As I mentioned in part 1, I wanted to look at the “dips” in the time-series in isolation to see if there are any similarities. By “dip”, here, I mean both the period of recession, where GDP decreases, and the period of growth where it increases to the level it left off. I will look at other forms of dip or recessional deviation from “the norm” later.
First off, let’s identify the dips to look at, namely those beginning in 1973, 1979, 1990 and the one we (the UK) are currently in which started in Q1 2008, then normalise them. In this case I have simply chosen a point on the latest dip by which to scale up the other dips. In the first graph you can see that I’ve used the start-quarter as the reference.
What can we see ?
- both the 1973 and 1990 dips have similar length and depth, although ’73 was clearly more noisey, or indeed a double-dip.
- The 1979 and 2008 appear to have different characteristics to the other two: depth is about the same, though slightly offset; the shape until the 12th or 13th quarter is not-dissimilar.
- Without thinking too much about what we know today about crises in the eurozone, US and their potential knock-on effects, we might predict that the 2008 dip would last roughly as long a the 1979 event : about 17 quarters, taking us up to the end of Q1 or Q2 in 2012. We’ll probably have first view on this in about June next year.
- With the sounds that governments and economists are making at the moment, it would appear that mid-2012 would be rather optimistic.
Let’s take another view on the data, and use the low-point of the dip as our reference.
Any more insight ?
- By doing this, the growth periods tend to bunch together and we might tentively say that these growth periods have similar trajectories once things start to pick up.
- Perhaps I need to align the low-points in time ?
Does this tell us more ?
- Well 1973 now looks a bit mad, bit does roughl finish at about the same point as 1990.
- 1979 and 2008 now look more similar during the growth period.
OK, probably nothing new here. Let’s look at deviations and patterns away from the “mean”. By “mean” I mean (ha-ha) the closest approximation to a long-term pattern that we found in the data in the last blog, namely y= 93293. exp(0.006x), which has an R2=0.9922.
IF there is some underlying system behaviour that dominates the GDP outcome, then we might say that booms and busts are simply deviations or indeed noise due to relatively short-term issues (or changes to inputs or parameters) and that the system outcome will inevitably return to it’s projected path, at some point. I suggest (at the moment) that this would not happen by iteslf i.e. do nothing and everything will be alright, eventually, but that it will be dependent on decisions made by key people on how to change system inputs (these decisions and processes in fact being part of the overall system). I will look at this using underlying behaviour using Causal Diagrams, and see what pops out.
So, back to the data. In the first graph, I’ve simply taken the expected GDP value due to the model away from the actual value to give us some sort of deviation view.
What can we see ?
- Well, the time-series is beginning to look more like a pendulum swing around the mean model, i.e. the x-axis in this case.
- The amplitude of swing is increasing over time (really agreeing with the exponential increase that the raw data shows), and appears to be a “under-damped” system. Most of this might be a function of something like inflation, not compensated for by using fixed 2006 prices to calculate the GDP in this time-series. I’ll return to this at some point.
- Interestingly we start to see other dips in the data; some above the mean or some below.
- This particular model only starts to become more stable beyond the mid 1970’s so perhaps we should not dwell on the early “noise” too much at this stage.
- I’ve highlighted the dips we’ve been looking at with star-shapes.
- Now look at the last few data points. With this perspective, I looks like we are still moving away from the mean, which on the basis of economic sentiment sounds like it might be the case. As we “flat-line” in absolute GDP growth, we are still moving further from the expected GDP level suggested by the long term model.
- Looking at previous dips, the time spent returning to the mean is much longer than the time taken to return to previous levels.
OK. Let’s go one step further and relativise the data according to the GDP at the time i.e. divide the deviation from the mean by the actual GDP (y) value, to give us a proportional deviation (in %).
What’s to see ?
- OK, we seem to have reduced some of the increasing amplitude, but the patterns remain the same
- The current period still appears to have the largest amplitude, at about 8% of GDP, and hasn’t bottomed out yet.
- Has the elastic snapped or is this just an effect of more global magitude?
- The shape of dips in this context are asymetric – a sharp fall, followed by a shallower rise. Having not reached the bottom yet, it’s trickier to predict the shape of things to come (but I’ll try) – let’s guess by eye to start with … the mean will be reached in 2019-20. Oh dear.
There’s lots more to think about so I will update further, but that’s it for now. Remember that GDP is a lagging indicator, only reflecting the overall outcome of the economy viewed in a particular way after many months lag. Thought: Is the use of GDP and growth an issue in itself, inevitably driving this system out of control in a positive feedback loop ?
Let’s go looking for other stuff : other measurables closer to real-time, underlying system structure & behaviour and maybe some key drivers or critical inputs.