Having acquired some GDP data (before the ONS website change), I had a go at drawing some graphs to look at patterns and features. The raw GDP data are quarterly seasonally adjusted chained volume measurements, using 2006 prices, for like-for-like comparison. The data set runs from Q1 1955.
I’ve added some “random” thoughts and questions to see if we can make sense of the system we’re observing here, with some cursory analysis. Perhaps you might wish to add you own, or comment on mine.
First off, plotted the time-series and look for features – see below. X-Axis : Quarters y-Axis: Quarterly GDP (consolidated) in £Millions.
In this case I’ve highlighted significant recessional dips or troughs, and highlighted where and how long it took GDP to return to the level just before dive, in quarters. I wanted to treat the whole dip as an event , rather than just the recessional part of it i.e. the sustained negative gradient part of the curve. I think that it is not until you reach where you left off that you can feel confident that you’re back on track or can pick up where you left off. Arguably that translates into a “confident economy” and a mentality of re-investment and sustained growth. I’ve also added lines linking the end one dip to the start of another, then compared them with the previous “non-dip growth period” (ndgp)
- Interesing to note that each successive non-dip growth period has a steeper gradient than the last.
- Also the 3 ndgp’s are successively longer.
Now if the overall behaviour of the graph was exponential i.e. the economic system output with the adjustments would inevitably show exponential pattern over time, then this might not be unexpected. However, another feature seems to be that bewteen dip-events the gradient of GDP is rather linear.
Q1. Does the nature of the dip have an influence on the gradient of subsequent period of linear growth ?
Q2. Is it inevitable that this growth will be steeper than the previous growth period ?
Q3. Are the growth periods actually linear ? Let’s check…
Q4 1976 – Q1 1979 : 0.9085
Q3 1983 – Q1 1990 : 0.9844
Q4 1993 – Q4 2007: 0.9985
Wow. Very linear – particular with the longer data more recent growth periods (noise lower). Q3 answer – yep, appears to be linear. Mark as “interesting” and return to later.
Just checking the whole set (1955 to 2010) it fits an exponential model slightly better than a linear one (R2: 0.9922 vs. 0.9647), which supports the initial observation. The equation that Excel spews out is : y = 93293. exp (0.006x). See below. Mark as “interesting” and return to later.
I suppose that if the underlying mechanisms or system within an economy driven by policy towards GDP growth result in an exponential time-series, then Q1 & Q2 are partly answered (though not the cause & effect bit – see Causal Diagrams for an idea)
Q4. If there is a overall exponential character to the long term trend can we make any predictions ? e.g. when will the current dip will return to the “norm” or “mean pattern”.
Looking at the latest event, we seem to have moved quite a way off the mean track. Has the system essentially changed and/or is the “perturbation” so large that we won’t see a return to the norm for some time ? Mark as “interesting” and return to later.
- A “tick-like” shape, from the start of one recession to the start of the next can be seen. It starts to create a step-like pattern around the underlying trend.
Q5. Do the dip events themselves essentially have the same character and/or are they a function of the underlying system behaviour ?
Let’s look at the shape of each dip and try to make some comparison. I’ve started on this already, but some other interesting results here might extend the analysis (or playing around with data) – Subject for another blog.
That’s it for now. I will return to this theme and expand.