Here are the charts for USD and EUR for the past three years (against a basket of currencies).
USD is still bumping along the bottom where it was three years ago, which is far lower than it has ever been in the last twenty years. Can anybody think of a good reason why it was so much higher in late 2008, early 2009? EUR slid nicely in 2010 (once the whole Greek/Euro-zone bail out tomfoolery kicked off) but it is still higher than it was ten years ago - between early 2000 and mid-2002, it was in the range between 0.8 and 0.9.
To sum up, not much excitement here at all. You can safely ignore what you read in the papers.
Here we go
1 hour ago
31 comments:
I'm sure you are right. But the sceptic in me can't help commenting. (Standard disclaimer: I recognise that I'm an amateur economist, so all the below could be utter bollocks).
Being an engineer rather than a financeer, my thoughts are that your charts and conclusions are highly dependent on the "basket of currencies" chosen. The basket is presumably chosen to operate to normalise the currencies you actually want to measure. In a global recession, we might expect all currencies to fall (on average) in which case charting a falling currency relative to a falling currency would show stability where there is none.
It seems not dissimilar to me to the process of normalising a signal before performing any kind of DSP work on it -- we would usually remove the signal mean and divide by factor that makes a standard deviation of one. Both of those operations are only reasonable if the number of samples is large.
That same limitation must surely apply to currencies too? I'm not sure it does when the comparison currencies are two of the biggest in the world, EUR and USD, how can you possibly find any basket that has enough "mass" to act as a ground for relative measures of these?
I think I would prefer to see currencies measured not against other currencies but against something more demonstrably stable. I would guess at gold or silver. I understand that an amount of gold (I forget what) would have bought you a fine tailored suit in ancient Rome, and would today buy you a fine tailored suit from Saville Row -- implying that gold has remained pretty stable. (That might not be true now the price has shot up because people are using it to escape inflation).
OP: "The basket is presumably chosen to operate to normalise the currencies you actually want to measure. In a global recession, we might expect all currencies to fall (on average) in which case charting a falling currency relative to a falling currency would show stability where there is none."
Exactly correct. But that is traditionally how you measure the 'stability' of a currency - how it's doing against the others, and of course domestic price inflation in each country.
"how can you possibly find any basket that has enough "mass" to act as a ground for relative measures of these?"
I use 8 - USD, CAD, GBP, EUR, CHF, HKD, JPY and AUD, and weight them equally.
Its not dependent really on the currency basket, the USD squeezed a lot higher at the end of 08, start of 09 and it can be explained in two ways.
1) That's the time of the Lehman's bust. If the world is on the edge of a potential collapse with a total breakdown of soceity, then the strength of the currency depends on the military strength. Would you rather have a store of wealth shared by US citizens, or would you rather have your wealth in Thai Baht?
2) Because of the Lehman's bust there was an enormous funding squeeze, every US bank wanted to hoard dollars.
What about plotting the currencies against gold? I'd set Gold as an 'index' or straight line and see what happens.
I use 8 - USD, CAD, GBP, EUR, CHF, HKD, JPY and AUD, and weight them equally.
I have three concerns:
1. You can't chart USD against a basket that includes USD, as that makes it look more stable than it is. At the very least you should use only the other seven when you are making that comparison. Take it to the extreme: if you compared USD against a basket that included only USD then the plot would be a horizontal line through 1.0.
2. If you are comparing USD with EUR by charting each against the basket then you can't include either of them in the basket, since the inclusion of EUR in the USD comparison basket (and vice versa) means that the two charts will correlate even if there is no correlation. So your basket should only include the other six. Again, take this to the extreme to see why. Plot USD as percentage of EUR and EUR as a percentage of USD (ignoring all others), you are not comparing the stability of each, you are comparing their similarity.
3. Weighting them all equally introduces specific bias toward the larger economies. If the British economy has a world influence (let's say) that is a hundred times more than the Swiss. Then by weighting them equally you allow GBP to dominate the basket. Each should be scaled by some appropriate factor (and that's where I give up and let the economists work out the factor).
Let me give you an analogy. Let's say I have research that calculated the average income of ten groups. I can't then calculate the overall average by simply averaging those ten averages, I must weight the group-averages according to the size of each group before summing them and dividing by ten.
I suspect your currency basket is making this same mistake. Your result that the EUR and USD show stability is exactly what we'd expect when comparing them against an average currency over which they exert a strong influence.
I'm not saying your conclusion is wrong, I'm saying that your data doesn't justify it.
BR, yes. I'd also assume it was 'flight to safety' (but strangely, this benefitted AUD and CHF as well, which are relatively small economies, but not Euro-zone, which is large economy.
L, I'm not interested in gold, never have been.
OP: "1. You can't chart USD against a basket that includes USD"
I didn't - see explanation here. Each currency is plotted against the other seven.
"3. Weighting them all equally introduces specific bias toward the larger economies."
That's being too clever for your own good. For each currency, it is assumed that that economy is one-seventh of the size of all the others put together. Sure, the USA may actually be a quarter, and Switzerland may only be one per cent, but it all comes out in the wash. I've tried weighting them and the picture is much the same.
Further, from the point of view of somebody speculating between USD and CHF, all you care about is relative movements, the fact that the USA is twenty-five times bigger than Switzerland is neither here nor.
AUD just got sold off a lot less because of its commodity exposure and the current (perhaps incorrect belief) in a degree of Chinese decoupling.
CHF because its pretty much backed by gold which benefitted at the time, but also because, well you'd feel based on history that a Swiss bank account is safe.
Whether you do it trade weighted, equal weighted or against a customized index, the story will be the same, because The Dollar is fundamental to the world economy.
MW No. I'm not 'interested' in Gold, but it's an intriguing graph to plot. I'll see if I can find it.
Lola, its as interesting as the reciprocal of the gold price chart.
That's being too clever for your own good.
That's pretty much a a given ;-)
For each currency, it is assumed that that economy is one-seventh of the size of all the others put together. Sure, the USA may actually be a quarter, and Switzerland may only be one per cent, but it all comes out in the wash.
I recognise that "it is assumed"; that's kind of my problem. That assumption is incorrect.
And I'm not sure what "it all comes out in the wash" means? Surely that argument is only valid if there is a reasonable justification that a change here is compensated by a change there?
I've tried weighting them and the picture is much the same.
I'm afraid that that implies to me that there is something wrong. Changing the weightings should change the result. If it doesn't then you have picked a basket of strongly-correlated currencies. i.e. they are all following some underlying super-currency. In which case, moving money from one to the other wouldn't provide you any security at all. They could all be collapsing together and we'd never know from these charts (which is what is actually happening I think).
Further, from the point of view of somebody speculating between USD and CHF, all you care about is relative movements, the fact that the USA is twenty-five times bigger than Switzerland is neither here nor.
You say that it doesn't matter that the USA is twenty five times bigger than Switzerland, but my argument is that by comparing as you have, CHF's contribution is huge (twenty-five times bigger than it should be), and therefore we can't make any judgement about relative stability -- the result is skewed toward the larger currencies.
My feeling is that your methodology will always come out with the result that the largest currencies in your basket are the most stable, because you are broadly measuring them against themselves. It's no surprise that the charts come out around 1.0; they were bound to by the choice of method.
I would be more interested to see these plotted in units of "loaves of bread", or against a localised inflation index.
So first off, gold is irrelevant other than as a measure of consumer confidence - it shows how many people have unimaginative, failed paranoia (note: if someone at a bank holds your gold, that's *exactly the same* as if they hold your currency. Unless you're actually in a shack lying on a bed of Krugerrands and brandishing a nine-barrelled elephant gun, you're no safer than me).
On everything else, Mark and Onus are both equally right/wrong. The size of the economy is completely irrelevant. The amount of cross-border exposure, however, is completely relevant. Suppose that instead of the USD, all international business, loans, etc, were denominated in CHF - at that point, the fact that the Swiss economy is about the size of New Zealand plus the Pacific Islands would be irrelevant - the CHF would still be bloody important.
Which, from the context of international trade, means that although Mark's basket isn't rigorous, it's still OK. China's external trade is conducted in JPY and USD, or currencies linked to them, so that's why USD rates are relevant.
Hell, all international trade (or over 95%, which works for modelling purposes) is done in USD, EUR, GBP or JPY - at which point, every other currency is a modelling irrelevance. And you could lose the pounds and the yen without losing much from the model overall...
My punt in NZD, much mocked when I mentioned it on the web two or three years ago, has prospered. Nya, nya, nya.
Onus Probandy,
" i.e. they are all following some underlying super-currency."
you mean like the dollar??
"In which case, moving money from one to the other wouldn't provide you any security at all. They could all be collapsing together and we'd never know from these charts (which is what is actually happening I think)."
a false conclusion. If you were to analyse thousands of stock returns, you'd find a strong principle component the "market" return. That however doesn't remove correlations within sectors and their secondary importance.
The point as I said before, the global price of the dollar is the driving factor, and your choice of basket won't change the story too much, especially if you're looking at a chart.
OP: "If [changing the weightings doesn't change the result] then you have picked a basket of strongly-correlated currencies... therefore we can't make any judgement about relative stability -- the result is skewed toward the larger currencies."
Well nope. The total value of all the currencies in the basket on any one day is always exactly 8, by definition. That's by intention, not by accident.
As to weighting, you have actually contradicted yourself. Let's take two reasonably 'similar' currencies, e.g. USD and CAD, one large economy, one small, the two otherwise move reasonably well in line.
If we had only two currencies in the basket (USD and CAD) and weighted them 1-to-1, they would both appear to be equally 'stable' (or equally 'unstable').
Were I to weight USD with 9 and CAD with 1 (to reflect relative size of GDP) and rework the figures, the relative stability or instability would be exactly the same because I only measure one currency against all others.
If I then bung in AUD (which is not correlated with USD), weighted 0.5, this makes USD appear less stable than just against CAD, because we are measuring it against 1 CAD and 0.5 AUD.
But if I weight USD against 1 CAD and 1 AUD, in relative terms, USD appears even less stable than measuring it against 1 CAD and 0.5 AUD. If anything, equal weightings understate the relative stability of the currencies of the larger economies.
JB, ta for back up. You forgot to mention the tin foil hat. As to all international trade being denominated in CHF, why would that make any difference? We'd still have an exchange rate for each currency to CHF and hence relative exchange rates for all currencies.
D, when did I mock it? AUD would have been even better, but such is life.
BR: "your choice of basket won't change the story too much, especially if you're looking at a chart."
Exactly, however scientific the maths, there is no evidence to say that currencies are worth what people think they are worth (see also 'purchasing power parity'). Charts are just a rough and ready visual tool which give you clues as to what might happen next.
If there is a 'super currency', I suppose the unit of value is not 'gold' or 'oil' but 'how credible are the tax and spending plans of each country's government' or 'how well is their economy doing, do they have a trade surplus or a trade deficit' etc.
@BR:
You mean like the dollar?
Yep, that would fit my analysis. It can't quite be the dollar as the dollar and this "super-currency" are plainly different. However, it could quite easily be "mostly the dollar", with the others thrown in.
a false conclusion. If you were to analyse thousands of stock returns, you'd find a strong principle component the "market" return. That however doesn't remove correlations within sectors and their secondary importance.
I don't really understand how we've disagreed here. My "all collapsing together" could be the "principal component" or "market return" that you speak of. Whether there are secondary correlations would be irrelevant for currencies in the light of that overall trend. The fact that people can still make money on a falling market is obviously a demonstration of that fact -- those secondary correlations can be important then. It doesn't change the fact that the whole market is falling though. A sinusoid imposed on a decaying line has ups and downs, but in the end it will hit zero. Countries care about the line; investors care about the sinusoid.
The point as I said before, the global price of the dollar is the driving factor, and your choice of basket won't change the story too much
I'm fine with this point; it's simply adding a practical reality to my theoretical conclusion that Mark's methodology will always show stability for the dollar (and by extension, the other large currency the euro) because he's measuring them against themselves (effectively).
OP: "Mark's methodology will always show stability for the dollar (and by extension, the other large currency the euro) because he's measuring them against themselves (effectively)."
But I keep explaining, that's exactly what I'm not doing.
EUR is measured against USD+CAD+GBP+CHF+SGD+JPY+AUD.
USD is measured against CAD+GBP+EUR+CHF+SGD+JPY+AUD
and so on.
OP,
The methodology for the dollar chart is irrelevant if you agree there is a principle component, its the super currency and whatever the basket you're comparing it to the USD side of the equation dominates.
The reason that the weighting doesn't matter for the euro chart is because by taking an average across a basket he's effectively stripping out the dollar exposure and then looking at a rough residual of the currency.
ie. lets say we have twenty stocks, we agree that they principle component of their returns will me the market return so we could write;
stock return = market return * beta + residual return.
by weighting one stock against a simple average of the other stocks we are forming a proxy for "market" as a simple average, we could market cap weight, price average, volume weight, or whatever, but the general picture will be the same. so the chart of the EUR is effectively a chart of the residual performance. Which is the part that contains the country (well currency) specific information, rather than the price of the dollar.
BR, I think that's a fair summary - a good measure of the relative performance of one stock is to measure its relative performance against all other stocks (it would be daft to measure it against currency movements, the number of gold medals a country wins in the Olympics or the height in inches of the Prime Minister).
OP, BR, if you want, send me an email and I'll email you the whole spreadsheet to tinker with at your leisure. You could reconstruct the raw data from oanda.com, but that takes hours.
MW, thats kind, but I have my own price feed already.
I think the more interesting question is what "should" happen to the EUR in the case of further bailouts or on the other hand a default, or Germany choosing to split off and re-introduce the DMark. The question then is what is your EUR actually worth?
But I keep explaining, that's exactly what I'm not doing.
You addressed point 1 of my "concerns" list (by confirming that you don't use the currency being compared in the basket), but not 2 (the fact that you include USD in the EUR basket and USD in the EUR basket, creating an implicit correlation between your two charts) and not really 3, other than to say "it all comes out in the wash".
I accept that on point 3 I'm talking about things I don't know enough about to make my own conclusions -- I am unable to make any estimate as to weightings. But I think I know enough to see that treating CHF as if it is equal to EUR in your currency basket is like including one car and one loaf of bread in the inflation estimate basket. Exactly how much would the loaf of bread have to change by before it has any impact on the inflation calculation?
If we had only two currencies in the basket (USD and CAD) and weighted them 1-to-1, they would both appear to be equally 'stable' (or equally 'unstable').
Stable relative to what? If there were only these two, then there is nothing you can weight, as there is only ever one other currency for comparison. Therefore your weightings of nine and one here are meaningless.
If I then bung in AUD (which is not correlated with USD), weighted 0.5, this makes USD appear less stable than just against CAD, because we are measuring it against 1 CAD and 0.5 AUD.
What does it matter what "appears"? We're looking for what's true. What we want is for the one that appears most unstable is the most unstable. If USD drops against AUD, then the chart should show that instability. Your example of increasing AUD from 0.5 to 1 is a simple expression of the fact that including more of an independent variable into a weighted sum naturally makes the result more independent. Isn't that obvious? The question is how best to measure that independence? We can't just make up a weighting.
The problem I was raising in my point 3 (perhaps unclearly) is that currencies are not independent. Hence, when you plot A against B, you're actually plotting A against (X * a + W * b), with "a" and "b" being the independent parts of A and B. (I'm actually still simplifying because there are almost certainly time lags in the system).
So, we have the observable signals of the currencies, which I'll label as the vector X; and the underlying (and unknown) independent currency drivers, which I'll label as the vector Y.
These are related like this:
X = M Y
Where M is some unknown mixing matrix.
So, we have two unknowns, and one series of observations. We can guess that M will be such that that the dominant currencies in X will be represented by a large component of driving currency in Y, but not much else.
Finding M is called "independent component analysis"; which is an extended version of "principal component analysis"; engineers would call it blind source separation. It would actually be terribly interesting to apply this technique to currencies.
I've gone off track a bit... unless this M matrix is a simple diagonal only case, then this plotting of currencies against baskets of others is a vast oversimplification and tells you nothing, since any conclusion you come to about "stability" is not valid if the underlying mix is such that every other currency is a proxy for some primary one (which BR says is the dollar).
BR, well, yes. And possibly, exactly.
OP your thinking is a little muddled, AUD isn't an object on its own, its price only exists wrt to other currencies.
MW I believe when talking about stability is not talking about some sort of cointegration, but by referencing the newspapers, is referring not to the volatility, but to the performance of the euro against other currencies.
OP: "I think I know enough to see that treating CHF as if it is equal to EUR in your currency basket is like including one car and one loaf of bread in the inflation estimate basket. "
That is a very good analogy, except by definition I take am taking, say one loaf of bread and one-ten-thousandth of a car; as well as one-twentieth of a pair of shoes; one-and-a-half newspapers etc.
When looking at the relative price of 'a car', I then compare the price (in pence) of one-ten-thousandth of a car with the total price (in pence) of [one loaf of bread + one-twentieth of a pair of shoes + one-and-a-half newspapers etc.]
When looking at the relative price of 'a pair of shoes' I compare the price (in pence) of one-twentieth of a pair of shoes with [one-ten-thousandth of a car + one loaf of bread + one-and-a-half newspapers etc.]
That fact that people spend £1,000s a year on cars (depreciation) and only £10s or £100s on shoes doesn't matter - if the calculations say that shoes have "gone up by ten per cent" and cars have "gone down by ten per cent" we know what we need to know about relative changes in the cost of shoes, cars etc.
OP, just to round that off, in that 'real goods' basket I also include one English pound, costing 100 pence.
So when valuing "one English pound" I compare it with the price of [one-thousandth of a car + one loaf of bread etc]. if the prices (in pence) of all these goods goes up by an average of five per cent, then the 'inflation' chart for "one English pound" would show it is now only worth +/- 0.95 of what it was worth before.
MW et al. In other words currencies are just another commodity?
Lola, when you can pay your taxes in orange juice and pork bellies then it will be just another commodity.
BR, well, in theory you can and do. If you don't pay in cash they'll come along and seize your assets. If all you have is pork bellies and orange juice, both of which will have a 'value' to someone, they'll take those.
Furthermore I seem to remember that 'Rupees' is a Hindu word for Oxen. In other words money worth was ex[ressed in terms of a commodity (oxen) of which everyone had an idea of value of.
L: "In other words currencies are just another commodity?"
Yes of course. And agreed to your comment after that.
@MW
That last post about currencies effectively being fractions-of-a-commodity has cleared it up for me.
Thanks.
Currencies are (I see now) effectively already implicitly weighted as I wished for in my original post. Therefore including them with equal weighting is fine.
Post a Comment