The short version
A life insurer and a property fund took Rs 51 billion from the Mauritian public. They lent most of what they kept to companies in their own group, wrote those loans up in value, and recorded the write-ups as profit — which is what made the next round of selling possible.
It is a circle, not a chain. That is the single most important thing about it, and the reason it ran for eight years while everyone got paid on time.
What this post is not. The report I am working from was commissioned by one of the two regulators that seized the group. The people it criticises were never given a right of reply. It makes no finding that any offence was committed by anyone, and no court has ruled on the substance. I have kept every hedge it uses.
Part 0The whole machine on one screen
Everything below is detail. This is the shape of it. Read the circle first — the arrow coming back up the right-hand side is the entire trick.
The mechanism, ¶225–233 and Figure 16 (¶230). Every figure in this diagram is the report’s. The circle is mine — the report draws it as three steps, but step three is what makes step one possible again, and a list cannot show that.
Five stages, in the order they happen. The rest of this post walks each one.
| Stage | What it does | Where |
|---|---|---|
| 1 · Raise | Sell certainty — capital back, a guaranteed rate, life cover, tax-free | The insurer |
| 2 · Move | Lend to companies you control, and call it an investment | The circle |
| 3 · Dress | Write those investments up. Record interest that never arrives | The insurer |
| 4 · Pass | Get through the solvency test, the capital test and the returns | The bank |
| 5 · Hide | Stop publishing when the numbers get too bad to publish | Going dark |
How to read the coloured strips. Every technique below gets one of three markers. Lawful on its face means the thing itself is an ordinary commercial act. Breaches a regulatory limit means a rule was crossed, with the rule named. Would be tested against a criminal provision means there is a section of an Act it could be charged under — not that it was, and not that anyone was convicted. Nobody has been convicted of anything arising from this report.
Part 1Two kinds of broke
Before any of the machinery makes sense, one idea has to land. It is the reason a company can be finished for five years while every customer is paid on the day they were promised.
There are two different questions you can ask about a business.
1. Does it own more than it owes? Sell everything, pay everyone — is anything
left? If not, it is balance-sheet insolvent.
2. Can it pay today’s bills today? If yes, it keeps trading, and nobody outside
sees a problem.
These are not the same question. A company can fail the first for years while passing the
second every single day — as long as new money arrives faster than old promises mature.
BAI failed the first test from 2010. The hole got ten times bigger in three years, and the group kept paying on time until the month it was seized.
¶5. The report adds that even those asset values were too high, so the real hole was bigger than either bar.
“Even though the BAI Group was balance sheet insolvent, it managed to continue operating primarily because it was able to raise enormous amounts of funds from the public to pay off the creditors, policyholders and investors when its obligations fell due.”
nTan report, ¶7Every customer who cashed out between 2010 and 2014 got their money. To them that was proof the company was sound. It was proof of nothing except that they were early.
Why nobody had the whole picture
Three companies could take money from the public. Each answered to a different Act and a different supervisor, and the only place all three were visible at once was a holding company in the Bahamas that publishes nothing.
| Company | What it sold | Governed by | Watched by |
|---|---|---|---|
| BA Insurance | Savings policies — Rs 45.8bn | Insurance Act 2005 | FSC |
| BPF | Preference shares — Rs 5.4bn | Securities Act 2005 | FSC, different regime |
| Bramer Bank | Deposits | Banking Act 2004 | Bank of Mauritius |
| Klad, Bahamas | — | nothing | nobody |
Split a group so each money-raising arm sits under a different rulebook. Every supervisor then sees a complete, well-behaved picture of its own piece. None of them can see that all three pieces point at the same set of related companies.
How it was used hereAn insurance supervisor saw premiums. A securities supervisor saw a fund. A central bank saw a bank. The consolidated view existed only at the Bahamas parent — and in 2012 the group tried to move its accounting consolidation there too ¶216.
Part 2The circle
The report calls what the insurer and the fund were running “Ponzi-like schemes”, and it hedges twice in one sentence — “Ponzi-like”, and only “for the larger part of the Review Period” ¶17. It never says when it started. Most news coverage dropped both hedges. I am keeping them.
| Step | What happens |
|---|---|
| 1 · Raise | Sell products promising high returns at low risk ¶19 |
| 2 · Spend | Pay earlier investors their returns. Send the rest to companies in the same group ¶20 |
| 3 · Write up | Declare those group investments worth more. Book it as profit ¶21 |
Step 3 is not the end. Step 3 is what makes step 1 possible again. The paper profits produce a company that looks strong, which is what attracts the next round of money and persuades existing customers to roll over instead of taking cash. The report’s own phrase is “vicious cycle”. A cycle that must keep accelerating has an end date built into it, whatever anyone intends.
And buried in a footnote is the sentence that explains why nobody stopped it for eight years:
“This impression was reinforced by the timely payment of returns, using fresh funds raised from the public.”
nTan report, ¶21 footnote 12Read it twice. The evidence that the company was sound was the mechanism itself. Paying people on time was not something the scheme managed despite being a scheme. It was the advertising, and it was funded by the next person through the door.
The three steps at ¶19–21, drawn as a loop. The steps and every figure are the report’s. Drawing them as a closed circle is mine — the report sets them out as a list, and a list cannot show that step three is what funds step one.
So is it a Ponzi scheme, or not?
The report gives its own definition in a footnote, and it is worth having in front of you before comparing anything:
Returns to existing holders were “paid out of funds raised from new policyholders and investors”; the returns promised were “higher than were sustainable”; the business “had to attract new investments at an ever-increasing rate”, and would “inevitably collapse under the weight of their liabilities”.
nTan report, ¶17 footnote 6 — the report’s own definition of what it means by “Ponzi-like”Now hold that against the textbook article. The differences are not cosmetic.
The distinction is not pedantry — it decides what can be charged. “Ponzi scheme” carries fraud from inception inside the word, so it points straight at a criminal provision. “Ponzi-like” describes a result that can be arrived at by degrees: an over-promised product, then a bad year, then a decision to value your way out of it, and at some point the new money is paying the old and nobody wrote that down as a plan. Where that line was crossed is the one question the report does not answer, and it is the reason its language is so careful.
It is also why the useful test is a ratio rather than a judgement. You cannot see intent from outside a company. You can see that the assets produced 3.3% in cash while the products promised up to 12.75% — and that gap has to be filled from somewhere. Part 9 works that check on the published numbers.
Move value out of a company to the people who control it, using transactions that each look ordinary on their own — a loan, an asset purchase, a fee for a service. The term comes from the academic literature on controlling shareholders. Nothing has to be secret, only unexamined.
How it was used hereAt least Rs 10.8bn from the insurer and Rs 2.8bn from the fund went to other companies in the group, recorded as investments ¶20 fn11. “At least” is the report’s word, in both lines.
A director of an insurer must “act honestly and in the best interests of the insurer and policyholders”, and must “exercise care, diligence and skill”.
Why this section and not the ordinary directors’ duty. Companies Act 2001 s.143 requires a director to act in the best interests of the company — and the company, here, was doing rather well out of the arrangement on paper. This section adds policyholders: a duty that runs past the company to the people whose money it actually is. On that test, sending Rs 10.8bn of policyholder money to companies the same people control is measured against the interests of the policyholders, not the interests of the insurer. Our reading
A saver does not compare headline rates. A saver compares what actually lands in the account. So if the state taxes your competitor’s product and not yours, you can advertise a lower rate than the bank and still hand the customer more money.
How it was used hereReturns on the insurance policies and the preference shares were not subject to income tax, where bank deposit interest was ¶17 fn8.
Nearly identical — and BAI got there while earning almost a full percentage point less. Turn it round and the edge is easier to see. To beat a tax-free 5.05%, a bank has to pay 5.94% gross, because 5.05 ÷ 0.85 = 5.94. And to match the top of BAI’s range — 12.75% tax-free — a bank would have to advertise 15%.Our arithmetic
But notice what the exemption does not do. It makes the product easier to sell. It does not put a single rupee into the investments underneath. BA Insurance still had to find 5.05–12.75% a year in real money to pay these people. It found 3.3% ¶36. The tax edge widened the sales pitch; it never touched the hole.
Part 3The insurer
BA Insurance was the biggest of the three taps and the subject of the longest chapter. One ratio carries the whole case.
That gap is the case. Everything after it is the accounting that hid it.
What they were actually selling
The product was Super Cash Back Gold. Minimum Rs 25,000, paid once. It promised four things ¶32:
- If you die, a payout of up to 110% of what you paid
- At the end of the term, all your money back
- A guaranteed bonus every year, 5.05% to 12.75%
- A final bonus linked to interest rates
Read those together. You get your capital back, a guaranteed return on top, insurance while you wait, and the return is tax-free. At no point is the customer’s capital presented as being at risk. This is a term deposit in an insurance wrapper, and the word “premium” is doing an enormous amount of work. It was about 80% of everything the insurer sold ¶33.
One caution before you use that. The report says the returns beat bank deposits. It says explicitly that it did not compare them to other Mauritian insurers’ products ¶31 fn15. You cannot say from this report that the product was abnormal for its own industry.
How a loss became a profit
Over eight years the insurer reported a profit every single year — small, steady, unremarkable. Rs 2.0bn in total. Here is what the report takes back out.
Figure 3, ¶40. Rs billions, over eight years. Note the third adjustment — +Rs 3.3bn of write-offs added back. The report is not cherry-picking downward.
Three things fall out of that chart. The insurance business itself lost money in all eight years — Rs 8.5bn of operating losses. Every reported year is a profit and every adjusted year is a loss, so this was not a business that drifted into trouble. And the reported profit is small: Rs 0.1–0.5bn a year on a balance sheet heading towards Rs 33.6bn.
A modest, steady profit is a far better disguise than a spectacular one. It invites no questions. That reading is mine — the row is theirs.
You own something. You decide it is worth more than you paid. You write the higher number into your accounts and call the difference profit. No cash moves. This is ordinary accounting, and it is only ever as good as the valuation behind it. When the thing being valued is a company you also control, you are on both sides of the estimate.
How it was used hereRs 11.4bn of fair value gains at the insurer alone ¶42. One asset was valued at 6.5× net asset value on unreviewed management forecasts, with growth counted twice ¶49. A hospital property was priced at Rs 2.5bn by working backwards from what the loss-making seller could afford to pay ¶77, and later supported by a valuer benchmarking a hospital on hotel room keys ¶81–82.
Someone owes you interest. You record it as income the moment it falls due, not when it arrives. That is normal accrual accounting. It stops being normal when the borrower plainly cannot pay and you keep recording it anyway.
How it was used hereRs 4.5bn of interest income on loans to group companies, with the report’s own emphasis: “which remain unpaid” ¶50. The main borrower’s debt grew from Rs 0.3bn to Rs 6.4bn at 11–13% interest, none of it ever paid ¶53.
One day: 31 December 2009
This is the most concrete event in the whole report, and the clearest single technique in it.
An insurer has to hold more assets than liabilities by a set margin, and the regulator tests that on the insurer’s own accounts, not the group’s. BA Insurance was failing that test at the end of 2007 and again at the end of 2008 ¶43(2).
Then, on the last day of the 2009 financial year, this happened — all of it in one day.
¶43(1). Rs 3.6 billion, five legs, one day. The lender’s name is blacked out in the public report. Le Mauricien named it as Banque des Mascareignes, and reported the facility as advanced at 15:45 and returned the same evening — press reporting, not a finding of the report.
By the end of the day the money was back where it started. But BA Insurance’s own accounts now showed Rs 3.6bn more in assets and Rs 3.6bn more in share capital. Assets up, liabilities unchanged. It passed. A deficit at the end of 2007 and 2008 became a surplus from 31 December 2009 onward.
Two people described it at the time.
“No actual cash was injected at the time of this transaction.”
KPMG, in the minutes of BA Insurance’s Audit Committee, 27 March 2012 — quoted at ¶43(5)The investment “appears to have been no more than a guise to round-trip Rs 3.6 billion of funds back to BA Investment.”
nTan report, ¶43(4)Arrange a transaction so that it is sitting on the books on the reporting date, and unwind it immediately afterwards. The picture is true for one day and false for the other 364. The mechanism inside it is a round-trip: money leaves and returns through a loop of parties, so the paperwork records activity the cash never really performed.
The famous case is Lehman Brothers’ Repo 105 — assets moved off the balance sheet days before each quarter-end and brought back after.
How it was used hereRs 3.6bn borrowed for one day, injected, “invested”, returned and repaid, all on 31 December 2009. Net cash nil. Every leg cancels when you consolidate the group — but the Rs 3.6bn survives on the insurer’s own balance sheet, which is the one the solvency rules test.
The auditor knew, and Rs 35 billion arrived afterwards
KPMG audited BA Insurance from 2004. The report found two of its presentations to the audit committee — March 2011 and March 2013. Both listed the same problems, among them that the insurance business was operating at a loss, that it relied on fair value gains to show a profit, that it was solvent only because of the 31 December 2009 injection, and that it relied on renewals to pay maturities ¶89.
KPMG’s own management letter said the company “is not using policyholders’ funds appropriately” ¶94. Mr Dawood Rawat was in the room for the 2013 presentation ¶95.
KPMG signed a clean audit opinion for both years ¶91.
¶97: BA Insurance then “had free rein to raise some Rs 35 billion from FY2010 to FY2014 in policyholders’ funds”. Against Rs 45.8bn raised across the whole period, that is about three-quarters of every rupee the public ever put in — arriving after the auditor had the whole picture in writing.
Part 4The fund
Bramer Property Fund raised Rs 5.4bn. It was sold as a property fund. It was not one.
What BPF actually sold. You give it Rs 100. It promises a fixed
7.3% to 20.62% every year. It promises your Rs 100 back at the end.
That is a loan. Your money does not rise and fall with the buildings. You are not an
owner. You are a lender. The report’s own words: the shares “resembled promissory
notes” ¶100. A promissory note is an IOU.
Which raises one question, and it is the whole chapter: if you owe the public Rs 5.4 billion, where does that show up in your accounts?
The one entry
A balance sheet has two sides. On the left, what you own. On the right, who has a claim on it — split between debt (money you must give back) and equity (what is left, and is genuinely yours).
BPF presented all three classes of its shares together, under one heading — “net assets attributable to holders of redeemable shares”. The preference shares were redeemable at par: a fixed amount, repayable. They were debt.
¶111(2), at 31 December 2014. The rule is IAS 32, which decides which side of the balance sheet an instrument belongs on by asking one question: can the holder force you to pay? Here the answer was yes, at a fixed price.
Rs 74 million is nothing. It means one bad investment tips the fund into insolvency — owing more than it owns. The report shows that happening twice: at the end of 2010 and again at the end of 2014.
Put money you are obliged to repay on the side of the balance sheet reserved for money that belongs to you. Nothing about the cash changes. What changes is every ratio anyone computes from the accounts — gearing, net assets, solvency, and whether the entity looks capitalised at all.
How it was used hereRs 3.7bn of “net assets” was, properly presented, about Rs 3.6bn of debt owed to the public plus Rs 74m of actual equity ¶111. The fund was not thinly capitalised. On its own books it was almost entirely borrowed money.
Where the money went
Over eight years, Rs 6.7bn came in and Rs 6.7bn went out. Two of those numbers matter more than the rest.
| In | Rs bn | Out | Rs bn |
|---|---|---|---|
| From investors | 5.4 | Returns paid to investors | 1.2 |
| Income actually earned | 0.6 | Original money repaid | 1.8 |
| Selling property | 0.7 | To companies in the group | 2.8 |
| Everything else | 0.9 | ||
| Total | 6.7 | Total | 6.7 |
Split the investor payments, because only one half is a return. BPF earned Rs 0.6bn in cash over eight years, and paid out Rs 1.2bn in returns. It paid out twice what it earned — before repaying a single rupee of anyone’s original money. Where did the difference come from?
“To cover the cash shortfall, BPF used funds raised from new investors to pay off existing investors.”
nTan report, ¶105The public paid to buy the public out
Of everything that left the fund, one transaction matters most. In May 2010, BPF lent Rs 450 million to Seaton, a company at the top of the group, “using funds raised from public investors”. Seaton used it for one purpose: to buy out the outside shareholders of BA Investment, the group’s listed holding company, and take it private ¶117, fn54.
¶117, footnote 54 and ¶214. Klad, the Bahamas parent, later took over the bonds. It never paid a single year’s interest — and BPF recorded the interest as income anyway, Rs 236m of it ¶118–119.
The classic direction for value extraction is outward, to the controlling family. This one runs upward: the money buys out the remaining outside owners, so control of the whole structure concentrates — financed by the very public being bought out.
How it was used hereRs 450m of retail investors’ money → Seaton → the buy-out of BA Investment’s minority → delisting in October 2010. The report’s footnote: “an example of using BPF’s publicly-raised funds to accomplish the BAI Group’s own purposes” fn54.
And when the regulator finally stopped it
The FSC licensed BPF in January 2014, after years of operating without one, with conditions: no new investors, no new money, no new investments. In the four weeks before the licence, BPF raised another Rs 128 million fn57.
Months later the group began raising money through a different company — HCL, unlicensed, same manager. It took in Rs 188m and passed Rs 111m to BPF. The report calls it the “Lernaean Hydra” ¶131–134: cut off one head, two more grow.
And the auditor. BDO signed a clean opinion on BPF’s 2014 accounts on 26 March 2015. Twelve days later it tried to take that opinion back. BPF went into administration in April ¶150, fn72.
Part 5The bank
Bramer Bank is the interesting one, because it is the only tap with a hard legal limit and a regulator reading its numbers every three months. So the technique had to change shape.
First, what “core capital” is. A bank runs on two
kinds of money: depositors’ money, which it owes back, and its own money
— what shareholders put in plus profits it kept. The second pot is core capital,
or Tier 1. It is the cushion; losses eat it first, and depositors only feel anything once it is
gone.
The rule: for every Rs 100 of its own money, a bank may lend at most
Rs 60 to companies in its own group — and must report that figure to the Bank of
Mauritius every quarter ¶189.
Why the cap exists: if your own group collapses, you lose the money and the
cushion in the same week. The two risks are not independent.
So the bank had far less room to move money to its group than the insurer or the fund did. The report says so itself ¶153. So it did not lend to its group. It bought things from it — a book of customer debts, and a warehouse of stock. Both are purchases. Neither counts as lending.
¶158–160. Except for one detail: Iframac kept collecting the money and was meant to pass it on. It stopped. The bank had the contractual right to force it and — in the report’s words — “for reasons unknown to us” never used it ¶169.
So the asset ended up exactly where the structure was designed to avoid. In a table later in the report that is not blacked out, it appears under its real name: “Account receivables from Iframac” — Rs 2,486 million Figure 16, ¶230. Not owed by the public. Owed by a company in its own group.
A limit is written to catch a form — here, credit. So do the same economic thing in a different form. Buy an asset rather than lend against it. The cash out of the door is identical; the line it lands on is not.
How it was used hereTwo devices. The bank bought Iframac’s hire-purchase book across twenty agreements from July 2013. And it ran “floor plan financing” — buying cars and retail stock, holding title, selling them back at a mark-up. Approval was sought for a Rs 100m, three-month facility to one dealer; it became Rs 688m over five years to three, none of the increases notified in advance ¶187. And from 2008 to 2015 those facilities went only to related parties ¶175.
Money out, capital back, the same day
Capital comes in two grades. Tier 1 is core — shares and retained profit, absorbing losses while the bank still trades. Tier 2 is the second layer: long-dated borrowing that ranks behind depositors and only absorbs losses in a wind-up. Both count toward the capital adequacy ratio, the headline test of whether a bank holds enough capital against its assets. Tier 2 is the cheap way to lift it.
It happened three times — 2013 with Iframac’s debentures ¶168(3), 2011 with BPF’s preference shares fn80, and 2010–12 when BA Insurance paid Rs 250 million for two bank debentures. That last one is not redacted, and the report states the purpose outright: “to help to increase the Tier 2 Capital of Bramer Bank” ¶84.
Double gearing is the same capital counted twice — one rupee supporting the insurer’s balance sheet and the bank’s. Bank regulators call the instrument a reciprocal cross-holding, or capital funded indirectly by the bank itself. Basel requires both to be deducted from regulatory capital, never counted — precisely because a bank that funds its own capital has added no ability to absorb a loss.
How it was used hereThe report says the 2013 leg meant the bank “inflated its capital”, and footnote 80 says the debentures and preference shares together “contributed to an improvement in Bramer Bank’s Tier 2 capital”. ¶228(4) totals it: more than Rs 1 billion of publicly-raised money went “to raise the share capital and/or capital adequacy ratio” of the bank.
One precision that matters. The 60% related-party cap is measured against Tier 1, so these Tier 2 round-trips did not widen it — they lifted the capital adequacy ratio. What widened the cap was Tier 1: BA Insurance and BPF buying 49% of the bank’s ordinary shares for Rs 615m in December 2010.
Two sets of books
None of it works unless the quarterly return stays clean. One clause makes that deliberate rather than careless:
“In fact, Bramer Bank had (correctly) treated these amounts owing from Iframac as exposure in its internal credit assessments, but conveniently omitted it from its disclosures to BoM.”
nTan report, ¶191(2)The right answer existed inside the bank. A different one went to the regulator. The table showing how far over the limit it really was is completely redacted — every number. But the black bars are drawn on top of the digits, so each bar is exactly as wide as the number underneath it. Measure the bars and you can count the digits. It checks out: the bar over the limit itself is two digits wide, and the report prints that limit in plain text elsewhere — 60%.
| Bramer Bank’s related-party exposure | as a share of Tier 1 |
|---|---|
| The legal limit | 60% |
| What the bank reported | two digits |
| What it should have reported | three digits |
Three digits means over 100%. The bank had not merely broken the cap — its lending to its own group exceeded its entire core capital. Which is why the report reaches for the word “shocking” ¶193.
A limit is only ever as good as the number reported against it. You do not have to lie about a figure — you can simply leave the exposure out of the calculation, and report a true number about an incomplete set.
How it was used hereAmounts owed by Iframac from the hire-purchase deal, the floor-plan facilities and two micro-financing facilities were left out of the quarterly returns, along with debentures the bank had “invested” in its own parent ¶191. Internally, they were counted.
Part 6The money that left
This is the chapter everyone reaches for first, and it opens by telling you how little it established.
“Due to time and information constraints, we have not been able to trace how exactly the funds were utilised by the related parties and exactly where (or, more precisely, with whom) the funds have ended up.”
nTan report, ¶196The report follows the money out of the three taps and into the group companies, and there the trail stops. It says it was “unable to determine” whether any of the Rs 17.3bn ended up with Mr Rawat ¶197. What it can show is a smaller, much more specific number: Rs 930 million.
| Traced to Mr Rawat, relatives, associates, or spent for their benefit | Rs m |
|---|---|
| The “Chairman Current Account” at BA Investment, 2008–2014 | 387 |
| Advances to Seaton — which owned BA Investment and had no operations at all | 280 |
| “Technical fees” to BA Holding, 2009–2014 — also no operations. One quarter’s fee was paid directly to Mr Rawat | 98 |
| Three further rows — entity, description, period and amount all blacked out | 165 |
| Grand Total ¶199 | 930 |
Those three redacted rows are recoverable. The report blacks out every cell but prints the total. 930 − 387 − 280 − 98 = Rs 165m. Always check whether another part of a document discloses what this part conceals.
What the Chairman’s account paid for, from BA Investment’s own management accounts: property Rs 16.9m, a vehicle Rs 5.3m, a boat, Rs 3.0m, furniture Rs 0.8m. And:
“Advances with no meaningful description of their purpose amounted to a staggering Rs 577 million.”
nTan report, ¶199The regulator said stop. So they used a stranger.
In late 2013 the FSC specifically instructed BA Insurance to stop sending money to related parties ¶203.
Enter Logandale — a company with no business of its own, whose former shareholder was an alleged relative of Mr Rawat and whose current shareholder was an employee of BA Insurance. Crucially, it was not disclosed as a related party in BA Insurance’s accounts ¶201.
Figure 14, ¶204. On paper BA Insurance bought a bond from an unrelated company. In fact it moved Rs 50m to its own parent and then to its Chairman — four weeks after being told not to.
A problem was left over: Logandale now owed BA Insurance Rs 50m, and had no business with which to pay. So over four days in January 2015 the money went round in a circle. BA Insurance bought properties from Seaton (Rs 31.4m) and paid “consultancy fees” to BA Investment (Rs 32.6m); Seaton passed its share to BA Investment; BA Investment advanced Rs 55.7m to Logandale; Logandale redeemed the bond ¶206.
The insurer funded the repayment of the debt owed to it. The report’s verdict: the transactions “appear to be part of a scheme which had questionable commercial purpose for any of the entities involved” ¶207.
Layering is the anti-money-laundering word for inserting an extra party into a chain so the two ends stop looking connected. The inserted company is a conduit — no business, no staff, one purpose. What makes it work here is a definitional gap: the rule bites on “related parties”, and this company was not disclosed as one.
How it was used hereA regulator’s instruction not to move money to related parties was met by moving money through a company that was not on the related-party list — and then by having the insurer fund that company’s repayment.
Part 7Going dark
Everything above is mechanism. This is the part that turns mechanism into intent — a document written inside the group, five years before the collapse.
“…it is estimated that the Group will show losses of over MUR 2.5bn by end of 2010… The Group cannot afford to report such losses and we will therefore take bold measures.”
The BAI Group’s Chief Operating Officer, “Transformation Strategy”, drafted April 2010 — quoted at ¶213Read the middle clause again. Not cannot afford such losses. Cannot afford to report them. It was written for three people: the Group Chairman (Mr Rawat), the Vice-Chairman and the Group President & CEO fn104.
A listed company must file audited consolidated accounts — the whole group in one set of numbers — on time and in public. A private company does not. BA Investment was listed.
¶213–217. Had BA Investment stayed listed, its accounts “would have revealed to the public in Mauritius” a Rs 3.3bn loss for 2010 and liabilities exceeding assets by Rs 1.2bn ¶219.
Disclosure obligations attach to the listing, not to the business. Take the company private and the obligation falls away — the accounts stop being filed because there is no longer anyone to file them to. Nothing about the underlying position changes.
How it was used hereOne month after the COO’s paper, Seaton began buying out BA Investment’s outside shareholders, funded with Rs 450m from BPF. Complete by October 2010; delisted 8 October. The report’s conclusion at ¶232: the group’s balance-sheet insolvency “was concealed from the investing public by the privatisation”.
And then they tried to move the accounts offshore
Delisting solved BA Investment. BA Insurance, as a licensed insurer, still had to consolidate somewhere. From the minutes of its Audit Committee, 27 March 2012: the KPMG audit partner advised that consolidating at the level of Klad — the Bahamas parent — would not be appropriate, because Klad was not required by law to prepare or file accounts and so could not satisfy the standard’s test that financial statements be “available for public use”.
Members “commented whether the possibility of having the financial statements of Klad audited and filed with the authorities in Bahamas would then satisfy the criteria… and requested the External Auditors to liaise with their technical team in South Africa regarding the acceptability of the above proposal.”
nTan report, ¶216The auditor said the route did not work. The committee asked the auditor to go and find a way to make it work. Not publishing the group’s true position was not an administrative drift. It was an objective, worked on in a minuted meeting, with the auditors in the room.
Klad’s accounts for 2010, 2011 and 2012 were eventually all signed off on the same day — 17 September 2014, the 2010 set 44 months after the year it covered. The 2012 accounts disclosed a loss of US$119m and liabilities exceeding assets by US$302m — about Rs 9.2bn. KPMG’s opinion on them was not qualified ¶217. They were filed in the Bahamas. Nothing was published in Mauritius, and clean opinions on the Mauritian companies kept coming for another seven months.
Accounting standards let a subsidiary skip consolidated accounts if a parent higher up produces them and they are publicly available. Push the consolidation to a parent in a jurisdiction that requires no filing, and the obligation evaporates on the way up.
How it was used hereAttempted at Klad in the Bahamas, in March 2012. The auditor identified the flaw — the accounts would not be “available for public use” — and was asked to seek a second view from another country’s technical team.
Part 8Every technique, with the law
All fourteen in one place. lawful on its face · breaches a regulatory limit · would be tested against a criminal provision.
| Technique | In one line | Provision |
|---|---|---|
| T1 Regulatory arbitrage | Split the group so no supervisor sees all of it | — |
| T2 Tunnelling | Lend to companies you control, book it as an investment | Insurance Act 2005, s.31(3) |
| T3 The tax edge | Beat bank deposits on the after-tax number | — |
| T4 Fair-value write-ups | Value what you control, book the gain as profit | Insurance Act 2005, returns offence |
| T5 Accrued income never received | Record interest the borrower cannot pay | Insurance Act 2005, concentration limit |
| T6 Window dressing | A round-trip across the reporting date | Insurance Act returns offence; FIAMLA s.14 |
| T7 Liability–equity misclassification | Present money repayable at par as equity | Securities Act 2005, s.116 |
| T8 Tunnelling upward | Investors’ money buys out the investors | Securities Act 2005, ss.115, 97 |
| T9 Form-over-substance | Buy instead of lend, so nothing shows as credit | Banking Act 2004, s.30(1)(a) |
| T10 Double gearing | Fund your own subsidiary’s regulatory capital | Banking Act 2004, s.31 |
| T11 Omission from the return | Record it internally, leave it out of the filing | Banking Act 2004, s.97(21) |
| T12 Layering | Route through a company not on the related-party list | Insurance Act s.31(3); FIAMLA s.3 |
| T13 Going dark | Delist, and the duty to publish falls away | — |
| T14 Jurisdiction shopping | Consolidate where nobody has to file | IFRS consolidation requirements |
Read that column carefully. “Would be tested against” is not “was charged under”, and it is certainly not “was convicted of”. No one has been convicted of anything arising from this report. The report itself makes no finding that an offence was committed — it asks four questions at ¶221–224 and answers none of them.
How each one actually works
One drawing each. The red marks are mine — they point at the step that does the work. Under every drawing is the provision it would be tested against, in the Act’s own words, with the operative phrase picked out.
act honestly and in the best interests of the insurer and policyholders. The policyholder limb is the one that bites: the company looked well served; the people whose money it was did not.
makes a representation or statement which he knows to be false in a material particularor who
recklessly makes a representation or statement which is false. Up to Rs 1m and 2 years.
misleading or deceptive or likely to mislead or deceive; s.116(3) says conduct includes
omitting a material fact. It carries a mandatory minimum — Rs 500,000 and one year — the only one here that does.
employ a device, scheme or artifice to defraudor engage in conduct that
operates as a fraud or deception, or is likely to operate as a fraud or deception. Read the second limb: effect, not only intent.
engage… in the wholesale or retail trade… or in any business other than the business for which the financial institution is licensed. nTan never raises s.30 at all.
significant interest; shares held without it are
null and void. The December 2010 purchase of 49% for Rs 615m sits squarely inside it — and the report never asks whether approval was given.
makes, with intent to deceive, any false or misleading statement or entry or omits any statement or entryin any book, account, report or statement. The omission limb is drafted for ¶191(2) almost word for word.
the proceeds of any crime— which requires a predicate offence first. Insurance Act s.31(3) needs no predicate at all.
And two structural observations that are not techniques at all, but explain a great deal.
Two things the case says about the rules themselves
- A statutory alarm can be wired to the wrong test. Companies Act 2001, s.162 obliges directors to act when the company is “unable to pay its debts as they fall due” — cash-flow insolvency. BAI was balance-sheet insolvent from 2010 and kept paying until 2015. On its face the duty never fired, through five years and a Rs 12bn hole. The alarm was wired to the one test the scheme kept passing — and passing was the mechanism.
- Recklessness is often enough. The Insurance Act returns offence, Securities Act s.116 and FIAMLA s.3 all catch “reckless” or “reasonable grounds for suspecting”. The bar is lower than the report’s careful language implies — which makes the absence of charges more interesting, not less.
Part 9The one check that would have caught it
Everything above came out of a report written afterwards, by people with subpoena powers, full access to the ledgers and eight months to read them. That is not a test anybody can run. The useful question is narrower: what could a person outside the company have seen at the time, from documents anyone could buy?
One ratio. And it is not subtle once you draw it.
¶225 and ¶19, drawn to scale. Both bars are the report’s figures. The gap is arithmetic, not interpretation: money promised that the assets did not produce has to arrive from somewhere, and the only other door is the next saver.
The rest of this section is the same test, broken into the four numbers it needs, and where each one lives in a set of published accounts. Every figure in the answers is BAI’s own.
1How much cash did the assets actually produce?
Where to look: the statement of cash flows — “cash generated from operations” — and hold it against the profit on the income statement. Not the profit. The cash.
BAI: Rs 1.6bn of cash income against Rs 45.8bn raised. 3.3%, against 5–12.75% promised ¶225. That single ratio is the entire case, and it needed no access to anything private.
2Who decided what the assets are worth?
Where to look: the fair-value hierarchy note. IFRS 13 makes companies say it out loud: Level 1 is a quoted market price, Level 2 is derived from observable prices, and Level 3 is the company’s own model. A balance sheet heavy in Level 3 is a balance sheet of opinions.
BAI: some Rs 12bn of fair-value write-ups and Rs 5bn of interest accrued and never received, across the insurer and the fund ¶232. When the report asked for the working, the answer was that the group “could not produce any valuation documentation when requested”.
3How much of the balance sheet is owed by the people who own it?
Where to look: the related-party transactions note, usually near the back and usually short. Add the balances up and divide by total assets. It is a five-minute calculation and almost nobody does it.
BAI: 83%. At that level the question is no longer whether the investments perform — it is whether the group can pay itself.
4Do profit and cash move together, year after year?
Where to look: five consecutive annual reports, one line from each — operating cash flow, and profit. Plot them on the same axis. You are looking for the year they stop tracking each other, not for a threshold.
BAI: profit climbed on write-ups and accrued interest; cash did not follow. Rs 5bn of income was recorded and never received. The divergence is visible from outside, in published documents, for years before 2015.
What this check would not have caught
This matters more than the check does. A reader running all four of the above in 2013 would have concluded that BAI was dangerous, and would still have missed most of this post:
- The round trip. It lasted one day across a year end. Annual accounts describe the position on that date, so a transaction designed around the date is invisible in them by construction. It took the report’s access to the ledgers to see it at all.
- The liability presented as equity. Visible only if you read the terms of the instrument itself — repayable at par, on demand — and then disagreed with how the audited accounts had classified it. That is a judgement against an auditor’s, not a number you can look up.
- The figure missing from the bank’s return. Returns to the Bank of Mauritius are not public. Nothing a reader could buy would have shown the gap between what the bank held internally and what it filed.
- Everything after the delisting. Once the publishing duty ended, the check ran out of inputs. That is the entire point of Part 7.
So the honest version is this. Public documents were enough to see that the returns being promised could not be earned. They were not enough to see how the gap was being covered, and they were never going to be. The check tells you to leave. It does not tell you what happened.
Part 10The people, and what happened to them
The public version of the report blacks out every name except Dawood Rawat’s. The names below come from the Mauritian press — principally Le Mauricien — and from the companies register. Every one of these is an allegation or an interrogation, not a finding. Where an outcome is unknown, I say so.
Right of reply. Nobody named here has been convicted of anything arising from this report, and the report itself makes no finding that an offence was committed. If you are named above and something here is wrong or incomplete, write to me and I will correct it.
What I take from it
The thing that stayed with me is not the fraud allegation. It is how ordinary most of the steps look. Selling a capital-guaranteed product is lawful. A tax exemption is lawful. Taking a company private is lawful. Valuing an asset upward is ordinary accounting, and accruing interest is required by the standards. Almost every individual move in this case is something a legitimate business does on a normal Tuesday.
What made it a machine was the sequence, and one number underneath all of it: the assets produced 3.3% in cash against 5–12.75% promised. Everything else — the write-ups, the round-trip, the delisting — exists to keep that gap out of sight for one more year.
And the gap was visible the whole time to anyone who compared cash received with returns promised. That is the part I intend to remember.