Category Archives: growth

Will the Real Real GDP in Turkey please stand up?

In the previous two posts on Turkey’s revised GDP statistics I showed 1) how revised GDP growth numbers diverge from the older version after 2009, 2) the importance of construction investments,  and 3) how standard policy variables failed to explain the revised growth figures.

Especially the previous post used cross-sectional data to show that global relationships between policy variables failed to predict Turkish GDP growth. In this post I will take a more direct approach at trying to validate the new official GDP statistics using time-series variation in Turkey. To the extent that Turkey’s revised GDP numbers reflect variation in economic output, there ought to be alternative measures of real output varying in a similar fashion.

One option would be to compare Turkey’s GDP growth with that of a standard leading indicator like industrial production (IP) or retail sales, all three which are published by the Turkish Statistical Institute (Turkstat). Yet the point of the revision appears to have been to improve statistical collection capabilities by using expanded administrative record of firm activity and so on. In this case, any difference between such a series and GDP data could simply reflect differences between recently revised statistics and soon-to-be revised statistics. It is therefore preferable to seek measures of economic output that are less sensitive to changes in Turkstat’s statistical capacity.

The recent debate over the quality of GDP statistics in China is the starting point here. Leaked Wikileaks cables from 2007 showed the now premier of the state council suggesting three alternative measures of economic output instead of the official GDP data: electricity, rail freight volume, and bank loans. Below I make some adjustments to this in order to incorporate factors more specific to the Turkish growth model.

For starters, rail is not the dominant mode of transportation. And electricity is but one source of energy. Therefore, it makes sense to further include total energy consumption and to focus on total inland freight (which would also include road and sea freight). Emissions of carbon dioxide is another potentially industry-relevant predictor. And finally, in addition to the credit used to financed these investments, I’ll also include construction permits, both as a measure of construction activity as well as a leading indicator for economic output.

The rationale underlying the choice of the above measures is the following: Turkey’s higher growth numbers have been bolstered by construction investments (on the expenditure side) and both industry as well as construction output on the production side. Consequently, new infrastructure or buildings – evidenced by more construction permits – ought to require increased energy use as well as moving goods and workers from one place to another. The associated economic activity (the actual construction process as well as any following activity in the completed investment projects) should simultaneously also result in higher emissions of greenhouse gases.

For this blog post, I employ mainly  the following eight variables (mostly in the form of growth rates):

  1. Industrial production index: not because it fits the purpose very well, as explained above, more as a reference to start with. The data is from Turkstat, available here.
  2. Primary Energy Consumption, in million tonnes of oil equivalent (mtoe), from the BP Statistical Energy Review.
  3. Electricity generation, in terrawatt-hours (TWh), from the BP Statistical Energy Review.
  4. Carbon dioxide emissions, in million tonnes, from the BP Statistical Energy Review.
  5. Total inland passenger traffic, in million passenger-kilometers, from OECD.
  6. Total inland freight, in million tonnes-kilometers, form OECD.
  7. Total bank loans, in CPI-deflated Turkish lira, from the Central Bank of Turkey
  8. Construction permits, in CPI-deflated Turkish lira, from Turkstat

At first, we would like to make sure the above variables are able to predict any of the Turkish GDP growth before 2010 – the breakpoint after which the new and old GDP series diverge.


The panels are ordered by how well they predict annual growth in Turkish GDP. IP has the highest R-squared at roughly 0.9 – unsurprising, as it’s likely drawn from very similar data as GDP and so less useful for my purposes here. Still, both energy use and electricity consumption predict a significant share  – 63 and 50 percent respectively – of the variation in growth. Interestingly, the worst predictor is inland freight. 

That said, overall these variables appear to do a fairly decent job of predicting real GDP growth. So how have they fared when comparing their growth rates during the periods 2004-2009 and 2010-2015, i.e. the two six-year periods before and after 2009/2010 in Turkey which I focused on in the previous post?

Below, I show how the growth rates from the energy-related outcomes stacks up against GDP growth (again, this is all Turkey-only data):


In the earlier period, most energy measures appear to grow at roughly the same rate per year as real GDP (although electricity generation is 1.2-1.5 percentage points higher). In the later period, although new and old series of real Turkish GDP both grow faster than the energy-related measures, it’s really the new GDP series that stands out, with a growth rate between 2.5-4.1 percentage points higher than the other measures. Most energy growth measures are also lower in the later period which would suggest lower, not higher, GDP growth in 2010-2015 compared to 2004-2009.

Thus, compared to the measures that explain the most of the variation in Turkish GDP (bar the IP measure), Turkey’s reported GDP growth is significantly higher than what alternative industry-related measures suggest. This is awkward, as we would expect extensive investments in construction to be more visible in these measures. It’s hard to believe that Turkey has undergone substantial energy efficiency to explain the difference in growth rates, especially when its annual power transmission losses as a share of output has remained at roughly 60-70% higher than the average among upper middle income countries. There has been some uptick in the use of renewable energy in Turkey, but it’s doubtful that this would have expanded as quickly as construction activity in the most recent years.

Below, I show the growth rates for freight and passenger transport.


This shows largely the same picture as the energy measures, with real GDP growing neck-to-neck with both freight and passenger traffic in 2004-2009 only to diverge substantially in the period 2010-2015.

The comparison with bank loans and construction permits below shows different levels to start with (these two have grown much faster than GDP in both periods) but in both cases their growth rates are nonetheless lower in the later period where revised GDP growth, in contrast, is much higher. This is also rather odd. We would have expected higher credit growth and higher growth in construction permits to be associated with higher real GDP growth, not the opposite.


Comparing these series to GDP growth one-by-one is illustrative to some extent but we might also want to use them to simultaneously predict what GDP growth ought to be.

For this purpose I first regress annual real GDP growth on annual growth rates for a subset of the above measures during the period 1981-2009: the first model includes, on the right-hand side, annual growth rates in primary energy consumption and electricity generation (as well as the natural logarithms of their values); the second model adds the growth rate and log level of bank loans; and the third model further adds the growth and log level of construction permits.  Omitted from this regression are the freight and passenger as well as the carbon dioxide measures, as it turns out that these have very little value in predicting GDP growth ones the other measures are included. (Including them changes nothing, it just eats additional degrees of freedom). The three models explain (have R-square values of) around 73, 82, and 83 % of the variation in GDP growth respectively during 1981-2009 (regression output is here).

I then use this model to predict out-of-sample Turkey’s GDP growth rate in the years 2010-2015 respectively. The average growth rates for the official data as well as those predicted from the three models are shown below.


Unsurprisingly, the model does a good job of predicting GDP growth during 2004-2009, but more interesting is that these composite regression models predict much lower growth rates than indicated by the new GDP figures from Turkstat. The three different models predict Turkey growing at around 4.1-7.5 percentage points lower than the most recent official series. In particular, the third model predicts Turkey’s GDP remaining entirely stagnant at on average zero growth between 2010-2015.

Curiously, in the previous post, the policy model I estimated predicted Turkey’s growth in GDP (per capita) to be roughly 4.5 percentage points lower than the revised official data. Thus, two very different methods of predicting growth in Turkish economic output suggest significantly lower growth rates than official data indicates.

Perhaps there are other measures of economic activity than the ones I have included that do a better job at explaining why the recently revised real GDP statistics coming out of Turkey shows such high growth rates during recent years. Yet even if there are such variables, it is rather odd that a Turkey’s phenomenal on-paper transformation from ‘Lounging House Cat’ to ‘Roaring Tiger’ cannot be discerned in commensurate growth dynamics with regards to energy consumption, carbon dioxide emissions, freight, passenger traffic, bank loans, or construction permits. Specifically, the former four tend to trace the old series’ lower GDP growth rates rather closely, and the latter two exhibit higher growth rates when GDP growth is lower and vice versa. Altogether, these variables predict GDP growth rates that are far below that of the official releases. At the very least, this ought raises some serious issues about the sustainability of economic growth in Turkey.

In conclusion, I have but one question: Will the real real GDP in Turkey please stand up?

Is New Turkey’s Growth Model From Outer Space?

The previous post focused the extent to which Turkey’s revised GDP data changed the recent history of economic growth in Turkey. A particularly striking fact of the new series is how much higher the growth rate in GDP in Turkey has been ever since the global financial crisis in 2008/2009. Of interest is then also how this changes Turkey’s economic performance in a comparative sense internationally, both in terms of economic growth as well as key economic indicators.

In this blog post, I take as the basis of economic performance the change in real GDP per capita obtained from the most recent October 2016 World Economic Outlook (WEO) from the IMF. For most of it I will show how this measure of economic growth – in two periods of six-year-averages for 2004-2009 and 2010-2015 respectively – correlates with a selected number of key economic indicators, most of which are included in the WEO database, and how much of Turkey’s (and other countries’) growth can be explained by a relatively simple regression model including a number of indicators of interest.

These indicators are: the natural logarithm of the average GDP per capita during the preceding six-year period, the natural logarithm of the average of population size during the preceding six-year period, the average growth rate in GDP per capita during the preceding six-year period, the current account balance as a % of GDP, the CPI inflation rate, the investment rate as a % of GDP, government debt as a % of GDP, and the unemployment rate. In addition, I also draw on the World Development Indicators database from the World Bank for labor force participation rate, domestic credit  to the private sector as a % of GDP, the age-dependency ratio, the urbanization rate, and from the IMF’s Balance of Payments database I also add the net international investment position (NIIP) as a % of GDP. For the investment rate, private credit, urbanization, and the NIIP-to-GDP measures, I also include a change variable with each measured as the change between the average during one six-year period and the corresponding average of the preceding six-year period. The IMF WEO indicators are quite standard and hopefully require little introduction. Most of added variables from the WDI are a bit Turkey-specific as they will show Turkey’s comparatively low labor force participation rate, the rapid growth of private credit in the economy, and the change in urbanization serving as a proxy for factor that could drive some of the large construction investments apparent in the new revised GDP series for Turkey. Continue reading

Constructing growth in New Turkey

The Turkish Statistical Institute recently released a revision to its GDP series (here and here), with some noteworthy consequences. Not only did the new series produce an upward revision of the level of GDP by around 20 percent (for GDP in 2015), but equally striking is the upward revision in the real growth rate of GDP after 2009 by an average of 1.8 % per year. The quarterly data is plotted below for new and old GDP and GDP growth rates (year-on-year) respectively.

The new statistics revision, taken at face value, arguably boost “the president’s economic arguments”, putting Turkey’s economy in a kinder light than previously thought. The timing is auspicious, as the government will likely try to revamp the constitution during the coming year.

The changes to Turkey’s GDP and growth rates are very large ones not only from an absolute perspective, but especially so in comparison with other cases of ESA (European standards) or SNA (United Nations standards) revisions. In OECD countries, such revisions have tended to have much smaller impacts on the levels of GDP and, at least on average, close-to zero effects on GDP growth. In most of these other cases, large revisions tend to be driven mostly by the changes in standards themselves, although in some of them, wider changes in how statistics are collected were more dominant. Continue reading

Picking Regional Winners and Losers in New Turkey

The Turkish Statistical Agency recently published a revised set of Gross Domestic Product (GDP) series, with some rather striking consequences for Turkey’s economy.

One less covered aspect of the new publication is the accompanying data on province-level GDP for the period 2004-2014. This is a welcome addition to students of Turkey, and I couldn’t help taking a look at it.

There are several interesting aspects of the Turkish economy that manifest itself in the data, such as the provinces of Istanbul and Kocaeli (the commercial and manufacturing centers of Turkey respectively) having US dollar GDP per capita levels in 2014 roughly similar to that of Saudi Arabia. Turkey’s economic output is also very heavily concentrated in the four largest provinces, Istanbul, Ankara, Izmir, and Bursa, which together account for almost exactly half of Turkey’s 2014 GDP, as can be seen below.  

As for the evolution of GDP per capita over time, the below graph plots the province-level values (deflated by the World Bank’s GDP deflator) indexed to their corresponding 2004 values. Interestingly, Kocaeli, has preformed rather well, whereas Ankara has growth the second-slowest of all provinces.


This graph shows amongst others the severity of the 2008/2009 crisis overall, as well as the sharp fall and subsequent sharp increase in GDP per capita in Kocaeli. In order to illustrate more clearly how individual provinces fared during the entire period, the below graph shows the growth rate in GDP per capita over the whole period 2004-2014, as well as two periods 2004-2009 and 2009-2014, by province. The left hand graph shows unadjusted growth rates, whereas the right-hand side graph adjusts the growth rates for the 2004 level of GDP ( in order to partial out convergence-driven growth).


As seen in the left-hand graph the period 2004-2009 exhibited lower growth rates than the subsequent 2009-2014 period (illustrated by the red dots lying more to the right of the red ones overall). Among the highest growing provinces are several small southeastern provinces, such as Siirt, Mardin, and Bingol. It is not surprising to see smaller provinces growing faster, as this is likely driven by forces of convergence, but even when one adjusts for the initial level of GDP per capita, this provinces still performed rather well.

One obvious difference between Siirt and Bitlis at the top of the growth ranking and Batman and Hakkari who are at the lower end of it, is that the former voted predominantly for the AKP, at least up until June 2015 (when Siirt and Bitlis produced two of the largest vote swings away from AKP that election). Similarily, pro-AKP provindes like Rize and Trabzon appear in the upper end with CHP-voting Denizli and Tekirdag closer to the bottom. Is this a coincidence or could it be that AKP voting predicts growth?

Below I show partial correlations from the following cross-sectional growth regression:

GDPpcGrowth_{i}^{2004-2014}=\alpha + \beta vshr_{AKP,i}^{2002}+\gamma X_{i}^{2004} + \epsilon_{i}

As for the set of controls X_{i} I use the variables already included in Turkstat’s province-level GDP data, namely the level of GDP per capita, population, the GDP shares of the agriculture and industry sectors respectively, and geographic fixed effects for whether the province is in the west, east, south, or north of the country. The below matrix graph shows a number of bivariate relationships being non-linear (in particular that between the AKP vote share in 2002 and both log GDP per capita in 2004 and industry share of output in 2004 respectively). For this reason I include quartile dummy variables for each of the plotted variables, allowing their correlations between the outcome and other explanatory variables to be non-linear.


Below, I show results from the growth regressions for three periods: 2004-2014, 2004-2009, and 2009-2014. Clearly, the results using the growth rate for the whole period hides the fact that it’s really in the first five years one observes a strong partial correlation between the 2002 AKP vote share and subsequent growth in GDP per capita, where a one percent increase in the vote share corresponds to a quarter of a percentage point increase in GDP per capita growth. In the second five-year period, the relationship is almost completely flat. There is furthermore no strong correlation between AKP vote share in 2002 and the 2004 level of GDP per capita.



The estimates for the full period as well as the 2004-2009 period are of medium magnitude: a one standard deviation increase in the AKP vote share, 13 percent, is correlated with a 2.6 percentage point higher growth during the period. Given that the standard deviation in the growth rate was 7.2 percent, such a vote swing would have had a meaningful bearing on economic performance.

These are merely partial correlations, and require strong assumptions to be interpreted as causal. There may very well be unobserved factors that correlate with both the political outcomes in the 2002 elections and subsequent growth performance that the model does not take into account.

It is noteworthy that several southeastern provinces appear to have significant pull on the regression line. Siirt, for example, experienced a higher-than-predicted (by the model) vote share for the AKP as well as higher-than predicted growth for the entire period. In contrast, both Batman, Hakkari, and Sirnak had the opposite experience. To see this more clearly I reran the growth regression only including the 24 provinces in the east (and using only linear controls to not soak up all degrees of freedom). The below graph shows that the partial correlations are quite pronounced even for the subset of eastern provinces in Turkey.


One (at this point speculative) interpretation of the overall results could be that the AKP in the early period focused on making sure the provinces that voted for it did as well as possible, or that the provinces that voted for AKP in 2002 also came out of the financial crisis relatively unscathed. In the later periods, as AKP seeks to grow its electoral base outside its core geographic constituencies, more resources might have been allocated to provinces in order to attract voters, partly at the expense of its existing core regions.

Moreover, if voters from swing provinces observed in the later period that their previous support for the AKP hadn’t resulted in comparably higher growth, this may have contributed to the shift away from the government in the June 2015 elections (pointing to an economic as well as political reason for abandoning the AKP in the southeast).




The Staggering Economic Costs of the Syrian Civil War

Lately, I’ve seen a lot of statistics about the economic consequences of the Syrian Civil War, much of it disturbing evidence as to the scale of the suffering. For example, one report by the Syrian Center for Policy Research (SCPR) published in March 2015 claimed that Syria had lost more $119bn in Gross Domestic Product (GDP) since the outbreak up until 2014, and that “total losses” amounted to $220bn when comparing to a scenario without the conflict. For a country whose GDP in 2007 was valued at $40bn, this represents an enormous dollar loss in Syria’s output.

Another report, published by UNWRA, made the claim that

“[e]ven if the conflict ceased now and GDP grew at an average rate of five per cent each year, it is estimated that it would take the Syrian economy 30 years to return to the economic level of 2010”.

These are all striking ways of describing the economic costs of the Syrian conflict. At the same time, neither the UNWRA report nor the SCPR is very specific about how it arrived at these quoted estimates and so I felt the urge to take a stab at this in my own way, while also expanding the alternative “non-crisis” scenarios a bit more.

As for GDP, there’s a disclaimer to be made about it only being just one measure – an imperfect one, at that – of economic output, and as for measuring living standards, its per capita variant is but one of many candidates, but as GDP remains the quintessential summary of an economy’s productive capacity, it is the focus of this blog post.

Continue reading