245.86%7911740.31%
KALKASKAMI11955.00%111936.67%
KENTMI11613446.40%14119941.47%
KEWEENAWMI00No refinances1150.00%
LAKEMI71041.18%10662.50%
LAPEERMI323150.79%285832.56%
LEELANAUMI1516.67%1516.67%
LENAWEEMI212842.86%255929.76%
LIVINGSTONMI576945.24%8112239.90%
LUCEMI2250.00%2433.33%
MACKINACMI5455.56%5645.45%
MACOMBMI19125542.83%22844134.08%
MANISTEEMI111052.38%211952.50%
MARQUETTEMI181652.94%233539.66%
MASONMI101343.48%121544.44%
MECOSTAMI151648.39%222546.81%
MENOMINEEMI7846.67%62320.69%
MIDLANDMI203735.09%465446.00%
MISSAUKEEMI51033.33%7653.85%
MONROEMI487040.68%658842.48%
MONTCALMMI312753.45%345936.56%
MONTMORENCYMI8850.00%10855.56%
MUSKEGONMI758247.77%8616634.13%
NEWAYGOMI193634.55%332457.89%
OAKLANDMI18322045.41%26240439.34%
OCEANAMI81633.33%151746.88%
OGEMAWMI10471.43%211558.33%
ONTONAGONMI6460.00%2433.33%
OSCEOLAMI121152.17%51229.41%
OSCODAMI20100.00%5838.46%
OTSEGOMI121348.00%141942.42%
OTTAWAMI507739.37%6211335.43%
PRESQUE ISLEMI3925.00%10952.63%
ROSCOMMONMI211853.85%222447.83%
SAGINAWMI396238.61%5310333.97%
SANILACMI191359.38%143230.43%
SCHOOLCRAFTMI4357.14%6366.67%
SHIAWASSEEMI283147.46%255332.05%
ST. CLAIRMI697348.59%8411043.30%
ST. JOSEPHMI101441.67%163531.37%
TUSCOLAMI172540.48%212942.00%
VAN BURENMI222744.90%254834.25%
WASHTENAWMI395641.05%557941.04%
WAYNEMI26127149.06%32154037.28%
WEXFORDMI24972.73%182442.86%

“Not listed” means the county does not appear in that year’s VA workbook (no data). A listed county with blank cash-out or IRRRL cells is counted as 0; “No refinances” means both were 0, so no share is defined. Counties are matched across years only by identical VA labels. Connecticut’s FY2026 rows mix legacy county labels with planning-region labels that carry only cash-out loans, so Connecticut shares are not comparable and are excluded from the ranking. The full county file is listed under Downloads.

5How much principal did VA cash-outs carry, and why does balance matter?

VA-guaranteed cash-out loans carried $30,407,451,735 of gross principal in FY2025 and $36,887,371,165 in FY2026 (+21.31%), faster than the 16.20% growth in cash-out count (VA Loan Volume by State).

Principal is the full amount of each new loan. For a cash-out, principal includes the payoff of the old mortgage, any financed costs and any cash the borrower received. The workbooks do not separate those parts. So these totals cannot tell us how much cash or equity borrowers withdrew. They tell us only how large the new loans were.

Mean cash-out principal is total principal divided by the number of loans. It rose from $357,524 to $373,252, or 4.40%. That increase is a weighted national average. It can rise because individual loans got larger, because more loans came from high-cost areas, or both. All amounts are nominal dollars, not adjusted for inflation.

Total principal grew faster than the count of cash-out loans: +21.31% against +16.20%. Put differently, roughly a fifth of the growth in dollars came from larger average loans and the rest from more loans. Because composition can shift, that split is descriptive. It does not mean that any given borrower took a larger loan.

Why the starting balance matters

Interest is proportional to balance. In words, one month of extra interest from a rate increase is the balance times the rate increase, divided by twelve. In symbols, for balance B and a quarter-point increase:

ΔInterest = B × 0.0025 ÷ 12

For a starting reference, we use the mean principal of new VA purchase loans, computed as total purchase dollars divided by purchase loans. It was $243,987 in fiscal 2015 and $410,567 in fiscal 2026. A quarter point on those amounts adds $50.83 and $85.53 of opening monthly interest. That is interest in the first month, not a change in the monthly payment, which also includes principal repayment.

Figure 4. Extra first-month interest from a 0.25-point higher rate at the mean VA purchase principal
Figure 4 values
FY2015 mean principal $243,987$50.83
FY2026 mean principal $410,567$85.53
Bolt calculation: (VA purchase loan dollars ÷ purchase loans) × 0.0025 ÷ 12, nominal dollars, from FY2015 and FY2026 state workbooks. New purchase loans, not balances of existing borrowers.

The comparison spans eleven years and uses nominal dollars. Part of the rise in mean purchase principal reflects general price inflation, part reflects home prices rising faster than general prices, and part reflects changes in which buyers used VA loans. The public workbooks cannot separate those parts, so the figure shows only how much a fixed rate change costs at each starting amount.

The point for refinancing is direct. A cash-out moves the entire existing balance to the new rate. If the new rate is higher than the old one, the increase applies to every dollar already owed, not only to the new cash. The larger the existing balance, the more that matters, which is why two households seeking the same cash can face very different costs.

6When does a cash-out cost less than keeping the first mortgage?

On opening interest alone, a cash-out costs less than keeping the first mortgage and adding a second loan whenever the second loan’s rate is above the break-even rate s* = r₁ + (B ÷ C) × (r₁ − r₀); fees, term and holding period can reverse that result.

This section moves from public data to hypothetical arithmetic. The calculator and the worked cases use numbers chosen for illustration. They are not VA data, Bolt offers, market rates or observed borrower outcomes.

Consider a household with a current balance B at annual rate r₀ that wants new money C. Option A is a cash-out refinance into one loan of B + C at rate r₁. Option B keeps the first mortgage and borrows C on a second lien at rate s. In words, the two options have equal opening interest when the cash-out loan’s yearly interest equals the old loan’s interest plus the second loan’s interest:

(B + C) × r₁ = B × r₀ + C × s

Solving for s gives the break-even second-loan rate:

s* = r₁ + (B ÷ C) × (r₁ − r₀)

The ratio B ÷ C is the leverage term. Each point by which the cash-out rate exceeds the old rate is multiplied by how large the existing balance is relative to the new cash. When r₁ equals r₀, s* equals the first-mortgage rate. When r₁ is far enough below r₀, s* turns negative: no second loan at a nonnegative rate matches the cash-out on opening interest.

All calculator inputs are entered by you. The starting values are hypothetical. Nothing you type is sent or stored.

Second-mortgage break-even calculator

Calculated in your browser. Nothing you enter is sent or saved.

Balance of your existing first mortgage

Additional cash you want to borrow

Rate on your current first mortgage

Rate on the new, larger first mortgage

Break-even second-mortgage rate

9.50%

A second mortgage on the new money priced below 9.50% means keeping the first mortgage plus a second costs less in opening interest. Priced above 9.50%, the cash-out refinance costs less in opening interest. At exactly that rate the two are equal.

Request a VA cash-out comparison from Bolt

Opens Bolt’s VA cash-out page. Calculator values are not sent and no application is started.

How to read the result. Suppose the calculator shows 9.50%. If a household could borrow the new money on a second loan below 9.50%, keeping the first mortgage would carry less interest at the start. Above 9.50%, the cash-out would. The comparison is about the first month or first year only; it is a starting point, not a verdict.

Worked case: equal opening interest

With B = $400,000 at r₀ = 5%, C = $50,000 and r₁ = 5.5%, the break-even rate is 5.5 + 8 × 0.5 = 9.50%. A $450,000 loan at 5.5% carries $24,750 of yearly opening interest. A $400,000 loan at 5% plus $50,000 at 9.5% carries $20,000 + $4,750 = $24,750. At r₁ = 4.5% the threshold falls to 0.50%. With r₀ = 6% and r₁ = 5%, it is −3.00%.

Worked case: fees, terms and time

Opening interest ignores four things that often decide the question: upfront costs, repayment schedules, how long the household keeps the loans, and rates that can change. Table 6 adds them using hypothetical values. Every structure delivers the same $50,000 of spendable cash, and each finances its own fees. The cash-out base is $400,000 + $50,000 + $6,000 of financed closing costs = $456,000. An illustrative nonexempt 2.15% funding fee on that base is $9,804, so the new note is $465,804, a 30-year loan at 5.5%. The fee is not charged on itself. A VA funding fee applies only when the borrower is not exempt, and VA publishes the current rates (VA funding fee and closing costs); the figure here is an assumption.

The alternatives keep the existing $400,000 loan at 5% with 25 years left. One adds a 15-year fixed second loan at 9.5% with an opening principal of $51,500, including $1,500 of financed costs. The other adds an interest-only line at 8.5% that resets to 10.5% after 2 years, with an initial draw of $50,500, including $500 of financed costs. Economic cost is cumulative interest paid plus the original fees counted once. “Fees incurred (financed)” are rolled into each loan, not paid in cash at closing; no additional cash fees are assumed.

Table 6. Hypothetical costs over three holding periods (existing rate 5%)
Holding periodStructureInterest paidFees incurred (financed)Interest + feesBalance still owed
2 yearsCash-out refinance$50,571$15,804$66,375$452,900
Keep first + fixed second$48,687$1,500$50,187$431,160
Keep first + resetting line$47,788$500$48,288$433,583
5 yearsCash-out refinance$123,568$15,804$139,372$430,685
Keep first + fixed second$116,949$1,500$118,449$395,881
Keep first + resetting line$119,115$500$119,615$404,821
10 yearsCash-out refinance$236,049$15,804$251,853$384,479
Keep first + fixed second$214,940$1,500$216,440$321,304
Keep first + resetting line$227,306$500$227,806$346,198

In this case, keeping the first mortgage costs less at every horizon. The cash-out also leaves a larger balance owed, partly because it restarts a 30-year schedule and finances its fees. Monthly payments differ too: the fixed second loan repays principal quickly and has the highest payment.

Change one input and the answer can reverse. Table 7 raises the existing rate to 6.5%, so the cash-out now lowers the rate on the existing balance.

Table 7. The same hypothetical, with an existing rate of 6.5%
Holding periodStructureInterest paidFees incurred (financed)Interest + feesBalance still owed
2 yearsCash-out refinance$50,571$15,804$66,375$452,900
Keep first + fixed second$60,652$1,500$62,152$434,426
Keep first + resetting line$59,754$500$60,254$436,849
5 yearsCash-out refinance$123,568$15,804$139,372$430,685
Keep first + fixed second$146,625$1,500$148,125$403,809
Keep first + resetting line$148,791$500$149,291$412,749
10 yearsCash-out refinance$236,049$15,804$251,853$384,479
Keep first + fixed second$272,784$1,500$274,284$335,652
Keep first + resetting line$285,150$500$285,650$360,545

Three features of these tables are worth noting. First, fees are fixed while interest accumulates, so short holding periods favor the option with lower costs. Second, a resetting line can look cheap early and become expensive later; here, the reset makes it the costliest option over ten years in Table 7. Third, balances owed differ, so a lower cost over a period may come with more debt left at its end. A complete comparison has to show all three.

Over two years, the financed fees still make the cash-out more expensive. Over five and ten years, its lower rate on the whole balance outweighs them. National figures cannot tell a household which case it is in; that depends on its own balance, rates, costs and plans.

Two further notes. A loan VA classifies as a cash-out can also refinance a non-VA mortgage into a VA loan, with or without much cash to the borrower (VA.gov). So a cash-out count is not a count of equity withdrawals. And for background on household take-up of cash-out refinancing and a Fannie Mae policy change, see Federal Reserve Bank of Philadelphia Working Paper 26-01 (doi.org/10.21799/frbp.wp.2026.01). We cite it as context only and rely on none of its results.

7What can these public counts not tell us?

VA’s public loan volume workbooks count guaranteed loans and sum their principal by type and place; they do not show cash disbursed, borrower identity, loan-to-value, interest rates, fees or later loan performance.

  • Counts are loans, not unique people. One household could appear more than once, and counts say nothing about who was eligible but did not refinance.
  • Principal is not cash proceeds. No public field separates the old payoff from new money or financed costs.
  • The files contain no interest rates, fees, credit or loan-to-value data, so the study cannot explain why borrowers chose one type over another.
  • Nothing here measures loan performance, defaults or whether a refinance helped a household.
  • Selection matters. The borrowers who refinance in a given year differ from those who do not, and from those who refinance in other years.
  • County labels are text without FIPS codes. Renamed or restructured areas, such as Connecticut’s planning regions, cannot be matched reliably.
  • These are complete administrative tallies, not samples. We therefore report no confidence intervals; uncertainty comes from definitions and coverage, not sampling.
  • Two adjacent fiscal years are not a trend, and twelve annual points cannot isolate causes.

Readers should also keep the unit of time in mind. A fiscal-year count gathers loans VA guaranteed over twelve months that straddle two calendar years. Comparisons with calendar-year market statistics, or with data organized by application or closing date, will not line up exactly even when both describe the same loans.

Finally, classification belongs to VA. When this study says “cash-out” or “IRRRL,” it means the category VA assigned in its workbooks. We do not reclassify loans or infer a borrower’s purpose from the loan type.

8How was the study done, and how can it be reproduced?

Every number comes from public VA workbooks linked on VA’s Loan Volume by State and Loan Volume by County pages, processed with a published Python script that checks totals before writing any table.

The study has four parts: a national series, a state decomposition, a county comparison and a hypothetical cost analysis. The first three use only VA’s public workbooks. The fourth uses only stated assumptions. No borrower records, Bolt pipeline data or third-party estimates are used anywhere.

Sources and coverage

National figures for fiscal 2015 through 2026 come from the Grand Total row of twelve annual state workbooks (Table 3 links each file). State figures come from the fiscal 2025 and 2026 state workbooks; county figures come from fy25-counties-total.xlsx and fy26-counties-total.xlsx. We confirmed each workbook’s internal title names the expected fiscal year. Coverage is the 50 states, the District of Columbia, Guam, Puerto Rico and the U.S. Virgin Islands.

Reconciliation

For fiscal 2026, state rows, county rows and both Grand Total rows agree exactly. For fiscal 2025, the national Grand Totals agree. The named state rows omit one cash-out loan, and the county file separately contains one cash-out loan in an unnamed row. Their amounts reconcile, but VA does not say whether they are the identical record. National figures use the Grand Totals. We do not use VA’s rounded percentage columns.

Replication steps

  1. Install Python 3 and the openpyxl package.
  2. Download replicate_va_refinance.py and run it. It downloads the VA workbooks and prints a checksum for each.
  3. For each state workbook it reads the Grand Total row’s cash-out, IRRRL and purchase counts and dollar totals.
  4. For each county workbook it asserts that every row’s categories sum to its total and that county sums equal the Grand Total.
  5. It computes shares, the three county cutoffs with integer cross products, and the decomposition, and asserts that the two parts add to the total change.
  6. Compare its four CSV outputs with the downloads below.

Data dictionary

fiscal_year:
VA fiscal year; FY2026 = October 1, 2025 to September 30, 2026.
cash_out, irrrl:
Count of VA-guaranteed loans VA classified as cash-out or IRRRL.
refinances:
cash_out + irrrl.
cash_out_share:
cash_out ÷ refinances, unrounded decimal (0 to 1).
cash_out_principal_usd, purchase_principal_usd:
Total loan amount in nominal dollars, as published by VA.
purchase_count:
Count of VA-guaranteed purchase loans.
official_state_file:
URL of the VA workbook used for that fiscal year.
county, state:
VA’s county label split at the final comma; “Unassigned” marks the one unlabeled FY2025 row.
weight_*, share_*, within_pp, mix_pp:
Jurisdiction weight, share and decomposition terms (percentage points).
minimum_refis_each_year, matched_counties, count_rose, share_fell, both:
County cutoff and counts of matched counties meeting each condition.

9What does this mean for a homeowner?

The FY2025–FY2026 VA data show more cash-out refinances inside a refinance market that grew even faster through streamlines; they do not show whether a cash-out is right for any household.

The share of refinances that were cash-outs fell, but the number rose. Most states and most large counties showed the same combination. Over twelve years, the share has swung widely, mostly because IRRRL volume rises and falls more sharply than cash-out volume.

For an individual decision, the useful lesson is about balance. A cash-out reprices the whole existing loan. Whether that helps or hurts depends on the old rate, the new rate, the size of the balance, the costs and how long the loans are kept. A fair comparison sets out cash flows, fees and remaining balances for each option over the same period.

Before comparing options, it helps to have five numbers in hand: the current balance, the current rate, the cash needed, the rates actually available for each structure, and the total costs of each. With those, the break-even rate gives a first answer, and a side-by-side table over the expected holding period gives a fuller one.

Several questions are worth asking along the way. Is a funding fee due, or is the borrower exempt? Would a second loan’s rate be fixed or variable? How long might the home be kept? Would restarting a 30-year term raise the total interest paid even if the monthly payment falls? No public data set answers these for a specific household.

Finally, the research questions remain open. Future public releases could show whether fiscal 2026’s mix persists. Data on rates, cash proceeds and loan performance, if they become public in aggregate, would allow stronger analysis than counts alone can support.

Authorship and commercial disclosure. This study was written and calculated by Bolt Home Loans LLC. Bolt originates VA cash-out refinances and has a financial interest in these transactions. VA published the source data and has not reviewed or endorsed this analysis. The study has not been peer reviewed and was not sponsored by any third party.

Downloads

References

  1. U.S. Department of Veterans Affairs. Loan Volume by State. Annual workbooks FY2015–FY2026 (see Table 3).
  2. U.S. Department of Veterans Affairs. Loan Volume by County. fy25-counties-total.xlsx; fy26-counties-total.xlsx.
  3. U.S. Department of Veterans Affairs. Cash-out refinance loan. VA.gov.
  4. Federal Reserve Bank of Philadelphia. Working Paper 26-01. https://doi.org/10.21799/frbp.wp.2026.01.

Cite this study

Bolt Home Loans. (November 20, 2026). Leverage in VA Refinancing: How the balance you refinance changes the cost of borrowing more (Public edition 1.0). Bolt Home Loans LLC. https://bolthome.loans/research/va-refinance-leverage

See also the VA IRRRL vs. VA cash-out comparison and Bolt’s editorial standards.

For educational purposes only. Not a commitment to lend and not a loan offer. Rates, terms, fees, and program availability change and vary by borrower, property, and state. All approvals subject to underwriting and lender guidelines. Bolt Home Loans LLC · NMLS #2784913 · Equal Housing Opportunity.