Monday, August 18, 2014

2Q2014 GDP: Into Orbit

My, oh my, how things have changed (log annual and quarterly SAAR changes; 2005=100):


We ain’t talkin’ bout no base effect no more. T’ain’t bout prices neither. At 6.4% in percentage terms, the economy has put up a pretty solid growth number. If the low level of output in 1Q2013 influenced growth this year, that’s less of a consideration for 2Q2014. And if export and commodity prices trended up in 1Q2014, they’ve been flat or trending down in 2Q2014 (log annual and quarterly SAAR changes; 2005=100):


I had my doubts about “sustainable” growth in 1Q2014, both from the two factors mentioned above, as well as the divergence between Malaysia’s export performance as against that for the rest of the region. Now, we’re seeing much better trade numbers coming out of the Philippines, Taiwan and Korea, which eases my mind as to how much of this growth is “real”.

There’s some domestic factors at play as well (log annual changes; 2005=100):


The demand side is a picture of contrasts, with private consumption holding up (in defiance of nearly everyone’s expectations) while public consumption tanked. Real imports have slowed, but investment has ticked up. Exports of course, have been accelerating since 3Q2013, and that’s really the main contributor to growth this time around – net exports doubled in 2Q2014 compared to 2Q2013.

Underscoring the strength of growth is the drop in inventories – I was expecting the opposite, what with the growing divergence between export growth and IPI growth. That inventories fell suggests stronger output growth down the road, as companies restock, which goes against market expectations. But then the whole forecasting community (including me) have been behind the curve this year.

One reason why is that a lot of forecasters aren’t taking into account nominal income growth, which has been pretty strong (log annual and quarterly saar changes):


After declining for two years straight, income growth has started climbing again – 2Q2014 NGDP is 10% higher than last year’s. For a better sense of that means for the man on the street, here’s the corresponding GNI per capita growth figures (log annual and quarterly saar changes):


Nominal income growth has been exceptionally strong over the past four quarters, with 8% annual growth around the lower limit. The employment series is noisy, but I think that translates to about 3%-5% real annual income growth i.e. over and above the rate of inflation. The data on GDP by income approach provides some corroboration to this, with employee compensation taking an increasingly larger share of GNI. That’s a good enough reason why private consumption growth has been sustained, despite the cuts in subsidies over the past year.

So what’s this all mean going forward? Most forecasters are expecting some degree of mean reversion to trend growth over the second half of the year. There’s a solid rationale behind this, as higher growth in 2H2013 means the economy has a much higher hurdle to climb to maintain growth at the pace we’ve seen in 1H2014. Then there’s the technical difficulty that forecasting models (including all of mine) by construction will show tendencies towards mean reversion as well. Forecasters going by intelligent guesswork will have a psychological rather than a statistical barrier to seeing anything different.

As against all that, exports have climbed this year despite a really bad 1Q2014 for the US, and the European economy falling flat on its face (again). So in technical terms, we could be facing a structural break point where the path of growth shifts permanently higher. Which is it, I don’t know, but it would interesting to watch this unfold.

Technical Notes:

2Q2014 National Accounts report from the Department of Statistics (warning: pdf link)


  1. Hi Hisyam, could you please explain the rationale behind using log?

    I don't have problem in understanding qoq, yoy or annualised.

    In maths, what I do know about log is that, it's used to transform a non linear data into a linear one.


    1. Log usually compresses a series in a more meaningful way.

      Doing the non-log way, 1% of 10 is much smaller compared to 1% of 1 billion, and you can't really compare anything. Log adjusts the problem that comes with magnitude, allowing you to compare numbers of different magnitude better.

      Log is also a very good approximation of percentages for very large numbers (essentially no difference for percentages, unless you're a really, really big fan of accuracy). In fact, you can see the percentages right off a log chart vs just normal numbers. (This is probably the biggest case for using log).

      The very long explanation is here:

    2. Also, if you're into regression, log is useful to make the series kosher. Doesn't always work and it isn't the only way, but it's a useful trick when it does work.

      Why? Because for the very reason you mentioned: it turns non-linear series into a linear one. In fact, I think that's the primary reason most economists present they stuff in log: usually they eyeball the series first and if it looks linear after being log'd, they'd use if for regression.

      If it doesn't seem linear, then regressing the series probably gonna be problematic later

  2. I shouldn't say linear, but it's something like that. I've forgotten the term already.