# FOREX Relative Buying Power — Part 3

Greetings!

I’m back with the third installment of this series about FOREX relative buying power. In the second part, I discussed my basic dataset and how I constructed a buying power chart for eight currencies. Warning! Red Alert! This chart is wrong!

Yep. I’ve rarely traded the USD/CAD, and for some reason I thought the Canadian Dollar was the base currency (CAD/USD). Oops! So the CAD data was all backwards, upside-down and sideways. The corrected chart is below. Also, here’s the corrected research spreadsheet.

### Moving Right Along…

Ok, so now that I’ve recovered from that setback and had a second cup of coffee with yummy girl scout cookie flavored creamer, it’s time to dig more deeply into the data and see if we can find anything useful. First I looked at the changes in each currency’s buying power from one week to the next. This procedure is similar to the one described in the previous post, but instead of always using the first week in the data for the comparison, we compare each week to the previous week.

This will yield small percentage changes in buying power for each of
the eight currencies in the data. For example, the first week’s changes are -2.02% for the USD, -0.32% for the EUR and so on, as seen on the “Weekly History” tab of the research spreadsheet, cells AD3, AE3, etc. To normalize the data onto a standard scale, I divide each positive change by the largest positive change, and the absolute value of each negative change by the largest negative change among the eight currencies. This puts the positive changes on a scale of 0% to 100% and the negative changes on a scale of 0% to -100%.

For example, if I only had three currencies and the buying power changes were -2% for USD, 1% for EUR and 4% for JPY, the normalized values would be:

USD: ABS(-2)/-2 = -100%
EUR: 1/4 = 25%
JPY: 4/4 = 100%

These calculations are on the “Weekly History” tab in columns AL-AS. Now we can sort the buying power changes as shown here.

### In search of correlations…

Next, I wanted to explore the correlation between a currency’s buying power change in one week and its price action (vs. the USD) in the following week. For the seven USD pairs, I normalized the following week’s pip changes. So again, the positive changes are sorted on a scale from 0% to 100% and the negative changes are sorted on a scale from 0% to -100%.

I put the following week’s price action on the same line as the current week’s buying power change. So we’re looking at the buying power change from week 1 to week 2, and comparing that to the price action from week 2 to week 3. I broke these out by currency pair starting in column AI.

### Results!

Now comes the fun part (not that all of this hasn’t been loads of laughs up until now of course). For each currency, I made a scatter plot of the buying power change vs. the price action. This is why I wanted to normalize both types of data onto bounded percentage scales. The scatter plots are down in the AA45 area of the “Analysis” sheet, and I’ve arranged them so that USD counter currency pairs are on the left and the USD base currency pairs are on the right.

We can clearly see a pattern here. While the correlations are slight (R-squared values averaging just over 6%), all of the USD counter currency pairs show negative correlations while the USD base currency pairs all show a positive correlation. The probability of this happening purely by chance is 1/(2^6) or about 1.5% (why? Think of flipping 7 coins to get either all heads or all tails. The first coin will be either one or the other, say tails. Then you’d need to flip 6 more tails to make them all the same. The probability of flipping one more tail is 50%, of flipping two more is 25%, etc.).

What does all this mean, jelly bean???? Well, first of all, stop calling me jelly bean. Secondly, what it means is that a currency tends to move in the opposite direction from its prior week’s buying power change. So when the EUR’s buying power goes up, the EUR/USD tends to drop in the following week. Thus, EUR/USD, GBP/USD, AUD/USD and NZD/USD all had negative correlations. Similarly, when the JPY has a strong rise in buying power, the JPY tends to fall during the following week, meaning that the USD/JPY goes up. Thus, USD/JPY, USD/CHF and USD/CAD all had positive correlations.

### Quick and dirty testing

So I said to myself, “Ok, so how would I trade this, Jelly Bean?” What if each week I had simply bought the currency with the weakest buying power change and sold the one with the strongest change?

I don’t have the currency cross data, so the test consists of buying or selling the corresponding USD pairs. For instance, if the JPY was strongest and the EUR was weakest, I would want to be short JPY and long EUR. In reality I would just buy the EUR/JPY of course, but in the test I had to buy both USD/JPY and EUR/USD. In the case where USD was the strongest or weakest, this issue goes away. So if the USD was weakest and the CHF was strongest, I would sell USD/CHF.

So there were one or two test trades each week. All the exciting Excel acrobatics for tabulating this are on the “Analysis” tab starting way over in column AX (note that there’s some more exploratory analysis even further over to the right, but I didn’t really get anywhere with that). For each week, I just totaled up the pips gained or lost from each trade.

It turns out that the average gain in pips per week was about 23, and the cumulative trade results are shown in the chart below. This isn’t nearly as sophisticated as most commercially available backtesting tools, but I’m not tring to build a Swiss watch here; I’m just looking for overall directional biases.

### Bai Nao!

In the next post I’ll talk about turning this research into a trading tool for the dashboards. As always, stay tuned and…keep pipping up!

by
This entry was posted in Research. Bookmark the permalink.