[R-sig-finance] Is oanda.com data trustworthy?
a.trapletti at swissonline.ch
Mon Nov 14 12:24:15 CET 2005
>Date: Sat, 12 Nov 2005 18:35:06 +0530
>From: Ajay Narottam Shah <ajayshah at mayin.org>
>Subject: [R-sig-finance] Is oanda.com data trustworthy?
>To: R-sig-finance <R-sig-finance at stat.math.ethz.ch>
>Message-ID: <20051112130506.GT11451 at lubyanka.local>
>Content-Type: text/plain; charset=us-ascii
>The EUR/USD currency market is very liquid and the data should be very
>But I see big differences between data on the US Federal reserve
>website and data on http://www.oanda.com (accessed using
>Here are some examples:
> US Fed oanda
>31-Oct-05 0.833681 0.8293
>1-Nov-05 0.833472 0.829
>2-Nov-05 0.828706 0.8325
>3-Nov-05 0.835352 0.8286
>4-Nov-05 0.845451 0.8374
>When expressed as 100*log(p2/p1), the returns look like this:
> US Fed oanda.com
>2005-11-01 -0.02507268 -0.03618163
>2005-11-02 -0.57346603 0.42130667
>2005-11-03 0.79877448 -0.46956922
>2005-11-04 1.20170199 1.05643239
>These differences seem huge to me! E.g. on 3 November, the US Fed says
>that returns were +0.798% and oanda.com says it's -0.469%.
>Here's the exact get.hist.quote incantation:
>>> get.hist.quote("USD/EUR", provider="oanda", start="2005-10-31", end="2005-11-04")
>trying URL 'http://www.oanda.com/convert/fxhistory?lang=en&date1=10%2F31%2F2005&date=11%2F04%2F2005&date_fmt=us&exch=USD&exch2=&expr=EUR&expr2=&margin_fixed=0&&SUBMIT=Get+Table&format=ASCII&redirected=1'
>Content type 'text/html' length unknown
>2005-10-31 2005-11-01 2005-11-02 2005-11-03 2005-11-04
> 0.8293 0.8290 0.8325 0.8286 0.8374
>What should one do?
> -- Ajay Shah ajayshah at mayin.org http://www.mayin.org/ajayshah
The prices are very different if you do not sample synchronously or if you take average prices as Oanda does (Go to http://www.oanda.com/convert/fxhistory get a rate and read the text at the bottom). For example the daily low of EUR/USD on 11/04 was around 1.1802 and the daily high around 1.1996 (194 basis points difference!).Hence both numbers - Oanda and Fed - you report above are "correct". Further if you look at correlations of increments between different sources and use the fact that there is almost no autocorrelation on each series it is not a surprise that there is low correlation between the returns of the different sources.
More information about the R-sig-finance