Mania, Panic and Capitulation: The Phases Named

Mania, panic and capitulation name three phases of a price cycle: rising conviction, a sharp reversal, and the exhausted selling that ends the fall. All three names are assigned afterwards, from a path that has already finished. On the invented Palash 100 the capitulation quarter fell 7.1 per cent while the panic quarter ahead of it fell 14.5, so the naming does not follow the arithmetic.

The three words rest on one distinction, and almost every argument about them is really an argument about that distinction. A description says what a completed path looked like. A rule says what to do while a path is still running. The dividing lines are chosen after the shape is known, so any finished price path can be divided into phases, and that is exactly why the divisions always seem to fit so well. Whether the same lines could have been drawn from the left-hand portion alone, with the right-hand portion still blank, is a completely separate question, and the answer is no.

What do mania, panic and capitulation actually name?

The shape is older than markets, so start outside them. Word goes round a neighbourhood on a Tuesday morning that the vegetable lorries have been held up and there will be no onions for a week. By eleven the queue at the stall is forty deep and people who normally buy half a kilo are buying five. The price triples and the buying speeds up rather than slowing down. The rising price is being read as proof that the rumour was right. At half past two a lorry turns into the lane. Within ten minutes nobody will pay the morning price, and the people who bought five kilos are trying to pass them on to neighbours at any price at all. By six in the evening the last of them has given up and sold at a loss, and the stall is quiet.

The afternoon has three parts, and everybody in the lane can name them by dinner time. ManiaA phase of rising prices and rising conviction, named after it has ended. is the first part: the price rises, and the rise itself is treated as evidence that the rise is justified. PanicA sharp reversal in which selling accelerates. is the second: everybody wants to be out ahead of everybody else, so the direction reverses and selling speeds up rather than slowing. CapitulationThe exhausted selling that ends a fall, named once the low is known. is the third: the last holders, the ones who insisted longest that the price would come back, finally give up and sell at whatever is on offer, and the selling stops because there is nobody left who still wants to sell.

Notice the tense in every one of those three definitions. The mania is the part that ended. The panic is the reversal that turned out to be a reversal. The capitulation is the selling that turned out to be the last of it. Not one of the three can be stated without a verb that looks backwards. The names come from a literature that was explicit about this. Hyman Minsky, in The Financial Instability Hypothesis in 1977, set out a sequence in which a long calm period changes what borrowers and lenders think is prudent, and Charles Kindleberger, in Manias, Panics and Crashes in 1978, gathered several centuries of episodes and showed the same stages recurring in each. Both were writing history. Neither was writing a timetable.

Who assigns the names, then? Whoever is telling the story afterwards. There is no committee, no register and no threshold. A fall of a certain size is not automatically a panic and no authority declares one. The absence of a threshold is not a gap in the definitions to be filled in later, but a property of what the words are for. The three names are labels for parts of a story, and a story needs an ending before its parts can be marked out.

Set each name against the thing it needs before it can be closed off. Every one of the three closes on an edge that only a later reading fixes, and the last of them cannot be completed until Q5, three quarters after the peak it measures from.

What did the Palash 100 do, quarter by quarter?

The Palash 100, an invented index, opens at 100.0 and is then struck at eight quarter ends: 118.0, 131.0, 112.0, 104.0, 116.0, 124.0, 121.0 and 127.0. The nine readings are the entire record. There is nothing between the quarter ends, nothing before the open and nothing after the eighth quarter.

The picture does the persuading and the arithmetic does the checking, so work every change out rather than reading the shape off the picture. The first quarter adds 18.0 points on a base of 100.0, so it is up 18.0 per cent. The second adds 13.0 on 118.0, up 11.0 per cent, and notice that the index rose by more points in the first quarter than the second. The third loses 19.0 on 131.0, down 14.5 per cent. The fourth loses 8.0 on 112.0, down 7.1 per cent. Then 12.0 on 104.0 is up 11.5, 8.0 on 116.0 is up 6.9, 3.0 on 124.0 is down 2.4, and 6.0 on 121.0 is up 5.0 per cent.

Points and per cent do not rank the same quarters in the same order, so one more thing falls out of that table. The largest move in points is the 19.0 point fall at Q3, the largest in per cent is the 18.0 per cent rise at Q1, and the unit alone decides which quarter gets called the biggest.

Where do the peak and the low actually fall?

The highest reading is 131.0 at Q2 and the lowest is 104.0 at Q4. Only two readings on this record can be called a peakThe highest point on a completed path. and a troughThe lowest point on a completed path., and both words are being used with the whole record already in hand. Label the moves by the quarters and not by counting the open as a quarter: from the open to Q2 the index is up 31.0 points on 100.0, so up 31.0 per cent, and from Q2 to Q4 it is down 27.0 points on 131.0.

The fall of 27.0 on 131.0 is 20.6 per cent, and it is the number a report would print as the drawdownThe fall from a peak to a later low, as a share of the peak.. Now hold that number beside a second one. The low of 104.0 sits above the open of 100.0 by 4.0 points on 100.0, or 4.0 per cent up. The same Q4 reading is a 20.6 per cent fall and a 4.0 per cent gain at the same time, and which of the two gets printed depends entirely on where the measuring started. A drawdown is measured from the peak by construction, so a large one says nothing about whether anybody is behind on what they paid. Somebody who bought at the open and did nothing was ahead by 4.0 per cent at the worst reading of the whole record.

The open and the peak are not the only places a measurement can start, so push the comparison further. The same Q4 reading can be set against any earlier reading on the record, and every base gives a different and entirely honest answer.

Which quarter was the worst, and is it the one the names imply?

A story about this path writes itself. The mania ran to Q2, the panic hit in Q3, and Q4 was the capitulation, the quarter when the last holders gave up. Every word of that sentence is available from the record. The sentence is still wrong about the one thing it most wants the reader to feel, that the capitulation was the worst part.

Q2 to Q3 falls 14.5 per cent. Q3 to Q4 falls 7.1 per cent. Divide the second by the first and the later quarter is 49.0 per cent of the earlier one, so the quarter a narrative calls the exhausted end of the fall is under half the size of the quarter before it. The name and the arithmetic point in different directions, and when they do it is always the name that was fitted to the story rather than the arithmetic that was fitted to the name. The mismatch is not a defect in this invented path but a consequence of what capitulation means. Capitulation is the end of a fall, and the end of a fall is by definition the part where the falling is running out.

Split the fall by where the points actually went and the gap widens. Of the 27.0 points from 131.0 down to 104.0, 19.0 went in Q2 to Q3 and 8.0 in Q3 to Q4. Seven tenths of the fall, 70.4 per cent of it, was over before the quarter the second name marks had begun.

What did turnover do at the top and at the bottom?

Price is only half of what a market leaves behind. The other half is turnoverHow much changed hands, here as a multiple of the eight quarter median., and this case records it for exactly two quarters. In Q2, the quarter of the highest reading, turnover ran at 3.1 times its eight quarter median. In Q4, the quarter of the lowest reading, it ran at 0.4 times. Divide one by the other and 3.1 over 0.4 is 7.8, so about eight times as much changed hands at the top as at the bottom.

The temptation is to read the pair as more than it is. Read it plainly instead. Activity was heaviest in the quarter where the path was about to turn down, and lightest in the quarter where it had already turned up. The crowd was at its busiest where the record shows the highest reading and at its quietest where the record shows the lowest. The pairing is a fact about what happened and not a method for finding either point. Think of the vegetable stall again: the queue was longest at the moment the price was highest, and there was nobody at the stall at six in the evening. Nobody in the queue knew at eleven that eleven was the busy hour, and nobody at the empty stall at six knew that six was the quiet one. Only the person writing it up that night knew both.

One caution attaches to those two numbers, and it is how little of the record they are. A multiple is given for Q2 and for Q4 and for no other quarter. Two readings cannot make a shape, and the two on offer are the extreme quarters, the pair that would flatter any story built on them.

What would somebody holding from the open have seen at each point?

Phase names describe the index. A phase name does not describe a person, and the two can differ sharply. Take somebody who bought at the open, did nothing at all for eight quarters, and looked at the statement each quarter end. Their reading is the level minus 100.0, so the sequence they saw is up 18.0, up 31.0, up 12.0, up 4.0, up 16.0, up 24.0, up 21.0 and up 27.0 per cent. Every single quarter end of the supposed mania, panic and capitulation showed that holder a gain, and the worst of the eight was still 4.0 per cent up. The panic was real and the capitulation was real, and neither of them ever put this person behind.

Being ahead the whole way is not the same as feeling nothing. At Q2 that holder was shown a gain of 31.0 per cent, and at Q4 the same holder was shown 4.0. Nobody went below what they paid, and 27.0 of the 31.0 points of shown gain went anyway, or 87.1 per cent of it.

Now the arithmetic that a phase story almost never states. Getting back to a peak costs more, in percentage terms, than the fall that lost it, and the gap widens with the size of the fall. Falling 27.0 points from 131.0 is a fall of 20.6 per cent. The base is smaller on the way back, so climbing those same 27.0 points from 104.0 is a rise of 26.0 per cent. The four quarters after the low delivered a rise of 23.0 points on 104.0, or 22.1 per cent, taking the index to 127.0. A rise of that size is a strong recovery and it is still 4.0 points short of 131.0, or 3.1 per cent of 131.0. Four rising quarters after the low, and the index has still not got back to where it stood at the reading a story would call the top of the mania.

Count the same record in quarters rather than in per cent and a second asymmetry appears. The rise took two quarters, the fall took two, and the climb back has taken four without finishing.

Why can a phase boundary only be drawn from a completed path?

Here is the mechanism, and it is simpler than it sounds. Every phase name is a division, and a division needs two edges. Mania needs a beginning and an end, and its end is the peak. Panic needs a beginning and an end, and its end is the point where the fall stopped. Only a stopped fall makes the last selling the last selling rather than the middle of some longer selling, so capitulation needs the fall to have stopped.

Now ask what fixes each of those edges. The peak is fixed by the fall that came after it. The end of the fall is fixed by the rise that came after that. Every edge that a phase name depends on is fixed by something that happens later, so no phase name can be completed with the right-hand end of the chart still open. The test for whether a phase name is available therefore has nothing to do with how dramatic the price action is, and everything to do with a boring structural question: what is the last plotted point, and is it the thing being named?

Run that test on every right hand end this record allows. The low gates every name, so the peak is markable from six of the eight and all three names from only four.

What happens to the names as the chart gets longer?

The question can be answered by hand, one quarter at a time. The control below plots the Palash 100 up to whichever quarter is selected and nothing beyond it. The peak marker, the low marker and the phase names do not appear together.

Why is a named phase not a trading signal?

The boundary between explanation and instruction matters more than any of the arithmetic above. A behavioural account of why a price moved is an explanation of a completed path, and it is never an instruction to buy, sell, hold, wait or avoid, and never evidence that acting on it would have paid. Thin turnover does not mark a bottom, a fall of about 20.6 per cent is not the usual size of one, and no index anywhere is bound to do what this invented one did. The sentence a reader must never assemble is: the effect exists, therefore trade it.

Three separate reasons keep that boundary honest, and any one of them is sufficient on its own. The first is cost. An effect measured after the fact is measured before the cost of acting on it, and this case has the cost in front of it: across the five turnover groups in the Palash decision log, gross returns sit inside 0.3 percentage points of each other at 11.2, 11.0, 11.1, 10.9 and 11.0 per cent, and net returns run 4.0 points apart at 10.9, 10.4, 9.6, 8.4 and 6.9. The difference between the two ends is what the trading cost, not what was picked, and 4.0 points a year is larger than most documented effects of this kind.

The two spreads are the whole of the first reason.

The second is publication. A pattern written up in a paper is read by everybody who reads that paper, so what it did before it was published is not evidence about what it does afterwards. The third is the deepest. The same obstacles that let a mispricing survive in the first place are the obstacles that stop a reader capturing it. Andrei Shleifer and Robert Vishny set that out in The Limits of Arbitrage in the Journal of Finance in 1997. The explanation and the obstacle are one fact seen from two sides. If it were easy to act on, it would not have lasted long enough to be described.

What does the decision log record about any of this?

Nothing, and that is worth saying plainly because it is the kind of gap a reader fills in without noticing. The Palash decision log is an invented record of 240 decisions taken by 60 investors over eight quarters: 96 buys, 84 sells, 36 switches and 24 pauses of a standing instruction, summing to 240. The log has no quarter by quarter breakdown. The Palash 100 is a separate invented record of nine index readings. There is no shared column between the two, so not one of those 240 decisions can be placed anywhere on the path drawn above.

The same discipline applies to the individual case. Meera Sundaram, the invented investor in this record, holds four positions that cost Rs 13,00,000/- in total and were valued at Rs 12,46,000/- when the eight quarter valuation was struck. The valuation belongs to her holding. The valuation is not a reading on the Palash 100, it cannot be converted into one, and the fact that both records happen to cover eight quarters joins them not at all.

How does somebody use phase language without turning it into a call?

Devika Rao, the adviser at the invented Palash Advisory Services Private Limited, meets these three words in almost every conversation, usually in the form of a question that has already answered itself. The useful move is not to ban the vocabulary but to make the tense explicit. When somebody says the market is in a panic, the reply that does real work is to ask which later reading fixed the peak they are measuring from. If the answer is that there is not one yet, the sentence is a mood and not a description.

For a person deciding alone, with no adviser and no committee, the same handle works and is easier to apply than it sounds. Write down, before reading any commentary, what the last reading on the chart is and what would have to happen after it for the phase name to be completable. If completing the name requires a reading that has not arrived, then the decision about to be taken rests on a story rather than on the record, and knowing that changes nothing about the decision except that the basis of it is now clear. A clear basis is a small thing to gain, and it is the only thing the check honestly offers.

The handle can be written down, and it is short enough to fit on the back of a statement.

The professional version carries one extra duty. An adviser describing a phase to a client is making a communication about a market, and how such communications may be framed is a matter of conduct requirements rather than of craft. Anybody writing them names the applicable requirement and confirms it at source.

Meera Sundaram, Devika Rao, Palash Advisory Services Private Limited, the Palash decision log and the Palash 100 index are invented.
Educational material. Not advice on any investment, tax, budget or market position.