So - considering that LIBOR was not an observed number, but a made-up number that was nonetheless useful, what did the LIBOR actually measure?

Bankers' consensus of what the overnight interest rate should be for a well-capitalized, creditworthy bank.

It measures the answer to the hypothetical question "If a another bank with good credit came to you right now to borrow (overnight/1W/1M/3M) in (USD/EUR/GBP/CHF/JPY) what rate would you offer them?" Which is a proxy for the banks' willingless to lend. (It had a less quoted counterpart, LIBID the London Interbank Bid Rate, where would you borrow at.)

No, LIBOR is your borrowing cost, the rate at which you'd be able to borrow (ie the rate that others would offer to you to lend you money).

LIBID is the rate at which you'd be willing to borrow. So the difference between the two is like bid vs ask.

ELI5 Why not simply measure the actual trading activity for exactly those activities the previous day?

Good question. I suspect this was because when the rate was defined in the 1980s, collecting that data would have been difficult.

After the LIBOR scandal, the EU brought the benchmarks regulation (BMR) which says that interest rate indexes have to be based on actual transactions, just as you say. Euribor, the equivalent of LIBOR for lending in euros, was reformed to be based on transactions:

https://www.emmi-benchmarks.eu/benchmarks/euribor/reforms/

The administrator of LIBOR proposed doing the same:

https://www.clarusft.com/rfrs-libor-is-changing/

But in the end, US and UK regulators decided just to abolish it, in favour of overnight indexes based on real transactions (SONIA for pounds, which already existed, and SOFR for dollars, which was created for this purpose).

I believe this divergence happened because of differences in the lending markets. In the euro area, there is still a lot of unsecured term lending, which is what Euribor measures. But in the UK and US, this kind of lending has largely dried up, but there is a lot of overnight lending, so they chose rates which measure that. I don't know why the euro area is different to the US and UK here. It's possible that the euro market will evolve to be more like the US and UK, in which case Euribor will stop being credible, and the euro will also move over to its overnight rate, ESTR.

Another fun quirk is that SONIA and ESTR measure unsecured overnight lending, whereas SOFR measures "repo", which is essentially lending secured with government bonds as collateral. There is a sterling overnight repo rate, RONIA, but i don't think it's used much. I think repo volumes are higher than unsecured lending volumes; if that difference gets stark enough, perhaps sterling and euro regulators will force another switch, to the repo indexes.

That's a day old. And the market may not be very active. Besides, I don't think they are centrally cleared? (Or perhaps in the early days they weren't, it was certainly pre-digitization)

The market - interbank rates - is quite small in terms of participants. I think the problem is the number was incredibly useful even if it wasn't accurate.

1) easier said than done - in general these are often OTC markets without clearing houses - that being said not insurmountable difficulties but not zero-cost either. 2) the previous days trades are backward looking, not forward looking. What’s the big deal you say, it’s just a day? We’ll think about driving by using your rear view mirror — you do fine until to hit a turn then things get a little sketchy, and a sharp turn can kill you.

The reason it's not 'simply' is that you could manipulate that pretty easily too.

It measured the consensus for what LIBOR should be, a rate that the banks as a whole were happy to use. Your question is a bit like asking what dollars are actually worth how that we've left the gold standard.

LIBOR is often part of the formula for calculating loan interest rates. Last time I saw such code it looked something like:

LIBOR + bank's minimum interest rate + a rate based on creditworthiness = your offered interest rate

Planet Money recently had a really good episode on how some banks are deciding on a replacement rate: https://www.npr.org/2021/10/08/1044598674/libor-pains

That's not what it's measuring, that's how it's used.

That's also what it measures, as measured by the people doing the estimation of creditworthiness, who are the experts in understand how to measure that value.

If anyone else was more accurate at this measurement, then they had arbitrage against those using the measure, giving those doing the initial measurement incentive to get it as right as is humanly possible, since they usually worked at places that use LIBOR to price things.

Since LIBOR underlied hundreds of trillions in assets, there are ample papers on all aspects of LIBOR, including those trying to see how well it was computed versus post outcomes.

It holds up well. https://scholar.google.com/scholar?hl=en&as_sdt=0%2C14&q=LIB...

It measures the backroom sensitivities of the cabal of bankers that set it, judging by recent news releases