Commander in Pips: One of the trading strategies that widely spread over the market is “news trading”, when traders just wait for significant macro data releases from Big 7 countries and maybe from some others. All that they do is make trades that are based on news releases. The foundation of their trades is the difference f the real number from expected number. The greater the difference is the stronger the move on the market could be. Although it looks simple in description, actually there is no rule of thumb that could be applicable at all cases – just use it and you will be richer than Mr. Buffet. No way. This trading system has its own stumbling blocks.
Pipruit: And what are they? For me this system looks nice and pretty simple – just trade in the direction of surprise, that’s all.
Commander in Pips: And when you will treat difference as surprise? What conditions will you apply to specify “surprise number”?
Commander in Pips: I see. Ok, let’s continue with the data release moment. Usually market reaction consists of two stages. The first one is very short and comes usually during some seconds after the data has been released. This stage could last for some minutes. When the first rush comes down a bit and traders make some preliminary assessment of released data, the market starts to show a longer-time move due to the released data. The first reaction could be as along with continuation or as opposite to the following move. Usually the second direction is the one that the market will follow in the nearest future to price-in new numbers. So, how to explain the initial splash, and why it sometimes agrees with the next move and sometimes is against it?
Pipruit: Hm, good question, Sir. Actually I have not thought about it yet…
The major explanation here first, stands in consensus forecast number of this data and second – how much different from that number market already has priced in right at the moment of actual number release.
Additional important question is how popular is the current consensus forecast number.
Pipruit: And what is consensus forecast?
Commander in Pips: Well, consensus is some kind of agreement between some famous traders, economists, analysts and big investors, financial institutions on expected news or macro data. Consensus on macro data usually represents the average number of those who has taken part in a head count. Usually consensus forecasts on macro data are provided by big information agencies, such as Bloomberg, Reuters and others and all other participants point them in their schedules. Here, for instance, is Yahoo’s schedule and information about an expected Non-farm payrolls release:
See “Market Expects” field – that is consensus forecast or market expectations.
Pipruit: And what is “briefing forecast”?
Commander in Pips: Briefing.com is a Yahoo’s partner, so it publishes their forecasts for major data. Almost any world-known financial institution (as JP Morgan or other similar size) has its own opinion on major news or data, but usually this information is not free and provided only to clients. Still, it could be in consensus forecasts, since big banks and other institutions will probably provide their expectations, to, say, Bloomberg, concerning Non-farm payrolls release.
So, as we’ve said numbers that were provided to create a head poll on some data will be averaged out by information agency and published at the eve of real number release. You even can create your own consensus forecast if you’ll find in the net data from respected authorities, major banks and solid financial companies.
This consensus forecast treated as the zero point for future actual number comparison. And there could be three different results:
- In a row with expectation or at expectation, when actual number has been released right at consensus or only slightly different no matter in which direction;
- Better than expected, when actual number is more positive in economy terms than forecasted. And this “better” is tangible;
- Worse than expected, when actual number is more negative in economy terms than forecasted. This negative difference should be greater than in “as expected” scenario and also should be tangible.
As a result we get two questions:
1. Will actual data meet consensus or not?
2. If not, how big difference will be?
Since consensus forecast number becomes well-known some time prior to the release date, the market usually prices it in before the moment of release. So the major potential fluctuations on the market at the moment of release will depend on how large difference will be between actual data and expected number. It’s obvious that the greater the difference the greater the moves and volatility and the faster the market will swing on the way in a row with that difference.
Unfortunately it almost impossible to predict, how much the market has priced in and has it priced in something at all. That’s why many traders skip the moment of release, stay flat and open positions only after the first fast reaction.
Usually a trader passes through heavy thoughts as “what will happen if…” and keeps an eye on data and market sentiment right before the release. Still, there could be very accurate work to be done, that will allow you to increase accuracy significantly. And that work is statistics. What the major issues of that work are:
1. Categorize macro data in groups, for example “extremely important”, “medium importance” and “low importance”. This is necessary, because far less macro data is very important to make the greater difference that we need to shift market;
2. Make an observation for all major macro statistics past moments of releases, what difference has taken place and what market move was. This will tell you, what the smallest difference should happen to move market with particular data – GDP or, say Retail sales etc.. For instance, if Non-farm Payrolls will release 10% lower it historically moves EUR/USD up for 50 pips. Also you will find the borders “as expected” results, say, if data releases with less than 5%, the difference from consensus market usually does not show any respect to that. And when you will see that Payrolls has been released just 3 % higher – you will skip this trade and save a lot of money. Peter O does the excellent work with this.
Pipruit: Commander, I’ll try to do that, but I still have one question. Sometimes, even if the data shows an upward surprise, th market turns in the opposite direction, at least initially, but sometimes permanently also. Why?
Commander in Pips: There could be many reasons for that, I’ll show you just some of the more common, ok?
1. When data surprise is insufficient, compares to consensus forecast. It usually happens, when the market prices-in a greater number than consensus forecast, because consensus could be conservative a bit. Then if market shows positive surprise, but not so significant as market expects or rumors tell – then it could lead to price movement in the opposite direction;
2. This also could happen due to position closing of those traders who entered the market prior to data release. When the market has not confirmed their expectations – the first reaction could be opposite, but then market could start to move in a row with actual data;
3. A bit different situation was during Japan’s biggest earthquake, when JPY has become even stronger during first 5 min after news has been released. That was due to closing of carry positions. We will speak about carry trade later. But the major idea tells us that investors borrow money with low interest rates, and invest them in currency with higher rates. Hence initially they borrow Yens, sell them and buy other currency. To payout the loan they need to buy back Yens first, because they were scared with uncertainty about perspectives of the Japanese economy. So, that “Back buying” of some part of carry positions was the reason for Yen fast appreciation.
4. And the last reason is data revising. Very often the government statistical authorities revise previous data due adjustments and more precise calculation. Let’s speak about it a bit.
Here is the same part from Yahoo schedule of macro statistics, but for August. You may see what the market has expected and what number actually was released – 40% upward surprise, that’s impressive. But that is not all – look at two last columns – “Revised From” and “Prior”.
This report releases monthly. In our table we mention the release of NFP for July. Two last columns are dedicated to June. At the day of June report release, it has shown that May (or prior) number was just 18K.
Then, the turn of July release has come and it shows 117K compares to 84K expected. But also it has shown that June number was revised upward to 46 K – that is additional +28K. Read it like that:
“previous month figure was revised from 18K to 46K”
So, total surprise is:
(Current month difference + revision)/current expectation= ((117-84)+(46-18))/84=72.6%
That is quite another picture. The market could just explode with that surprise. You have to keep an eye on revised data and always use it in surprise calculation. Otherwise you can be confused by results. Say, the market expects +15K, but real figure comes at 5K – this is negative surprise, but if previous data was revised upward for 20K, the picture turns drastically – it becomes positive!
And vice versa, the revision could totally negate the current surprise, if it was in the opposite direction. So, that is very important and made for different data, not only for Non farm payrolls.