What Bal Harbour Condominiums Have Actually Returned: 621 Resales, 2006 to 2026
By Ariel Lima ·
I am a real estate investment advisor at Douglas Elliman Real Estate, operating under the ViveFlorida brand, with twenty years in the Bal Harbour and Sunny Isles Beach oceanfront corridor. I earn a commission when a client closes, and every building in this record is one I transact in. Last verified: October 16, 2026.
Do oceanfront condominiums in Bal Harbour make money? Over 621 completed round trips — the same apartment bought and sold again — the median result was +29% after selling costs, and 16% of them lost money. Both figures are computed from recorded closing prices, not reported by anyone.
The number that matters more is what separates the two groups. Of the 84 apartments held ten years or longer, six lost money. All six were bought inside the same twenty-five months, between November 2006 and December 2008. The other 60 long holds — spanning twelve of the thirteen buildings — produced no losses at all.
That is the finding this page exists to publish. Everything below is how it was built, where it breaks, and what it cannot tell you.
What this record is
A repeat sale is the same apartment, bought and sold again, both legs arm’s length. It is the cleanest measurement available in residential real estate, because the apartment is its own control: no adjustment for floor, line, view, size or finish is required, since those are identical on both legs.
621 residence pairs across Bal Harbour’s thirteen oceanfront properties — fourteen condominiums, because the St. Regis address holds two of them under separate declarations — plus 28 hotel-condominium pairs at One Bal Harbour counted separately. The record begins in 2006, which is as far back as the closing data runs.
Four rules, applied before anything was computed:
A 6% selling cost is deducted from every resale. Commission, documentary stamps and the ordinary costs of closing. Gross appreciation is not a return; what the seller kept is. Every figure on this page is net of that deduction, and every one of them is computed rather than recorded.
Pairs under six months are excluded. They measure something other than holding an asset.
Cabanas, hotel-condominium units and combined apartments are held separate from their tower stacks throughout. A cabana is not a residence. A hotel-condominium unit at One Bal Harbour is a different asset class recorded in a different condominium. Two apartments joined into one have no comparator.
Both legs must be arm’s length. One pair at The Palace is excluded from every figure here pending a deed read — an eight-month round trip at +67% gross that the 2014 market does not explain, and that is either a gut renovation or a first leg that was not arm’s length. It is quarantined rather than deleted, because deleting it without reading it would be a choice about the answer.
The holding period
This is the central result, and it is not what most people expect.
| Held | Pairs | Median net | Lost money |
|---|---|---|---|
| Under 3 years | 163 | +16% | 19% |
| 3 to 6 years | 194 | +25% | 21% |
| 6 to 10 years | 180 | +33% | 13% |
| 10 years and over | 84 | +69% | 7% |
All medians are computed after the 6% selling cost.
The first six years are close to a coin flip, and they do not improve as you move through them. Nineteen percent of sub-three-year holds lost money; twenty-one percent of three-to-six-year holds did. The band that sounds safer is marginally worse.
The improvement arrives after six years and sharply after ten. It is not a gradual curve, and anyone describing it as one has not looked at the three-to-six-year row.
What this means for a buyer is a condition of entry, not a prediction. A ten-year floor more than halves the loss rate by itself — 7% against 16% across all pairs — and of the variables on this page it is the one that sits inside the buyer’s control. Someone who cannot commit a decade is buying a different asset than they think they are, and saying so is part of the work.
The finding that governs
Every ten-year loss in the complete record was bought inside the same twenty-five months.
10-year holds bought Nov 2006 – Dec 2008 24 pairs 6 at a loss 25% median +62% 10-year holds bought at any other time 60 pairs 0 at a loss 0% median +70%
Six losses. All six purchases fall between 30 November 2006 and 29 December 2008. Four of the six are apartments in a single building.
The sixty clean long holds span twelve of the thirteen buildings — Oceana, completed in 2016, cannot yet produce a ten-year pair. Across all 621 pairs the same window reads 31% at a loss, against 14% for purchases from January 2009 onward.
The circularity, stated plainly, because it is the first thing a careful reader will test. The window was drawn around the six losses. That those six fall inside it proves nothing by itself — any six points define a range.
Two things are not circular, and they are what carry the finding. The first is the converse: sixty long holds outside that window, across twelve buildings, with zero losses. The second is that November 2006 to December 2008 is the ordinary dating of the last cycle’s peak, arrived at long before this file existed and not by anything in it. The window was not discovered here. It was already named, and the losses landed in it.
Building by building
| Building | Delivered | Pairs | Median net | Median hold | Lost money |
|---|---|---|---|---|---|
| Bal Harbour 101 | 1977 | 38 | +54% | 5.3y | 3 of 38 · 8% |
| The Kenilworth | 1975 | 43 | +46% | 5.3y | 7 of 43 · 16% |
| Bal Harbour Tower | 1990 | 9 | +42% | 10.0y | 1 of 9 · 11% |
| Balmoral | 1977 | 86 | +36% | 6.0y | 8 of 86 · 9% |
| One Bal Harbour, residences | 2007 | 47 | +34% | 7.5y | 5 of 47 · 11% |
| Harbour House | 2007 | 164 | +30% | 5.0y | 21 of 164 · 13% |
| The Palace | 1994 | 9 | +24% | 7.5y | 2 of 9 · 22% |
| Majestic Tower | 1998 | 23 | +23% | 6.5y | 5 of 23 · 22% |
| The Tiffany | 1982 | 29 | +22% | 4.8y | 5 of 29 · 17% |
| St. Regis, North and South | 2011 | 31 | +16% | 3.0y | 9 of 31 · 29% |
| Oceana | 2016 | 18 | +14% | 2.9y | 5 of 18 · 28% |
| The Plaza | 1965 | 97 | +11% | 4.7y | 20 of 97 · 21% |
| Bellini | 2004 | 14 | +11% | 4.9y | 6 of 14 · 43% |
| St. Regis Center | 2011 | 13 | +10% | 4.9y | 5 of 13 · 38% |
| One Bal Harbour, hotel-condominium | 2007 | 28 | +40% | 6.0y | 3 of 28 · 11% |
Read the pairs column before the median. A 43% loss rate on 14 pairs and a 13% loss rate on 164 pairs are not the same kind of fact, and treating them as comparable is the easiest mistake to make with a table shaped like this one.
Read the median hold column second. The two newest buildings sit at the bottom of the return column and carry the shortest holds — 2.9 and 3.0 years. A building delivered in 2016 cannot produce a long hold yet, so its figure describes short-hold behaviour in a recent market and nothing else. It is not a verdict on the building. Where a building’s full transaction history matters more than its eighteen repeat sales, it is published separately: every Oceana sale since delivery covers 126 closings against the 18 pairs that appear here, and the building’s verified unit count, areas and recorded works are set out on the Oceana Bal Harbour page.
St. Regis appears twice because it is two condominiums. The North and South residences and the Center tower are separate declarations with different rules, and the record separates them: 29% at a loss against 38%, same address. A listing will not make that distinction. The recorded declaration does.
Loss rates run from 8% to 43% across the fourteen condominiums. Half a mile separates one end of that range from the other.
Where the pattern breaks
The Plaza is the one building of the thirteen where holding did not protect the buyer. Under three years, 8% at a loss. Ten years and over, 33%. Inverted at both ends, on 97 pairs — the third-deepest sample in the record, so this is not a small-sample artifact.
Four of the six decade-long losses in the entire corridor are Plaza apartments, all bought at the peak. Its four strongest long holds — +156%, +123%, +116% and +111% — were bought between 2011 and 2013. Same building, same decade of holding, different entry. It is the clearest illustration on this page that entry level does work the ten-year rule does not do for you.
Bellini carries the highest loss rate at 43%, on 14 pairs. Five of its six losses are three-to-six-year holds bought between 2013 and 2017. Both of its ten-year holds are positive. On a sample that size the loss rate is a description of those fourteen apartments, not a property of the building.
Price does the work, not income
| Held | Pairs | Price | Income at 2%/yr | Total | Price share |
|---|---|---|---|---|---|
| Under 3 years | 163 | +16% | +4% | +19% | 82% |
| 3 to 6 | 194 | +25% | +9% | +33% | 74% |
| 6 to 10 | 180 | +33% | +16% | +49% | 68% |
| 10 and over | 84 | +69% | +25% | +94% | 74% |
The income column is computed at a flat 2% annual net yield, which is near the middle of what the corridor’s measured net yields actually run: 1.15% to 2.25% after association fees and property tax, on a cash purchase with no debt, with the association’s master policy inside the fee. A financed buyer carries lender-required coverage on top and is not described by these figures.
Across all 621 pairs, price contributes +29% against income’s +10% over a 5.1-year median hold. Price is 74% of the total return.
One limit on that income column, stated because it cuts against the figure. It is computed at a positive rate by assumption. In practice a unit that lets for only part of a year can produce a negative net result after association fees and tax, and the corridor’s measured figures show exactly that at short furnished occupancies. The column describes a unit that lets on an annual basis, not one let seasonally.
What survives that limit is the proportion. Price movement is roughly three times the size of the income leg over a typical hold, in either direction. A buyer evaluating this corridor primarily on rental yield is examining the smaller quarter of the outcome.
Which rules actually change the outcome
Each rule applied to all 621 pairs, added one at a time, in the order of how much work it does:
every pair 621 pairs median +29% 16% at a loss + held ten years or more 84 pairs median +69% 7% + not bought in the Nov 2006 – Dec 2008 window 60 pairs median +70% 0% + in a condominium with a sub-13% loss rate 40 pairs median +70% 0%
Taken alone rather than stacked: the ten-year hold gives 7% at a loss, the building filter 11%, the peak-window filter 15%, against a 16% baseline.
The fourth line is the one worth sitting with. Choosing the building adds nothing once the first two rules hold — the same median, the same zero losses, on a third fewer apartments. That is an uncomfortable result for someone who publishes building-by-building analysis, and it is published here because it is what the record says.
What it does not mean is that the building is irrelevant. It means the building matters most to the buyer who cannot hold ten years, which in practice is a large share of them. For that buyer the spread between 8% and 43% is the whole decision.
Five ideas tested against this record and withdrawn
Each of these was plausible, was tested, and failed. They are published because the tests that fail are the ones a reader cannot reconstruct, and because a record that only ever confirms its author is not a record.
“Entry year is what matters.” Withdrawn. Holding period does more work, and the two interact — the peak-window effect appears almost entirely in long holds, which is the opposite of what the idea predicts.
“Hold four years and you will not lose money.” Withdrawn, and it is worse than wrong. The three-to-six-year band is the weakest in the set at 21% at a loss — higher than the sub-three-year band it was supposed to improve on.
“The cheaper buildings outperform.” Withdrawn. Price does not order the loss column in either direction. Two of the least expensive buildings in the record sit at opposite ends of it — one at 8% and one at 21% — and the two dearest sit at 28% and 29%, which looks like support for the idea until you notice both are recent deliveries whose pairs are all short holds. Price at entry is confounded with entry year, and for that reason no price-per-foot column appears on this page.
“The declaration’s leasing rule predicts how often a building trades.” Not supported. Measured over each building’s own lifetime the bands separated cleanly, which looked like a finding. Re-measured over a single common window — 1 January 2020 to 12 October 2026, controlling for building age and market era — the bands overlap substantially. An exact permutation test across all 495 ways of splitting twelve buildings returns a one-sided p of 0.038. The group means still order correctly and the strictest building is still slowest, so the tendency is real; it is not a rule that classifies an individual building, and the earlier claim that it was has been withdrawn.
“The leasing rule predicts returns.” Not supported. Six-month buildings show 14% at a loss against 19% for twelve- and twenty-four-month buildings, but on only 70 pairs, and the direction reverses at Bal Harbour Tower.
What this record cannot tell you
Everyone still holding is invisible. This method sees completed round trips only. An owner sitting on a loss who has not sold does not appear. That makes every loss rate on this page a floor, not an estimate — the true rate is higher by an unknown amount, and it is highest in the short bands, where the people who would have sold at a loss are most likely to still be waiting.
The most recent buyers cannot be assessed at all. Purchases from 2022 onward that have already resold did so at a median hold near two years, which is close to a definition of a motivated seller. That group’s figures describe motivated sellers, not 2022 buyers, and are not published here for that reason.
It carries no forecast. The record states what happened to people who have already sold. Nothing on this page claims any pattern continues, and the twenty-five-month window is a description of the past, not a tool for recognising the next one. The peak cannot be called forward. What can be done is to price a purchase against a building’s own closed record instead of against its asking prices — a measurement rather than a prediction, and the only defensible meaning of “the right entry level.”
It says nothing about any individual apartment. A building median is not a floor plan. Line, elevation, exposure and condition move a result further than the building median does, and none of them are in this method, because the method’s whole strength is that it holds them constant.
Four buildings cannot support a carrying-cost figure and so the income column above is corridor-level rather than building-level where that matters.
Questions this record answers
Do Bal Harbour condominiums go up in value?
Over 621 completed resales the median was +29% after a 6% selling cost, over a median hold of 5.1 years. Sixteen percent lost money. Both figures are computed from recorded closing prices.
How long should I hold a condominium in Bal Harbour?
The record supports a ten-year floor as a condition of entry. Holds of ten years or more show 7% at a loss against 16% across all pairs, and holds under six years show 19% to 21%. The improvement is not gradual — the three-to-six-year band is slightly worse than the band below it.
Can you lose money buying a condominium in Bal Harbour?
Yes. One hundred and two of the 621 completed resales lost money after selling costs. By condominium the rate runs from 8% to 43%, and because owners holding at a loss who have not sold are invisible to this method, every one of those rates is a floor.
What happened to people who bought at the top of the market?
Twenty-four apartments bought between November 2006 and December 2008 were held ten years or more. Six of them lost money — a 25% loss rate against 0% for the sixty long holds bought outside that window. The median for the peak group was still +62%, so most of them recovered; a quarter did not.
Which Bal Harbour building has the strongest resale record?
On loss rate, Bal Harbour 101 at 3 of 38. On median net return, the same building at +54%. Both are computed on 38 pairs over a 5.3-year median hold, and neither figure predicts the next sale in that building.
Is rental income or appreciation the bigger part of the return?
Price, by roughly three to one. Across 621 pairs price contributed +29% against approximately +10% from income over the median hold, making price about 74% of the total. Measured net yields in the corridor run 1.15% to 2.25% on a cash purchase after association fees and tax.
What is a repeat-sales analysis?
It measures the same property bought and sold again, so the property acts as its own control and no adjustment for size, floor, view or finish is needed. It is the method behind the major national house-price indices, applied here to one half-square-mile corridor.
Why is a 6% selling cost deducted?
Because gross appreciation is not a return. Commission and closing costs come out of the seller’s proceeds whether or not an analysis accounts for them. Every figure on this page is net of 6% on the resale leg.
Does a cheaper building perform better?
Not in this record. Price does not order the loss column in either direction — two of the least expensive buildings sit at 8% and 21%, and the two dearest sit at 28% and 29% only because both are recent deliveries whose pairs are all short holds. Entry price is confounded with entry year, and no price-per-foot column appears on this page for that reason.
Does a building’s leasing rule affect what owners earn on resale?
Not detectably. Six-month-minimum buildings show 14% at a loss against 19% for stricter buildings, but on 70 pairs and with the direction reversing in one building. The claim is not supported by this record.
Why does St. Regis appear twice?
Because the address holds two separate condominiums with different recorded rules. The North and South residences show 29% at a loss; the Center tower shows 38%. Pooling them would hide a real difference between two assets a listing describes identically.
What does this analysis not cover?
Owners who have not sold, buyers since 2022, any individual apartment’s prospects, and anything forward-looking. It is a record of completed transactions and makes no claim about what happens next.
Method and sources
Every closed sale for each of the thirteen addresses was pulled from the South Florida MLS over the longest window the system permits, approximately twenty-seven years; the usable record begins in 2006. Pairs were matched by apartment identity, both legs required to be arm’s length, and a 6% selling cost applied to the resale leg. Pairs under six months were excluded. Cabanas, hotel-condominium units and combined apartments were held separate from their tower stacks throughout.
Building identity and the separation of St. Regis into two condominiums were taken from the recorded declarations at the Miami-Dade Clerk of the Court and Comptroller. The index was searched building by building on October 12, 2026, and on that date the Clerk’s database reported coverage of instruments recorded through October 7. Anything recorded after October 7 is outside this work, not anything after October 12.
Three data-quality findings worth stating, because they affect what can be claimed. The MLS year-built field is agent-entered and does not agree with itself — one building carries 1977 on 107 rows and 1978 on 37 in a single export. The recording date of a declaration is the source used here for a building’s year, not the MLS field. One sale was excluded as not credible: a $210 recorded price against a $199,000 asking price. Duplicate listing numbers appear in the exports and were deduplicated before matching.
All figures are computed from recorded closing prices. None is reported by a developer, an association or a brokerage.
An opinion, marked as one
Everything above is a measurement. This paragraph is not.
I think the ten-year floor is underused as a filter, and that most of the loss in this record was avoidable not by better timing but by a longer stated horizon at the point of purchase. The buyers who lost money were, with few exceptions, buyers whose circumstances changed inside six years. That is not a market risk. It is a planning one, and it is the part of the decision an advisor can actually affect.
I also think the corridor is mispriced on income and correctly priced on scarcity, which is why I spend more time on the declaration and the entry level than on the rent roll.
Those are judgments. The 621 pairs are not.
Applying this
The method here is not specific to Bal Harbour. Any submarket with enough repeat sales to matter can be read the same way — matched pairs, selling costs deducted, holding periods banded, the peak window tested rather than assumed. This analysis can be applied to any neighbourhood between West Palm Beach and Coconut Grove, Sunny Isles Beach included. What it would show there, I do not know; I have not run it.
That is the part worth saying plainly: this is built to order, and the published version is only the version I chose to publish.
The question someone actually brings is almost never the question a page answers. It is narrower, and it is theirs. Which of two buildings has held value better for the people who actually sold. What a ten-year hold has looked like in a submarket that is not covered here. Whether one line behaves like the building it sits in, or nothing like it. What the record says about a specific purchase year, a specific size band, a specific price tier. Those are all the same analysis pointed at a different set, and the work behind them is the work behind this page.
Three things, then.
Your own apartment’s position in this record. If you own in one of the thirteen, I will tell you what the pairs in your building and your holding band actually did, and where your purchase year sits against the twenty-five-month window. Send the building and the year you bought — that is enough, and you do not need to say anything else. No cost, and nothing is asked of you.
The pair-level detail behind any row above. Which apartments, bought when, sold when, held how long. That is not published, because an individual apartment’s round trip is a conversation and not a page. It exists for all 621. No cost.
The same record, built around a question that is not on this page. A building, a submarket, a line, a price tier, a holding period you have in mind — scoped to what you are actually deciding, with the sources named and the limits stated the same way they are stated here, including the ones that cut against the answer. That is a conversation rather than a form, because the first thing to establish is whether the data exists at the depth your question needs. Sometimes it does not, and saying so early is cheaper for both of us than a thin answer.
If you are an adviser and the decision belongs to a client, the arrangement works differently and is set out on how this practice works alongside a wealth advisor. Nothing reaches your client that you have not seen first.
I build the same reading for any building in the corridor, and the same method anywhere the record supports it.
The monthly corridor analysis. One a month.