by

                                                                       Chih Kwan Chen
                                                                        (June  30,  2004)

 
                                                                              Abstract

                               The correlation between the long term trends of productivity gain and US economic
                               growth is studied by stripping away wide gyrations due to the economic booms and
                               busts from both sets of data. For the preglobalization era, a strong positive correlation
                               is observed as expected. However, in the current globalization phase the positive
                               correlation has changed to a negative one, implying that higher productivity gain
                               rather induces slower economic growth. This surprising outcome is further traced
                               down to the long term trend of increasing US trade deficit that undermines the long
                               term potential of job creation due to both the forced automation and offshore job
                               outsorcing in order for US domestic industries to compete with foreign rivals.
 

1. Introduction

       It is often claimed that the productivity gain drives the economic growth. In this paper we will study the validity of this claim. First of all we must know what are the "productivity gain" and the "economic growth" mentioned in the phrase. Most people think that they are just statistics released by the government every quarter. In the next section we will show that the jumpy annual data (don't need to mention the quarterly data) can not be reconciled with the claim that the productivity gain drives the real GDP growth. Actually what we should study and the opening claim is refering to are the long term trends of the rate of productivity gain and the long term real GDP growth rate that are buried beneath the widely gyrating annual data pushed around by economic ups and downs and need a careful analysis to reveal.

       Section 3 is devoted to the detailed study of the long term trends of productivity gain rate and real GDP growth rate, and the relation between those two data sets. During the preglobalization era, that is, before Reagan era, there exists a strong positive correlation between the two data sets. However, in the current globalization phase, this positive correlation has turned into a negative one.

       The surprising negative correlation discovered in Section 3 is further analysed in Section 4. There is a strong indication that the negative correlation is the result of long term trend of expanding trade deficit that has damaged the long term potential of the society of USA to create jobs and thus reducing the long term growth potential of the economy.
 

2. Uncovering the long term trends of productivity gains

       We start the analysis by listing the mechanisms that can produce gyrations in the annual data of productivity gain rate.

Mechanism 1: At the top of an economic boom, most people, including corporate managers, are extremely optimistic about the
                      continuation of the economic boom. Corporate managers will hire large number of workers to gear up their
                      production in order to meet the expected ever surging demands of their products. Then the boom peaks out,
                      the growth of sales stagnate, and the productivity of corporations plunges. When it becomes obvious to
                      most people that the peak of the economic boom has passed, corporate managers will panick and start
                      to lay off workers. As the race between the layoff and the falling sales heats up, the productivity gain rate will
                      arrest its sharp fall and eventually hits the bottom. As economy starts to recover, most people, including the
                      corporate managers, will be skeptical, and the corporations will not start to hire new workers. As sales pick
                      up, corporate managers will drive their bare bone work force to work to full capacity to meet the surging
                      demand. Thus productivity will raise rapidly ahead of the real economic recovery. This kind of gyrations in
                      productivity gain rate, if plotted on a graph, will look like preceeding the graph of real GDP growth rate.
                      However, as is obvious from the explanation, it is the economic booms and busts that are driving the gyrations
                      of this kind of productivity gain rate, not the other way around as required in the phrase that the productivity
                      gain drives the economic growth.

Mechanism 2: As an economic boom proceeds, corporations will put their idle capacity back to the production lines. The
                      natural way to do is to put the most efficient capacity among the idled back to service first. Thus as economic
                      boom matures and the sales growth slows, so are the productivity gains. When sales growth turns to negative,
                      capacities employed in the production lines need to be idled. This will be done by idling the least efficient
                      capacity first. Thus as the sales growth turns to negative, the productivity gains rate will turn to positive and
                      rises. The productivity gain curve from this mechanism also will lead the curve of annual growth rate of real
                      GDP. From the explanation it should also be obvious that it is the economic ups and downs that is driving this
                      mechanism of productivity gains, not the other way around as required in the phrase.

Mechanism 3: During an economic expansion, corporate managers have a natural tendency to want to invest in new
                      technologies to improve the efficiency of their productions. However, it takes time to bring those new
                      technologies into real production, so the productivity gains from this mechanism will lag behind the curve of the
                      annual growth rate of real GDP. The productivity gains of this mechanism apparently can induce higher
                      economic growth in future and belongs to the category that we want to study.

There are other mechanisms of productivity gains, and they will be discussed later in appropriate places since they will not cause wild gyrations in the productivity gain curve.

       Now let us look at the actual data themselves. The annual growth rate of real GDP from 1949 to 2003 (Ref. 1) are plotted as red dots, the annual growth rate of productivity (Ref. 2) as blue dots, and the annual growth rate of nonfarm payroll employment rate (Ref. 2) as green dots in Fig. 1 respectively. By a close inspection of the red dots and the blue dots we will

see that the blue curve has a tendency to fall first when the red curve is still hanging on at the high ground. This is a clear indication that the blue curve, the productivity gain, preceeds the red one, the real GDP growth. This is an indication of the domination of Mechanisms 1 and 2 in the gyrations of the productivity gain curve in Fig, 1as discussed above. Thus we need to eliminate the wide gyrations present in Fig. 1 in order to study the underlying long term trends of the data. The green curve, the change of nonfarm payroll employment, keeps up well with the real GDP growth rate, the red curve, except in the period after 1995; this discrepancy will be discussed later.

       The annual data of productivity gain and real GDP growth rate are plotted in the style of a correlation graph in
Fig. 2 (a). The correlation there is rather poor as expected. By taking 2 year moving averages of both data sets as in
Fig. 2(b),   the positive correlation becomes unmistakable, whereas gyrations from one data point to the next is reduced somewhat. By going into 4 year moving averages as in Fig. 2(c), the positive correlation is further improved.

With 4 year moving averages the gyrations of both data sets, of course, are reduced further. Average length of an economic cycle during 1949 to 2003 is longer than 4 years but shorter than 8 years. Thus the gyrations in 4 year moving averages, that is not presented here, are still due the ups and downs of economic cycles. To extract the underlying long term trends buried beneath the economi cycles, correlations between 8 year moving averages are shown in Fig. 2(d). In principle we can go even longer by taking more than 8 year moving averages. However, if we go too long, all the resulting data points will cluster around the center of the correlation graph and nothing interesting will be revealed. We consider 8 year moving averages as a good compromise and will regard 8 year moving averages of various data sets as revealing their long term trends throughout the remainder of the paper.
 

3. Correlations in the preglobalization and the globalization phases
 
        To further analyse the long term correlations between the productivity gain and the GDP growth, 8 year moving average data sets are divided into two time periods, from 1956 to 1978 and from 1979 to 2003, and are plotted in Fig. 3. The

graph for the period from 1956 to 1978 shows a clear positive correlation. We may confidentally conclude that the inovations in technology and the opening up of new high productivity industries have steadily pushed up economic growth during that period, and the quote at the beginning of this article that productivity gain drives economic growth is really refering to this kind of positive correlation between two data sets.

       For the period from 1979 to 2003, data points from 1979 to 1998 are plotted as black dots, and those of 1999 to 2003 as red dots. As for the black dots, though some kind of positive correlation still exists, the range of productivity gain becomes much narrower, from 1% to 2%, compared to the previous period where productivity gains ranged from 2% to 3%, and consquently the average economic growth is also lower than the previous period. If superposed on the graph of the period from 1956 to 1978, black dots in the period of 1979 to 1998 fall into the region labeled as "A". The productivity gain in the period from 1999 to 2003 has accelerated, but the correlation with the economic growth has turned into a negative correlation. This means that higher productivity gain rather has induced lower economic growth; this is totally in contradiction to the starting phrase to proclaim that productivity gain drives economic growth. It should be noted that the data points in Fig. 3 are all 8 year moving averages, thus the data point of 1999 is rather measuring the condition centered around the year of 1995. The red data points showing the surprising negative correlation centered around the period of the latter half of 1990's, supposedly the era of PC and internet.   In the next section we will analyse the cause of this surprising negative correlation revealed here.
 

4. The cause of the negative correlation

       The most desirable productivity gain is due to the emergence of highly efficient new industries; this kind of productivity gain will add to employment and will contribute to economic growth. The automation of an existing, but rapidly expanding industry will not cause the loss of jobs so it will also contribute to economic growth. The automation of a matured industry will result in layoffs; with the total sales stagnating with or without the automation and less employment as the consequence of the automation, this kind of productivity gain will reduce jobs and thus hurt the economic growth. Then there are a kind of forced automation; a struggling industry undergoes automation to compete with imported goods manufactured in low labor cost countries. This type of automation will result in massive layoffs and will definitely hurt the economic growth.

       Besides automation, productivity will change by moving manufacturing facilities out of the country whereas moving the facilities within a country from a high wage area to a low wage area will not change the productivity. To explain this statement, let us consider the following example: XYZ Co. sales its products to consumers so that its annual sale of 1 billion dollars is directly tabulated into GDP as the final sales. It employes 1000 workers, 900 in the manufacturing facility and 100 in the headquarter. Its manufacturing facility is in a high wage area. The expenses of XYZ Co. are as follows: 300 million dollars a year is for the buying of raw material used in the production. To simplify the argument here, let us assume that all the 300 million dollar raw material are imported from overseas. The labor cost for the manufacturing is 500 million dollars a year, the expenditure of the headquarter is 100 million dollars a year, and the profit of XYZ Co. is 100 million dollars a year. The contribution of XYZ Co. to GDP is 700 million dollars a year since 300 million dollars of imports  for the raw material must be subtracted as net import from its total final sales of 1 billion dollars. Thus the productivity of XYZ Co. is 700 million dollars divided by 1000 workers, that is $700,000/worker/per year. Now suppose it lays off 900 factory workers, move the manufacturing facility to a low wage area within the same country, rehire 900 workers in the new area, and consequently reduces the labor cost for manufacturing by 100 million dollars a year. Also let us assume that XYZ Co. keeps the price of its procducts steady. Thus the profit of XYZ Co. doubles to 200 million dollars a year. Now for the productivity, the contribution of XYZ Co. to the final sales category of GDP is still 700 million dollars a year, and the number of workers is still 1000, so the productivity of XYZ Co. does not change. Suppose XYZ Co. outsources its manufacturing facility to a developing nation. The labor cost in the new host country is now only 200 million dollars a year. Suppose that XYZ Co. still keeps the price of its products at the same level. The profit of XYZ Co. becomes 400 million dollars a year. How about its productivity? Since XYZ Co. must reimport 500 million dollar worth of products to be sold to the consumers in this country, its contribution to GDP is reduced to 1 billion dollars - 500 million dollars = 500 million dollars. Its domestic workforce consists of 100 workers in the headquarter, so its productivity is now 500 million dollars divided by 100 workers = 5 million dollars/worker/year. Thus the productivity of XYZ Co. jumps almost 7 folds by just moving its manufacturing facility outside of the country.

       As clear from the above discussion we suspect that it is the run away trade deficit that is boosting the productivity by forcing many struggling domestic industries into automation in order to compete with inexpensive imports and inducing many to outsource their not very high tech manufacturing facilities to low labor cost countries; this kind of activity, though boost productivity sharply, will erode the long term job creation potential of this country and thus reduces the long term economic growth as we observed in the red dots of Fig. 3. To confirm this suspicion, 8 year moving averages of the ratio of merchandise trade deficit over nominal GDP are plotted versus 8 year moving averages of the productivity gain in Fig. 4. Various periods are plotted with

different colors. The blue dots are for the period from 1957 to 1969; during this period USA had a merchandise trade surplus about 1 % of GDP, and the productivity gains scattered around from 2% to 3%, indicating the lack of correlation between those two data sets. Red dots are from 1970 to 1983. There is a broad trend that larger the trade deficit, lower the productivity gain. However, a close inspection reveals that the red dots are grouped into three clusters and with two huge gaps separating the clusters. Two gaps between the clusters of red dots correspond to two energy crises. When oil price jumps, trade deficit will explode. However, trade deficit expansion due to commodity price jump will hurt the economy and lower the productivity gains, not like the trade deficit based on the manufactured goods that will in general boost consumption and stimulate economic growth (Ref. 3 and 4 ).  The black dots are for the period from 1984 to 2003; a strong correlation between the rising productivity and the expanding trade deficit is clearly displayed. Most of the black dots not labeled correspond to black dots of the section 1979 to 2003 of Fig. 3. The black dots from 2000 to 2003 correspond to red dots in Fig.3 and are the ones forming the negative correlation between the long term productivity gain and the long term economic growth. The points after 2000 are showing a much stronger correlation between the productivity gain and the expansion of the merchandise trade deficit.

       In Fig. 5,  8 year moving averages of yearly changes in nonfarm business payroll employment are plotted against 8 year moving averages of the ratio between the merchandise trade deficit and the nominal GDP. Various periods are tagged with different colors and with explicit labellings. Here we direct the attention only to the blue dots labeled from 97 to 03.

At first within the period of blue dots the rise of productivity and the rise of job creation potential went hand in hand, indicating that the opening up of new industries, that is, reflecting the proliferation of PC and the emergence of internet society, has created many new jobs more than compensating the job loss due to the steadily expanding merchandise trade deficit. However, as 8 year moving averages of trade deficit ratio moves above 3% of  GDP, the erosion of job creation power is clearly displayed. In calculating 8 year moving average of the coming dot of 2004 we need to subtract the number of jobs created in 1996 and add the number of jobs created in 2004 to the sum of jobs from 1996 to 2003. In 1996 totally 2.4 million jobs were created. If the jobs created in 2004 is less than 2.4 million, then the job creation ratio will certainly drop further. On the other hand the trade deficit ratio will move toward 5%. This means that the 2004 dot will slide down the curve toward the lower left corner further. Considering the numbers of jobs created in 1997, 1998, 1999 and 2000 are all around 3 million, the slide down of future blue dots along the curve toward the lower left corner of Fig. 5 will probably continue for quite a while.
 

4. Conclusion

       The observed negative correlation between the long term trend of productivity gain and that of economic growth in recent years is traced down to the rapidly expanding merchandise trade deficit of USA. The run away trade deficit is causing the rapid rise of the long term productivity gain through forced automation and offshore job outsourcing, but is steadily eroding the long term job creation potential of the society and is causing the decline of the potential of the long term economic growth.
 

References

1.  Bureau of Labor Statistics

2.  Bureau of Economic Analysis

3.  Article No. 1 posted on this website

4.  Article No. 2 posted on this website