By Claus Vistesen: Copenhagen
(click on pictures for better viewing)
Last week, we learned that industrial production rose yet again in Japan clocking in at 1.4% month-on-month in September after having increased by 1.6% in August.
Companies said they planned to increase production in October and November as well, indicating the recovery from a record export collapse in the first quarter is holding up. Growth in China is generating sales for manufacturers including Hitachi Construction Machinery Co., which this week said it has worked off stockpiles that piled up during the recession.
“The pace of the recovery is faster than expected,” said Hiroshi Miyazaki, chief economist at Shinkin Asset Management Co. in Tokyo. Withdrawal of stimulus in the U.S. and Europe may cause output and exports to slow down this quarter, Miyazaki said, “but so far, today’s production report showed few signs of that.”
This is good news for Japan's economy even if it seems that Japan may simply be re-deploying old tricks in which companies are leveraging external demand but the domestic economy remains unable to pick up on the momentum. As ever, the disconnect between the level and flow of domestic activity (and price pressure) created by the domestic economy and the additional boost from external demand and asset income remains one the main perspective through which to look at the Japanese economy.
Within this context, the notion of Japan being dependent on exports to grow has emerged; initially as a strong market discourse and since in a more formal theoretical light in the form of the humble contribution of yours truly. It still represents a powerful market discourse and in fact, the idea of export dependency or reliance on external demand has been propelled to the main scene of the current economic turmoil as it has slowly but surely dawned on market participants and policy makers that the extent to which global imbalances need to be resolved, we have to find someone to run the deficits. And although this may seem a simple task, it has proved decidedly difficult to make the puzzles match in a world where deleveraging remains a key driving force on both the microeconomic and macroeconomic level.
In this entry I thought it would be interesting to look at a topic which combines the two perspectives above, that is; both the theoretical and the more market oriented narrative. On the former, this analysis would seek to move the analytical perspective down a notch from the pure macroeconomic level to a microeconomic level linking data on the company level (company accounts) with macroeconomic data (national accounts). On the latter, the analysis would provide some empirical foundation for the often cited relationship between a positive reading on industrial production/capex and a pick up in external demand, or more precisely the link between corporate activity and exports.
The analysis will be based on data from the Japanese trade ministry (METI) and OECD and will cover the period 1960Q1-2008Q4 (mail me for the excel sheet). On the company side, I will use data on sales (topline) and as well as profits (operating and ordinary). I will also distinguish between the manufacturing and non-manufacturing sector since one might expect, in Japan's case, the accounts of the former to be considerably more sensitive to external demand than in the case of the latter. With respect to national accounts I am using the OECD CARSA methodology which essentially signifies that we have current prices at annual levels with seasonal adjustment.
In line with the spin traditionally served here at Alpha.Sources, I will be looking at an increase in the connection between corporate sales and profits and external demand as an implicit function of age. This is to say, that this disconnect between domestic momentum and the ability of Japanese companies to generate revenues and thus growth based on external demand is a function of the increase in Japan's median age.
The main results of the analysis can be summarized in the following points.
- The positive relationship between the change in Japanese companies' profits/topline and the change in exports or the current account has increased markedly in a post 1990 and specifically post 1998-2000 context. This effect is predominantly a phenomenon observed amongst manufacturing companies.
- The empirical analysis suggest that Japanese manufacturing companies are now highly reliant on external demand to generate sales, profits and thus in some sense investment activity.
- The sensitivity of the sales of manufacturers to the volume of exports has increased by a factor of 60% from 0.25% to 0.4% around the period where Japan breaches a median age of 40 years.
- The sensitivity of the ordinary profits of manufacturers to the current account has equally increased markedly in the period where Japan has moved to a median age above 40. In the period after 1997 results indicates that a 1 unit (JPY) increase in the change of the current account has led to a 0.23 unit (JPY) increase in the ordinary profits of Japanese manufacturers which compares with a corresponding non-significant relationship in a pre 1998 context.
A Look at Company Performance, the Current Account and Correlation
Regardless of whether one squares the outlook on the Japanese economy, there is no doubt that the dent which the corporates have taken as a result of the economic turmoil is unprecedented.
Notice that I have indexed the charts with 1995 as a base year and then realize that current nominal value of company revenues has dropped to a level comparative to the one observed in 1993-1994 in relative terms. In absolute terms, the aggregate value of sales of Japanese manufacturers stood at some tn 84 and 82 JPY in Q1-09 and Q2-09 respectively which is value not observed since 1989 in nominal terms. This should give us a clear picture of drop in activity and then also the difficulty with which Japan will have in restoring productive activities back to normal whatever this might mean as we move forward.
With respect to Japan's external balance it is a bit more complicated, but there are some important points to remember as we move through the charts. Japan has been running an external surplus since the beginning of the 1980s, but as I have spent an entire academic paper explaining, it is only in the latter part of the 1990s and into the 2000s that this external surplus seems to be connected strongly with output growth. Moreover, it is important to distinguish between net exports and the income balance since the latter has been particularly important driving Japan's external balance in a from the 1990s and onwards
In light of the graph above, we can say that given the sharp decline in domestic growth in a post 1990 context external demand has take over, so to speak, both in terms of keeping national savings higher as well as contributing to headline growth. It is along the same axis that we would then expect the relationship between Japanese companies and external demand to have increased.
Moving on to some simple correlation analysis the following two charts which show the correlation between company sales and exports as well as the trade surplus/GDP will give us a nice initial overview of the data in question.
The representation is in changes (which is not unimportant) and smoothed by taking the correlation as a 4 year moving average (i.e. 16 quarters). The y-axis is ending period which means that a correlation for e.g. 1997 means correlation between 1993-1997.
Simply eye balling these charts does not seem to provide decisive evidence of the hypothesis of export dependency. Sure, we can easily see that the period 1997-2008 has seen a sharp increase in the relationship between company sales and the flow of exports as well as the share of external demand and GDP, but the key point to take away from these graphs is that they appear to be mean reverting (with a very weak positive time trend in the case of the second). This would mean then that the connection currently observed between exports and corporate sales is not unique. However, if we focus the attention on the second graph, it is also pretty clear that it is only in a post 1980 context that we have observed periods in which the correlation between sales and the trade surplus has been consistently and strongly positive. This would then seem to lend some evidence to the idea of export dependency and how this may be a distinct characteristic of contemporary Japan.
Moreover it would seem that it is not possible (except in the case of the correlation between sales and the trade surplus) to distinguish decisively between manufacturing and non-manufacturing. In later sections and using simple ordinary least squares analysis, it is however possible to differentiate this statement considerably.
Before we come to that though, it would be apt to use the initial conclusion above and have a closer look at the post 1980 period. Moreover, and courtesy of a more richer dataset on the macroeconomic level we can now augment the analysis with the income balance and thus the current account. This may seem trivial, but is very important in Japan's case since the income balance in particular has driven the external balance in recent years. From a company point of view and in order to be consistent, I will correct for the importance of the income balance by including company profits as the main gauge for company performance.
This chart (in level form and only for manufacturers) seems to be more supportive of the evidence of export dependency at least if we allow ourself the luxury to look only at the period from 1980s. The chart shows however that the strong positive relationship between the current account and the profits of companies is a relatively recent phenomenon which took off somewhere around 2000. Consequently, in the period from 1983 to 2000 the correlation between the current account and company profits in the manufacturing sector has been negative and in some cases strongly negative.
From this brief look at correlations, we should be satisfied that when it comes to the period post 1998 (more or less) the performance of Japanese companies have been strongly linked to external demand and income derived from external assets. Yet, this does not provides decisive evidence for export dependency measured as a strong and growing link between the performance of companies and external demand. In order to show this we must turn our attention to a bit more sophisticated statistical techniques although I can promise you that it won't be very fancy.
Some Models to Go With That?
The analysis which proceeds will center on the two following simple models which take the first difference or percentage change as a linear function of the change in either the value of exports or the current account.
The first regression will also be run with the sales of non-manufacturers as dependent variable in order to check the initial conclusion above that it is really not possible to distinguish between manufacturers and non-manufacturers.
Now, if you don't care about statistical analysis, you may stop here and move straight to the conclusion or go back to the summary in the beginning where the main results are reported. If you decide however to move on, rest assured that, following the models above, I never move beyond univariate OLS, so things should not get too complicated if you are a little bit familiar with statistical analysis.
Note that throughout, the full period will be Q1 1960 to Q4 2008, period 1 signifies the period where Japan had a median age below 40 and period 2 is consequently defined as the period in which Japan had a median age above 40.
If we begin with the first model that plots sales of manufacturers as a linear function of the volume in exports (both in % changes), the results for the full period, period 1, and period 2 regressions return the following results .
For those of you who are familiar with the results presented in my earlier work on Japan, these results should be well known. In this way, it appears that the relationship between the sales of manufacturers and the volume of exports has increased markedly, both in terms of the marginal effect as well as in the context of the overall fit of the model. Since this model is a log-log model, we can interpret the coefficient in percentages and in this way, the estimation indicate that the sensitivity of the sales of manufacturers to the volume of exports has increased by a factor of 60% from 0.25% to 0.4%. In words, it means that in the second period the estimation suggests that a 1% increase in exports will lead to a 0.4% increase in the sales of manufacturers whereas the corresponding number is 0.25% in period 1.
Looking at the overall fit of the relationship, the results clearly indicate that this representation leaves out a considerable source of the variation in the sales of manufacturers which is entirely to be expected. As always, it is essentially a qualitative and theoretical question whether the increase in the sensitivity as well as the goodness of fit (from 0.08 to 0.13) represents de-facto export dependency or simply indicates an increased reliance over and above other more important factors.
In relation to the second model which plots operating profits as a linear function of the change in the current account, it is important to note that this model is estimated in the first difference (and thus not log-log) because the current account in some cases has been negative. Moreover, the sample period is shorter than for the first model (1980-2008) since OECD does not have data for the income balance prior to 1980.
On an overall basis these results underpin those from the first estimation although they seem to confirm the hypothesis to a much higher degree. Abstracting from the full period result which serves as an anchor for the overall significance of the relationship, the difference between the model estimated for period 1 and period 2 is striking. Consequently, the first period estimation signifying the period where the median age of Japan is below 40 shows no significant relationship whatsoever, in this sense it appears that the apparent negative relationship implied above from the correlation charts do not pass the simple causality test which OLS represents. The second period regression on the other hand returns a strong and significant relationship which indicates that a 1 unit (JPY) increase in the change of the current account will lead to a 0.23 unit (JPY) increase in the ordinary profits of Japanese manufacturers. On the goodness of fit measure the only thing we can say is that it has increased considerably to signify the increase in relationship between the profits of manufacturers and the current account. However, whether a goodness of fit of 0.15 is high in an absolute sense here is difficult to say without a more thorough and comparative study. But since we have a univariate framework, I believe this result to be quite extraordinary.
I hope by now that I will have either convinced you or scared you off in terms of the importance of whether Japan is dependent on exports to grow or not. I would also hope that the connection to events closer to the market is not too difficult to see. For example, Bloomberg is running the story today that the BOJ, like most other central banks, is either willingly or, dragged kicking and screaming by market sentiment, moving towards the formulation and near execution of the famed exit strategy from extraordinary monetary policy measures. Clearly, interest rates are to remain low for as far as the eye can see, but it is interesting to ponder whether the decision by the BOJ scale back corporate debt purchases is related to optimism on the companies ability to leverage domestic growth or whether it is because they see export markets reving back up in which case it would be back to the same old growth strategy. Another example would be the latest inflation reading which suggests, more than anything, the extent to which the domestic economy in Japan is not able to provide an environment in which companies can operate profitably as well as it indicates how overall domestic momentum is essentially contractory.
It is within this general economic context that the notion of export dependency becomes important and specifically how this might relate to the ageing of Japan's population. In this entry I have tried to take this idea down a notch from the strict macroeconomic level in the form of an analysis of the relationship between external demand measured through national accounts and aggregate corporate accounts. The results, I believe, speak for themselves and strongly suggest I think that Japan indeed is becoming increasingly dependent on external demand to create the growth and income the economy needs to maintain economic growth. Following from this, a number of questions present themselves, not least the most crucial general question relating the issue of export dependency to ageing in a general sense and then on to the discourse on global macroeconomic imbalances. But for now, I will let you digest the data at described and analysed above.
 - Results of this regression is not formally reported in the text; please mail me if you want my excel sheet and results.
 - In order to be really rigorous I would have to formally test for the difference between the two periods (e.g. through a Chow Test or related method), but here it will suffice to look at the change over the period without putting a label of statistical significance on it.