You will be able to name the three most common ways a money answer goes wrong before you read it.
Open the most recent money answer an assistant gave you, whether it was about a fee, a savings rate or a tax relief. Mark each number, date, rule and source you can find in it. Each mark is an item to check before you do anything based on that answer.
Money answers usually fail in one of four ways: invented numbers, old facts, a wrong country, or fabricated sources. Once you know the four, you go into an answer expecting them, and they become much easier to catch. Of the four, a fabricated source is the one you'll meet most often in money answers. Whichever failure is involved, every number, date, rule and source needs a check before you act on what the assistant said.
The first type, invented numbers, covers figures with a specific look and no real basis. They tend to appear as a fund's expense ratio quoted to two decimal places, the interest a savings account pays, or what an investment has returned on average over ten years. A fee of 0.45% (an invented example) reads exactly like a real fee would. Precision makes a figure feel trustworthy, but it gives you no information about whether the figure is right.
The cause goes back to Lesson 1.1, "What an assistant is actually doing when you ask about money": the model writes by predicting what usually comes next. When a number is due, it produces one of the right size, in the right format and inside the usual range. A number can meet all three of those and still be false.
Expect this whenever you ask about a fee, rate, return, premium or other figure for one particular product. A narrow product has had less written about it, so the model has seen less text on it and fills more of the gap with a figure that fits.
Old facts are the second type: figures that were correct at one point and have changed since. A model learns from text up to its training cutoff date, and Singapore's money rules keep moving after that date. CPF interest rates, contribution caps, Singapore Savings Bonds rates and housing grant amounts are all revised on schedules set by the agencies responsible for them. Income tax reliefs change a different way: the Minister for Finance announces changes in the annual Budget and Parliament makes them law, while IRAS administers the reliefs and publishes the current amounts but does not set them. An assistant can repeat a rule word for word from its training text and still give you a figure that is wrong today.
Old facts are harder to spot than invented numbers because searching for the figure can turn up the same number in a blog post written three years ago, which feels like confirmation even though it is the same outdated figure in a second place. Some assistants can search the web themselves. That helps, but a search can land on an old blog post as easily as on the official page, so the problem remains.
For any rate, cap, limit or threshold, treat it as something to look up, and look it up on the site of whichever body sets it. Module 6 goes into old facts and the limits of web search in more detail.
The third type, wrong-country answers, apply another country's rules to you, and the other country is usually the US. Far more is written online about American personal finance than about Singapore's, and answers lean the same way. A question about saving for retirement can come back with 401(k) plans and Roth IRAs. A question about your credit score can come back with advice that depends on how US credit scores work.
Start by saying where you live at the top of the question. "I live in Singapore" is enough. That shifts a great deal of the answer, though a foreign rule the model has seen many times can still slip in, so keep watching for one as you read. Lesson 6.2, "Answers that quietly assume you live in the US", has a checklist for catching them.
Ask an assistant where a figure came from and you may get a report title, an author, a publisher, a year and a link. Some of these are genuine. Others are fabricated sources, assembled from parts that sound real: an organisation that exists, a believable title, a date that fits, and a web address in the correct format that goes nowhere. A citation has a familiar shape, the same way a fee or a rate does, and producing familiar shapes is what the model does well.
A fabricated source also makes the other three types more dangerous, because a wrong figure or an old fact with a citation beside it looks as if someone has already checked it.
Darren (an invented example) wasted a Sunday in an earlier lesson. This time he asks something narrower: what share of young Singaporeans have an emergency fund, since he wants to know whether he is behind. The reply gives a percentage, names a survey by a local bank with a year beside it, and includes a link, all written like a paragraph from a newspaper. When he clicks the link, the page does not exist. He searches for the survey's title and nothing by that name turns up. Darren has run into a fabricated source, and he did nothing wrong. Getting one does not mean you asked the question badly.
Asking the assistant "are you sure?" seldom gets you further. It may apologise and offer a different source that also doesn't exist, or stand by the first one. Opening the source is the only real test, and until you have opened it, you don't yet have a source. To check one:
click the link search for the exact title inside quotation marks for a government source, go to the agency's own site and find it there
Go through an answer one sentence at a time and place each sentence in one of two piles. Pile one holds explanations of how something works, and you can judge most of these by whether they make sense. Pile two holds specific figures, dates, rules and sources, and each item there stays unconfirmed until you have checked it somewhere other than the assistant.
In Darren's answer, the paragraph about why an emergency fund matters made sense and belongs in pile one. The percentage, the survey and the link belong in pile two, and none of the three held up when he checked.
Try the same thing with a statistic of your own. Ask for a money figure, then ask where it comes from, and see what happens when you go looking for each source it gives you.
Ask an assistant where a money statistic comes from. Open every link or title it gives you, and write down which ones exist and whether they say what it claimed.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).