3. Emergence of Taiwan’s Real / Nominal GDP Spread
3.4 International Price Levels
3.4.3 Anomaly Taiwan
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
-
Singapore: The large share of foreign migrant workers in low paying jobs implies that sectors with low productivity growth do not need to compete for employees with the high growth sectors, therefore the assumptions for the Balassa-Samuelson-Effect are not given. In a sense, ‘non-tradeables’ are cheaply imported bySingapore, which makes it a special case and disturbs comparability with other countries.
-
Luxembourg: As a city state in the center of the European Union it faces a situation similar to Singapore. Low geographic extension in combination with free labor and goods movement and the use of the same currency as neighbouring countries, many of the goods consumed within country borders are likely to not be bought within the country and many jobs are likely to be given to workers living outside the country, including jobs in the ‘non-tradeable’ sector.Doing so leaves 32 countries (Figure 3.4.2) with a GDP (PPP) per capita between roughly 20,000 and 65.000 and a ratio between 0.64 (Norway) and 1.99 (Taiwan). The logarithmic trend function becomes:
y = -0.7452ln(x)+8.8753 R2 = 0.3925.
(s.e.) (.1693) (1.7794)
3.4.3 Anomaly Taiwan
If we now remove Taiwan from the sample, the function becomes:
y = -0.8145ln(x)+9.569502 R2 = 0.6955 (s.e.) (.1001) (1.0515)
It is worth mentioning that the next largest outlier is the US with a ratio of 1 (as defined) and a predicted ratio of roughly 0.75. This is 2013 data, looking at the exchange rate shift of the US dollar at the end of 2014 and early 2015, we can expect the US now having a roughly 20 percent higher price level compared to most other currencies then at 2013. Therefore the US has recently moved much closer towards
‧
the ratio that is predicted by the model, hence supporting the purchasing power parity hypothesis and making Taiwan an even more extreme outlier by increasing the
distance between the difference of Taiwan’s predicted and actual ratio and the country with the second largest difference, and significantly increasing R2 (to roughly 0.74).
Therefore, not only visually, but also statistically, Taiwan is the only strong outlier. It is however possible, that those random deviations temporarily occur over time. I already showed that Taiwan’s ratio grew over time, it is hence not a random short-term/temporary movement or shock. Nevertheless it makes sense to make a time series analysis in order to see whether or not other deviations with comparable impact occurred.
Unfortunately, four major effects blur any look into the past of PPP ratios:
Firstly, as the calculations change, so does their outcome. With regular revisions of how prices are assessed and compared between countries and regions, data between years becomes less comparable the larger the time difference. The obvious reason for those calculations to be changed is that they are improved. This might be put into question at a detail by detail inspection - parts are controversial - but on a larger 15 scale the tendency is unambiguous. Consequently we can expect more errors in the price level estimations in the past, than today. Therefore, slightly higher deviations are to be anticipated, and we have to bear that in mind when comparing that data to Taiwan’s situation today.
Secondly, the definition of which country is developed and which is not, is quite arbitrary. That counts for today as it does for the past, however defining a year for which a country started to be a developed country is a difficult step to make, and one that no matter what definition is used, will always be impossible to defend. I will therefore stand back from doing so and consider all countries that the IMF defines as developed today, as developed countries for the whole time series. This is an approach shared by the IMF, which also does not give definite dates, but only the categories, as the IMF does not imply a country became developed the moment or year the IMF
Especially the recent change in 2011, which resulted in a huge boost of PPP income
15
for emerging economies relative to developed economies. This is another reason why this research focuses in developed countries as those are only marginally affected by this change.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
recognises the fact it now is. Hence, the further back our observation goes back, the more countries will show characteristics that we associate with a lower level of development, as they have been catching up during most of the period, and therefore the data becomes less reliable.
Thirdly, PPP calculations start with gathering huge amounts of information on which goods are sold how frequently and for what price. This requires a high level of collaboration between ICP staff and local authorities. It is not unlikely to assume a notable learning curve here, as management practises might strongly differ due to different sociocultural backgrounds, and while the private sector is fast to adapt best practises, the public sector - which is the one involved here - is notorious for slow adaption and more legacy work practises. Also on the individual employee level we can assume a learning curve, with its effect stronger the further we go back in time.
Fourthly, a significant number of countries has not been part of the First World during the period of the Cold War. As we know, Communist countries usually do not provide an efficient market, the price mechanism is distorted and in some instances not existent at all, which means the assumptions we need for the law of one price are not met. It also is plausible to assume a time lag, as the reforms of former Soviet Block countries took many years and their effects took many years more.
On order to assess whether there have been other anomalies besides Taiwan, I will compare a time series analysis conducted with Taiwan and one without Taiwan.
Using an ordinary least square regression in the form of : 16
!!
cf. Agresti and Finlay (2008)
16
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Starting in 1990, R2 starts off very low and Beta (excl. Taiwan) jumps from extremely low to extremely high and finally to a plausible level (Figure 3.4.3). This is due to bad data quality and the step by step addition of the former Soviet controlled Eastern European states. Hence it makes sense to only interpret the findings from about 1995 onwards. As can be seen, until 2003, R2 and R2 including Taiwan are very close, but then spread permanently. This is caused by Taiwan’s increasing ratio at that time. Also this is the only significant drop in R2, implying another outlier like Taiwan has never occurred during that period, i.e. Taiwan’s ratio relative to its income level is
unparalleled taking into account the current price diversity within the developed world.
Interestingly, Beta shows a declining trend. As the function is logarithmic, the implications of this change are more formidable than they appear on first sight. Also, the period of strong decline in Beta between 2002 and 2008 has its timing
conspicuously close to the decline in R2 including Taiwan (roughly 2003 to 2008).
This concludes Taiwan became an anomaly at around the time when the price differences between the other countries became smaller and the spread between
Figure 3.4.3 Price Level Time Series
0.25 0.50 0.75 1.00
1 2 3
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 Beta (excl.Taiwan) R^2 (incl. Taiwan) R^2 (excl. Taiwan)
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Taiwan’s actual ratio and the expected ratio stabilised at around the time price level differences between the other countries stabilised too.