A Long Term View on the Short Term Co-movement of Output and Prices in a Small Open Economy

Ola H. Grytten*

Arngrim Hunnes† April 9, 2010

Abstract According to a Keynesian view, short term output fluctuations are normally demand side led. Since prices reflect demand, they should mirror output fluctuations. Thus, prices and output are expected to move in the same direction in the short run. The present paper investigates the historical co-movements of output and prices for a small open raw material based economy, in this case Norway 1830 – 2006. We find little evidence of a positive relationship. On the contrary, we find negative correlations between the two variables, indicating that supply side shocks, often initiated by the foreign sector, were more important for historical business cycles in Norway than assumed hitherto. JEL codes: E31, E32, N10, N13, N14 Keywords: Prices, output, business cycles, demand, supply, economic history, Norway.

*

Department of Economics, Norwgian School of Economics and Business Administration (NHH), Helleveien 30, N-5045 Bergen, Norway. Phone: +47 55 95 93 45. E-mail [email protected] † Department of Economics and Business Administration, University of Agder (UiA), Service box 422, N-4604 Kristiansand, Norway. Phone: +47 38 14 15 14. E-mail [email protected] 1

1. Introduction The notion of corresponding inflation and output fluctuations has been a central assumption in economic history writing. This is in line with the empirical findings of Nicolai Kondratiev, who mapped long cycles in the economy by price series and John Maynard Keynes who explained historical business cycle downturns in general by negative shifts in aggregated product demand.1 Empirically, it can easily be illustrated e.g. that prices fell during the long depression from the mid 1870s till the late 1880s, during the post-war depression in the 1920s and during the great depression in the 1930s.2 In fact these examples of falling output and prices have given name to the term depression. Nevertheless, it is not at all difficult to point out the opposite. Prices tended to fall also during years of significant economic growth from the mid 1870s to the mid 1890s, and even in many countries during the booming 1920s. One may also find several examples of prices rising rapidly, despite output downturns, e.g. during the 1970s, when the combination of increasing inflation and stagnating output was named stagflation. Most writers on Norwegian economy history seem to agree that short term output fluctuations have been demand side led, when most will agree that it is normally supply-side led in the long run.3 However, for Norway very little work has been done in order to test the relationship between prices and output historically. Thus, in this paper we examine if there historically has been any empirical correspondence between short-term price movements and output volumes in the small, open raw material based economy of Norway using relatively newly published data from 1830 till present. To the authors best knowledge this is the first paper to use such a long time series to study this relationship for the Norwegian economy. Positive co-movements between the variables indicates that the business cycle historically most often is demand-side driven, when negative co-movement would indicate that the business cycle historically is more likely to be supply-side led. In fact, the latter is what we somewhat surprisingly basically find during the twentieth century until present days. Earlier empirical work on Norwegian data has for most parts used contemporary data. In the study by Bjørnland who uses data from 1967 – 1994 she finds that “the business cycles properties vary considerably with the de-trending methods used.”4 Hence, she finds both a 1

Kondratiev 1926, 573 – 609, Keynes 1936, 23 – 24. Kondratiev 1926, 573 – 609, Schumpeter 1939, 87 – 139. 3 Sejersted 1973, Hodne 1983, Bergh 1983, Furre 1991. 4 Bjørnland 2000, 369. 2

2

pro-cyclical and counter-cyclical pattern depending on the type of filter applied. Research by Husebø and Wilhelmsen using data from 1982 – 2003 indicate that “output and consumer prices in levels are negatively correlated, with prices leading output.” 5 This can probably, at least partly, be explained by Norwegian dependency on petroleum and petroleum prices, which can act counter-cyclical to international business cycles. Petroleum prices constitute a significant cost for the Western economies. Thus, demand will shift from output produced in the Western economies over to a necessary input to the economy. In consequence, prices increase and outputs decrease. Existing research on other countries include the study by Cooley and Ohanian who looks at the relationship between prices and output in US for different sub-periods 1822 – 1987.6 For the post-war period they find a negative correlation and the data shows a positive correlation in the interwar years. They conclude that “with the important exception of the inter-war period, these data are not at all suggestive of the stylized fact of pro-cyclical prices that many macroeconomists holds.”7 Smith finds in his study of ten countries (US, UK, Canada, Australia, Sweden, Italy, Denmark, Norway, Japan and Germany) that the relationship is procyclical until WWII and counter-cyclical for the post-depression period.8 A frequency domain analysis is performed on US data 1875 – 1994 by Pakko.9 He observes that negative correlations tend to be associated with low frequencies while positive correlations are associated with higher frequencies. Den Haan who introduces a new methodology by analyzing correlation coefficients of VARs at different forecast horizons finds positive correlations in the short run and negative correlations in the long run for US 1948 – 1997.10 The same conclusion is reached for the G7 countries (Canada, France, Germany, Italy, Japan, UK and US) in the postwar period.11 Parker extends the work by den Haan by looking further back in time.12 That is, he studies the years 1875 – 1914 and 1920 – 1941 in the US and concludes that pre WWI the relationship between output and prices are strongly pro-cyclical. For the interwar years he also looks at UK, Belgium, Canada, Germany and Sweden and finds that the price behavior is “overwhelmingly pro-cyclical across a number of different

5

Husebø and Wilhelmsen 2005, 11. Cooley and Ohanian, 1991. 7 Ibid., p. 47. 8 Smith, 1992. 9 Pakko, 2000. 10 Den Haan, 2000. 11 Den Haan and Sumner, 2004. 12 Parker 2005. 6

3

countries, confirming what it appears the economic literature has determined.”13 Further, Parker concludes “it appears as if the emergence of counter-cyclical price behavior is a postWWII phenomenon.”14 However, the same pattern of lacking positive correspondence between prices and business cycles during the last decades is also discovered for the US, which in nature is a quite different kind of economy.15 On the other hand it is also found evidence of co-movement between prices and output in the US economy since World War II.16

2. A small open raw material based economy Ever since the first steps towards an international Norwegian economy, when the Hanseatic League established itself in the foremost Norwegian city, Bergen, in the 1350s, Norway’s export sector relied upon raw materials. Fisheries and forestry made up the two most important export industries until the nineteenth century. Thereafter, the merchant fleet showed a rapid growth in the mid-decades of the 1800s and became the predominant export industry together with fisheries and forestry.17 The growth of the Norwegian merchant fleet was heavily dependent on worldwide transport of raw materials. Since the 1970s the petroleum industry has become the predominant Norwegian export industry. However, still fish and fish products and wooden based products, like pulp and paper are important Norwegian export products. Due to its dependence on foreign markets both regarding exports and imports, Norway has in the large with a few exceptions been in favor of free international trade.18 Thus, one can definitely claim that Norway always has been a small, open, raw material based economy. This would apply that the Norwegian business cycle should very much depend on the business cycles within our most important trading partners. Huge demand of raw materials should normally give both high prices and a boom in the economic activity. On the other side the abundance of raw materials would be decisive for the level of supply and thereby both the price levels and economic activity in the economy. E.g. small fish catches would normally cause fish prices to increase, when output volumes decrease. In such 13

Ibid. Ibid. 15 Stock and Watson 1998, King and Rebelo 2000. 16 Den Haan 2000, 3 – 30. 17 Brautaset 2002, 197 – 205. 18 Hanisch 1999, 17 – 20. 14

4

a case the business cycle is caused by a supply side shock, and prices and output should be negatively correlated. It is a complicated task to decide whether business cycles are demand or supply led. In this paper we focus on the statistical correlation between prices and business cycles. On the basis of these results, we can draw some conclusions on the likeliness of the business cycles to have been demand side led in the small open raw material dependent economy of Norway 1830 – 2006. Further, in order to conclude on the possible correspondence between short-term output and price movements in Norway 1830 – 2006 one should look at different sub-periods of time. This paper deals with three such sub-periods. 1830 – 1913 The first sub-period stretches from 1830 till 1913, and represents the pioneer period of modern economic growth within a liberal economic order in Norway.19 For most of this period the monetary system was fairly stable with a real silver standard 1842-1873 and a real gold standard 1874-1913. During the first years of the sub-period, i.e. 1830-1842 the central bank monitored a nominal silver standard, with a careful deflationary monetary policy in order to obtain the par silver value of the speciedaler. 1913 – 1952 The second sub-period starts in 1913 and ends in 1952. This was a very volatile period for both the international and the domestic economy, characterized by two world wars and their corresponding post-war periods, crises and growth, and inflation and deflation.20 During this period several years can be characterized as turbulent regarding the monetary situation. 1952 – 2006 Our final sub-period covers the years 1952-2006 and is characterized by a significant and growing public sector and a social-democratic economic planning regime. Domestically, these features were inspired by John Maynard Keynes and Ragnar Frisch.21 Admittedly, Norway gradually returned to a more neo-liberalistic economic world order since 1979. Nevertheless, economic planning in a mixed economy is still a dominant economic regime. The monetary policy was fairly stable with the exception of some turbulence in particular in the 1970s and 1980s.

19

Hodne and Grytten 2000, 59 – 276. Hanisch 1996, 141 – 156. 21 Søilen 1998, 417 – 446. 20

5

3. Data As indicator for the general price level we use a newly published combined cost of living index (CLI) and consumer price index (CPI) for Norway, hereafter denoted as a historical CPI.22 As for business cycles, we use newly published historical gross domestic product (GDP) figures, reflecting total output in the economy.23 The available data sets allow us to compare annual figures for all years 1830 till 2006. Prices The combined historical CLI-CPI, constructed for and published by the Norwegian central bank and quality controlled by Statistics Norway, is stretching back to 1516. It is constructed by a traditional Laspeyres approach, which is common for historical price indices.24 In fact it is a mixture of a cost of living index and a consumer price index till 1959, meaning it does not only reflect market prices, but also the costs of providing necessities for working class families. From 1959 onwards, it stands as a pure consumer price index. The series for the period in question in this paper is in fact spliced together of six different indices. The first, constructed by Ola H Grytten, covers the period 1819-1871 and includes 29 commodities within eight consumption groups 1819-1830 and 47 commodities within nine consumption groups 1830-1871, and includes most kinds of consumption less services. Almost all observations are monthly or quarterly retail or market place prices reported all over the country by governmental decree. Thus, the annual figures in the index are made up from several thousands of market price observations with high reliability.25 The key source for these data is the Professor Dr. Ingvar B. Wedervang’s Archive on Historical wages and Prices, kept at the Norwegian School of Economics and Business Administration in Bergen, Norway. The second and third cost of living indices are basically compiled from data on prices and consumption patterns in the Oslo area. The first of these, covering the years from 1871 till 1901, was constructed by Jan Ramstad. It was made up of an even richer price material reported by and for public servants for 55 representative commodities, recorded in the Wedervang Archive. This index is very well documented and proved reliable.26 Thereafter, we use the cost of living index from the Statistical Office of Kristiania (Oslo) for the period 22

Grytten 2004a, 47 – 98. Grytten 2004b, 241 – 288. 24 Grytten 2004a, 47 – 98. 25 Sircular, 4th Royal Norwegian Ministry, January 20th 1816 and Wedervang Archive, file 272. 26 Ramstad 1982, 471 – 493. 23

6

1901-1916, including about 70 items.27 Again all major consumption groups are included in these monthly figures, except for services, which are lacking till 1900, and thereafter are underrepresented. The fourth index series was constructed by the Ministry of Social Affairs during a short period stretching from 1916 till 1919. It covers 16 of the major urban areas of Norway at the time and includes 60 retail commodities, fuel inclusive. The observations were taken on a monthly basis.28 From 1919 Statistics Norway took over as the major provider of cost of living indices. They conducted several consumption surveys, covering up to 31 urban areas and collected retail price data on some hundred consumption items, i.e. from 120 in 1920 to 700 in 1959. Data were compiled all over the country in order to construct this index. Finally, in 1959/1960 the cost of living index for working class households was fully replaced by a consumer price index (CPI) representing all kinds of households and products, which can be bought in retail shops at market prices. The range of data compilation has been increased gradually. Today Statistics Norway collects retail prices of a set of about 1000 representative commodities from all over the country.29 During the last years they have monitored a CPI constructed as a geometrical series. However, in our historical series we still use an arithmetic approach. GDP In order to map the business cycles we use gross domestic product per capita in fixed prices, which expresses the total volumes of output or value added in the economy. The historical national accounts for Norway stretch back to 1830, and are calculated in several steps. In 1965 Statistics Norway published GDP per capita for the years 1865-1960.30. The calculations were carried out from the production and the expenditure side based on available data kept or published by the bureau. Thereafter, we use new calculations made of GDP back to 1830 by a group of scholars from the Norwegian School of Economics and Business Administration in Bergen. They are well documented in publications from the central bank.31 Again, an important source, both of volumes and prices, is the Wedervang Archive. Also here, the calculations are made both 27

NOS 1978, 518 – 519. NOS 1994, 289 – 292. 29 NOS 1994, 289 – 292. 30 NOS 1965, 64 – 371. 31 Grytten 2004b, 272 – 289. 28

7

from the production and the expenditure side, and to reach at estimates in fixed prices we have in principle used a double deflation technique. On the basis of both input and output data from the main industries within both the public and private sector and annual price observations it has been possible to calculate relatively trustworthy series of value added in the Norwegian economy for the period. Central sources are records from locally and centrally located civil servants published by ministries and other public bodies connected to the government along with private archives and records and reports given by senior county officials. In particular the foreign trade statistics and industrial and population censuses have been of great importance. Thirdly, we use the newly revised contemporary GDP-figures for 1970-2006 calculated and maintained by the national account department at Statistics Norway, considered some of the most precise national accounts in the world. 32 Finally, the historical GDP-calculations for 1830-1865, 1865-1950 and the revised contemporary figures 1970-2006 are spliced together. These series were first published by the central bank of Norway as part of a project on historical monetary statistics.33 They were also quality controlled by a research network on the construction of standardized and harmonized Nordic historical national accounts. Together with the price data they should constitute valid and reliable sources for examination of the correspondence between price movements and business cycles in Norway 1830-2006. Figure 1 shows ln CPI (henceforth CPI) and figure 2 shows ln real GDP per capita (henceforth GDP) 1830 – 2006 both in levels and first differences. The figures show large variations from year to year for both prices and output. The output trend is ascending the entire time span, while a trend growth in prices appears after WWI.

32 33

www.ssb.no/emner/09/01/nr/index.html Eitrheim 2004, Eitrheim 2007. 8

0.4

9 CPI (right) D.CPI (left) 8.5

0.3 8

0.2

7.5

7 0.1 6.5

0

6

5.5 -0.1 5

-0.2

4.5 1840

1860

1880

1900

1920

1940

1960

1980

2000

Figure 1: ln CPI in levels and first differences 1830 – 2006. Source: Grytten 2004a, 92 – 93. 0.15

13 GDP (right) D.GDP (left) 12.5

0.1

12

0.05 11.5

0

11

10.5 -0.05

10

-0.1 9.5

-0.15

9 1840

1860

1880

1900

1920

1940

1960

1980

Figure 2: ln real GDP per capita in levels and first differences 1830 – 2006. Source: Grytten 2004b, 285. 9

2000

4. Annual and business cycle fluctuations The new data for the GDP and CPI series enable us to look at the correspondence between annual changes in output and prices as plotted in figure 3. It is surprisingly difficult to spot any significant persistent correlation between the two, in particular not a positive relationship. Thus, we cannot trace any clear tendency by just looking at these graphs. Another problem with looking at annual fluctuations is that these changes are not measurements of business cycles, due to their different time span. A business cycle is always of some length of time, which is more than one year, traditionally more often four to nine years. Hence, the next step in our analysis will be to use filters to extract the business cycle component of the time series. 0.4 D.CPI D.GDP

0.3

0.2

0.1

0

-0.1

-0.2 1840

1860

1880

1900

1920

1940

1960

1980

2000

Figure 3: Annual fluctuations in CPI and GDP for Norway 1830 – 2006. Sources: Grytten 2004a, 92 – 93. Grytten 2004b, 285. Separating trend and cycle It is common to think of a macroeconomic time series, for example the GDP, as something that grows along a trend. This trend growth represents the mechanisms of economic growth, e.g. productivity changes, and which has a permanent effect on the economy. But, in reality the economy will deviate from the long-run trend because of positive and negative shocks to

10

the economy.34 To study business cycles one must find a way to break down a time series into these two components. The technique used to decompose a time series into a trend and a cyclical component is called filtering. Generally and formally, a filter separates an observed time series, { yt }Tt =1 , into a smoothed or a trend component, g t , and a cyclical component, ct : yt = g t + ct .

(1)

One filter that is very much used in applied work is the Hodrick-Prescott (HP) filter35, where the trend component g t is, for a given sample size T, determined by T

T −1

min ∑ ( yt − gt ) 2 + λ ∑ [( g t +1 − g t ) − ( gt − g t −1 )]2 . t =1

(2)

t =2

To estimate the trend the HP filter tries to minimize two terms. The first term is the deviation of the observed value of GDP and the trend. That is, we want to minimize the cyclical component yt − g t = ct . The term in square brackets in equation (2) represents the approximate change in estimated trend between two time periods given that we use the logarithm of the variable in question. The relative weight of these two terms is given by the parameter λ which is also called the smoothing parameter.36 The larger the value of the λ the smoother the estimated trend growth, that is, λ approaching infinity implies that the estimated trend will approximate a straight line.37 Although the filter has been influential and is used extensively in the economics and finance literature it has some drawbacks. First, HP filter is known for “unusual behavior of cyclical components near the end of the sample.”38 Second, it is not straightforward to choose the size of the smoothing parameter lambda (λ). In this paper we apply a value of 100 for the λ, which is a ‘normal’ value for annual data. Third, the filter “can generate business cycle dynamics even if none are present in the original data.”39 To overcome these drawbacks several other filter techniques have been proposed in the literature. In this paper we apply two other filters in addition to the HP filter. But, we start by reviewing a few concepts from time series analysis.

34

Nelson 2008. Hodrick and Prescott, 1997. 36 Sørensen and Whitta-Jacobsen, 2005 p. 405. 37 Ibid. 38 Baxter and King 1999, p. 576. 39 Cogley and Nason 1995, p. 253. 35

11

When analyzing the correlation properties between different points in time of a time series one is performing what is called a time domain analysis. But, it is also possible to analyze a time series in the frequency domain (spectral analysis) and its objective is to determine which cyclical frequencies are important in determining the behavior (variance) of the time series. Let us start by studying a time series Yt with a periodic sinusoidal component which can be modeled as Yt = R cos(ωt + φ ) + Z t

(3)

where ω is the frequency, R is the amplitude, φ the phase and {Rt } is a stationary random series.40 The frequencies are measured in radians where π is equal to 180°. Hence, 2π represents a full cycle. The simple model in equation (3) can be extended by the fact that the variation in an observed time series can be explained by several different frequencies. In sales figures there may for example be weekly, monthly and yearly cyclical variations.41 Then we may write the model with several frequencies as k

Yt = ∑ R j cos(ω j t + φ j ) + Z t

(4)

j =1

where R j , φ j are the amplitude and phase at the frequency ω j . Using the fact that

cos(ωt + φ ) = cos ωt cos φ − sin ωt sin φ the previous model can be written as k

Yt = ∑ ( a j cos ω j t + b j sin ω j t ) + Z t

(5)

j =1

where a j = R j cos φ j and b j = − R j sin φ j .42 This states that the time series Yt can be approximated by a weighted sum of cosines and sines of different frequencies.43 Generalizing this by letting k approach infinity, that is, employing an infinite number of frequencies in explaining the time series, it is possible to find the spectral density function f (ω ) (also called the spectrum).

40

Chatfield 2004, p. 107. Ibid. p. 108. 42 Ibid. p. 109. 43 Hamilton 1994, p. 152. 41

12

As stated above, we are interested in finding the frequencies that are important in determining the behavior of the time series. Luckily, the theory of spectral analysis also states that it is possible to extract these components using what is called the ‘ideal band pass filter.’ The filter transforms the data linearly by keeping “intact the components of the data within a specific band of frequencies and eliminates all other components.”48 Other ideal filters involve the low-pass filter which keep intact the low frequencies of the spectrum and the high-pass filter which wipes out the low frequencies of the spectrum. In practice, however, it is not possible to compute the ‘ideal band pass filter’ since it requires an infinite amount of data. Some sort of approximation is required and in this paper we apply two band pass (BP) filters. The first one is the Baxter-King (BK) and the second one is the Christiano-Fitzgerald (CF) band pass filter.49 The main difference between these two filters is how they approximate the ‘ideal band pass filter.’ In general, BP filters have become popular in recent years as an alternative to the HP filter. “BP are appealing because they make the notion of business cycle operational by selecting fluctuations in the pre-specified range.”50 Low frequencies imply longer cycles in the data while higher frequencies imply shorter cycles. Which frequencies that should be allowed to pass through the filter depend on the a priori expectations on the duration of a typical business cycle. Hence, the band pass filter “passes periodic components that lie within a prespecified frequency band and eliminate everything else.”51 According to the business cycle theory pioneer Joseph Kitchin the economy moves in inventory cycles of three to five years.52 According to another pioneer in the field, Clement Juglar, investment cycles of seven to eleven years are quite common.53 These seem basically to be demand driven. Business cycles have tended to become shorter during the modernization of the economy. In applying the BK and the CF band pass filters, we have chosen periodic components between two and seven years.

48

Christiano and Fitzgerald 2003, p. 436. Baxter and King, 1999. Christiano and Fitzgerald 2003. 50 Canova 2007, p. 94. 51 Cogley 2008. 52 Kitchin 1923. 53 Juglar 1916. 49

13

0.15 CPI (BK cycle) GDP (BK cycle)

0.1

0.05

0

-0.05

-0.1 1840

1860

1880

1900

1920

1940

1960

1980

2000

Figure 4: CPI and GDP cycles 1830 – 2006 estimated by the Baxter-King band pass filter allowing periodic components between two and seven years. Sources: Grytten 2004a, 92 – 93. Grytten 2004b, 285. In figure 4 we plot the estimated cycles from the BK filter of the time series. Still it is difficult to trace any consistent pattern in the co-movement between GDP and CPI. This could be explained by time lags or by the fact that business cycles last for several years. Thus, there is not necessarily strong contemporaneous correlation between the two variables, but there may nevertheless be correlations within the typical cycle period.

Correlations We now compute the cross correlations between the cyclical components of the GDP and the CPI in order look at the strength and direction of their (linear) relationship. That is, whether, the CPI is pro-cyclical or counter-cyclical. In addition we compute the correlation between the two when we allow that the CPI is a leading or lagging variable of the GDP by one or two time periods. The calculations are done for all sub-periods plus the entire time span.

14

Table 1: Correlation coefficients between GDP and CPI for Norway 1830 – 2006. Lag

1830 – 1913

1913 – 1952

1952 – 2006

1830 – 2006

HP

BK

CF

HP

BK

CF

HP

BK

CF

HP

BK

CF

-2

.311

.146

.223

-.124

-.042

-.135

.182

.182

.254

-.005

.006

.014

-1

.328

.285

.301

-.250

-.010

.103

-.063

.029

-.076

-.111

.078

.145

0

.052

-.029

-.209

-.351

-.151

.079

-.327

-.421

-.539

-.271

-.118

-.083

1

-.056

-.157

-.183

-.312

.085

.145

-.302

-.302

-.379

-.281

-.067

-.015

2

-.081

-.055

.096

-.242

.059

-.005

-.062

.103

.233

-.237

-.005

.068

The numbers show the correlation between the cyclical components of current GDP and current, leads [-] and lags of CPI. Lag equal to zero (middle row with numbers) indicates contemporaneous correlation. Numbers in cursive denotes significant coefficient at least 10% significance level. Cycles are estimated with the HP, BK, and CP filters.

According to the results presented in table 1 about 58% (35 out of 60) of the estimated correlations are negative. There is significant contemporaneous negative correlation. In other words, CPI tends to be counter-cyclical. We also see that there is a significant strong negative correlation between current GDP and lagged values of the CPI. This holds in most cases regardless whether we apply the Hodrick-Prescott, Baxter-King or the Christiano-Fitzgerald filter. For the whole sample, we observe that for the contemporaneous and lagged correlation coefficients the HP filter generates correlations coefficients that are several times the magnitude of the correlations based on the BK and the CF filter. The contemporaneous and lagged correlation coefficients for 1830 – 2006 are all statistically significant when the data are filtered with the HP filter. During the first sub-period 1830 – 1913, we have in fact none significant negative coefficients for the HP and BK series. The CF series gives statistically significant negative correlation coefficients for the correlation between current GDP and current and the one period lagged CPI. The HP series on the other hand, indicates a statistically positive relationship between current GDP and leading CPI (one and two periods). That is, prices moves ahead of output. This positive relationship on the aggregate national level is found despite a clear negative relationship between output and prices from agriculture, as shown in figure 5.54 The negative correlation coefficient in the agricultural sector (-.40), clearly indicates that supply shocks

54

Grytten 2004c, 47 – 76 and Grytten 2004a, 47 – 98. 15

were decisive for the annual development of output and prices in this important industry, while the opposite seems to have been the case for the economy in general for the HP series. 30 Prices (HP cycle) Volumes (HP cycle) 25

20

15

10

5

0

-5

-10

-15

-20 1830

1840

1850

1860

1870

1880

1890

1900

1910

Figure 5: HP cycles (λ = 100) for prices and volumes in agriculture in Norway 1830 – 1910. Sources: Grytten 2004a, 90 – 91. Grytten 2004c, 47 – 76. Ramstad 1982, 493. For the second sub-period 1913 – 1952, the HP series gives only negative correlations, while the BK and CF series are more mixed. However, only the HP series gives statistically negative correlations for contemporaneous and one period lag. All the other coefficients are not statistically significant. This was a period with several shocks to the economy.55 In the first place two world wars took place during these years. Inflation grew rapidly during the war years when output contracted. Secondly, two huge international depressions hit the economy devastatingly. Thirdly, Norway ran a strong deflationary monetary policy for most of the 1920s, causing a sharp and long period of deflation in the 1920s in addition to the international deflation in the 1930s. Fourthly, during the post WWII period till 1952, the government widely subsidized and directed the economy in order to prevent high inflation and thereafter a postwar depression.56 Between these years of crises and abnormal economic

55 56

Klovland 1998, 309 – 344. Hodne and Grytten 2002, 77 – 196. 16

policy, significant growth took place. From 1913 till 1952 the recorded per capita GDP growth rate was impressively 2.24 per cent.57 This implies that the years 1913 – 1952 constitute a period of substantial economic growth both when compared to Norwegian historical growth rates and growth rates of other countries during the same period. However, due to the two great wars, their aftermaths and the long period of deflationary monetary policy, one can easily see that the economy in many years must have been influenced by heavy supply side shocks and that prices could not necessarily have mirrored the business cycles. For the third sub-period 1952 – 2006, all the coefficients that are statistically significant are negative for the contemporaneous and one period lag. This holds for all three series, and the size of the correlations are in the interval -.30 to -.54. Price leads are for most part positive, but they are with one exception not statistically significant. Despite the dominant Keynesian paradigm in this period, we chiefly find negative relationships between prices and output.58 There was not any significant primary sector to impose huge supply side shocks to the Norwegian economy after 1952. Still prices and business cycles were negatively correlated. This could partly be explained by the adoption of technology from the US, causing productivity and prices to increase and the economy to boom. During the 1970s, the economies of the OECD countries were heavily influenced by the considerable jumps of petroleum prices caused by OPEC I in 1973 (petroleum embargo), and OPEC II in 1979 (Iraqi-Iranian war). Oil prices per barrel stepped up from three to forty dollars during the 1970s. In consequence, the OECD area experienced high inflation and fall in demand for domestically produced goods. Thus, changes in prices and output moved in opposite directions. As for the entire period 1830 – 2006, our calculations give negative correlations for the most part. However, only the HP series gives negative correlations that are statistically significant for contemporaneous and lagged correlation coefficients. Summing up, on the evidence presented in table 1 we conclude that for the historical period 1830 – 2006 it is more common with negative correspondence between prices and output than positive.

57 58

NOS 1965, 348 – 351. Hanisch 1999, 17 – 28. 17

5. The foreign sector In order to get a better understanding of the observed phenomena in the small open raw material based economy of Norway it is of interest to see if the development has been influenced by the foreign sector. Thus, in this section of the paper we take a closer look at the correspondence between export and import prices versus output from the economy. The export and import prices used are59 taken from the historical national accounts. We use the implicit price indices. Sources for the export and import price series in the historical national accounts are public records and the Wedervang Archive. Figure 6 shows the Baxter-King cycles for the GDP and the implicit price indices for export and import prices. 0.3 GDP (BK cycle) Export prices (BK cycle) Import prices (BK cycle)

0.2

0.1

0

-0.1

-0.2

-0.3 1840

1860

1880

1900

1920

1940

1960

1980

2000

Figure 6: GDP and implicit price indices for export and import prices cycles 1830 – 2006 estimated by the Baxter-King band pass filter allowing periodic components between two and seven years. Sources: Grytten 2004b, 285. Wedervang Archive, files W272, W276 and W383. In the first place the graphs reveal huge volatility in prices on exports and imports. Thus, these might have a substantial influence on the trade balance and the economy. Huge price fluctuations are more likely to cause and reflect shocks to the economy than smaller price fluctuations. 59

Wedervang Archive, files W272, W276 and W383. 18

Nevertheless, it is not easy to read from the graphs if there is a general trend, negative or positive, between export and import prices on the one hand and output on the other. Thus, we will have to calculate correlation coefficients and their significance. An important aspect of this examination is to find out how external shocks have influenced Norwegian output through import or export prices. Additionally, this examination can inform us on how other supply side shocks, e.g. low or high fish catches, output from forestry or oil and gas relative to demand has influenced the relationships between prices and output. We conduct this examination with further calculations of correlation coefficients. When simultaneous and price-lagged series are the most relevant when examining a possible Keynesian relationship between output and prices, simultaneous and price-led series would be the most relevant in this analysis. The idea is to examine if external supply side shocks can explain fluctuations in output.

Exports First we present correlation calculations for export prices versus GDP per capita in table 2. As in the previous section we apply the HP, BK and CF filter and look at the cyclical components. Our output data are taken from the historical GDP series for Norway, where the price series is the implicit deflator for exports. Both the output figures and the price observations for the foreign sector are among most valid and reliable data in the Norwegian historical national accounts. Thus, these series clearly should be considered trustworthy.60

60

Grytten 2004b, 281 – 283. 19

Table 2: Correlation coefficients between GDP and implicit price deflator for Norwegian exports 1830 – 2006. Lag

1830 – 1913

1913 – 1952

1952 – 2006

1830 – 2006

HP

BK

CF

HP

BK

CF

HP

BK

CF

HP

BK

CF

-2

.094

-.075

-.110

-.147

-.351

-.373

-.034

-.237

-.172

-.039

-.231

-.224

-1

.206

.049

-.025

-.137

.126

.264

-.160

-.228

-.139

-.062

.057

.084

0

.182

-.038

-.116

-.204

.197

.383

-.124

.111

.126

-.129

.104

.147

1

.182

.173

.118

-.359

-.251

-.212

-.100

.093

.107

-.238

-.138

-.097

2

.073

.155

.172

-.277

-.123

-.277

-.085

-.128

-.115

-.218

-.118

-.115

The numbers show the correlation between the cyclical components of current GDP and current, leads [-] and lags of the implicit price deflator for Norwegian exports. Lag equal to zero (middle row with numbers) indicates contemporaneous correlation. Numbers in cursive denotes significant coefficient at least 10% significance level. Cycles are estimated with the HP, BK, and CP filters.

According to the estimates reported in Table 2 there was a positive correlation between export prices and output 1830 – 1913. The HP series, which is the only series giving correlation coefficients that are statistically significant. The correlation is significant for the contemporaneous and one period price lead and lag series. This indicates that positive shifts in export prices were mirrored in positive shifts in output during this sub-period. Thus, the export sector of the supply side contributes to explain the positive co-movements of output and prices in this period. This is fairly understandable for a small open raw material based economy. As for the turbulent sub-period 1913 – 1952 we find for most cases negative correlations. The estimates for the price lags give the strongest results. These results are thus in line with those reported in table 1, suggesting considerable negative counter-cyclical movements for general prices versus output 1913 – 1952. Also for the third sub-period we find for most cases negative correlations. However, all but one of these are statistically insignificant. For the entire time span 1830 – 2006, there is again for most cases negative correlations. For the HP series the contemporaneous and lagged correlations are statistically significant. Looking at the period as a whole there is a clear negative co-movement between export prices and output from the economy. This is very much in line with the pattern for the total economy. Hence, supply side shocks from the export industries seem to have had an effect on 20

the economy. Supply side shifts from the export sector which have made production costs fall have caused input volumes increase.

Imports The estimated co-movements of import prices and output are reported in table 3. Again, we find mostly positive correlations for the first sub-period 1830 – 1913. For the HP series, the statistically significance pattern is exactly as for the export prices (cfr. table 2). The positive correlations indicate that negative supply side shifts from the import side had a negative effect on output. This can in a broader perspective be explained by international price movements, i.e., when international prices fell Norwegian output fell. In this way Norwegian output in the nineteenth century may mirror the international business cycles.61 This again confirms our findings that prices and output basically moved in the same direction 1830 – 1913.

Table 3: Correlation coefficients between current GDP and implicit price deflator for Norwegian imports 1830 – 2006. Lag

1830 – 1913

1913 – 1952

1952 – 2006

1830 – 2006

HP

BK

CF

HP

BK

CF

HP

BK

CF

HP

BK

CF

-2

.102

.014

-.018

-.056

-.082

-.029

.101

-.074

-.033

.008

-.043

-.026

-1

.269

.003

.018

-.117

.401

.577

-.038

-.125

-.145

-.027

.227

.303

0

.414

.229

.205

-.462

-.245

-.062

-.073

.037

-.011

-.228

-.088

.022

1

.327

.192

.134

-.559

-.363

-.363

-.110

-.029

-.052

-.323

-.209

-.169

2

.035

.044

-.024

-.330

.058

-.037

-.067

-.048

-.039

-.255

.009

-.030

The numbers show the correlation between the cyclical components of current GDP and current, leads [-] and lags of the implicit price deflator for Norwegian imports. Lag equal to zero (middle row with numbers) indicates contemporaneous correlation. Numbers in cursive denotes significant coefficient at least 10% significance level. Cycles are estimated with the HP, BK, and CP filters.

For the next sub-period 1913 – 1952, we find mostly negative correlations. Looking at the coefficients that are statistically significant, most of them are for the contemporaneous and lagged values. Especially this is true for the HP series. This indicates that negative price shocks on imported goods gave fuel to domestic output. 61

Hodne 1983, 262 – 269. 21

As for the years 1952 – 2006, we again observe a negative co-movement for all but two estimates. However, none of the estimates for this period is statistically significant. Thus the relationship between import prices and output from the Norwegian economy in this period is unclear. Finally, looking at the full time span 1830 – 2006 and the HP series, we find mostly negative correlations, and most of them are statistically significant. The correlations for the BK and CF series do not show any clear pattern. But, the overall picture from table 3 is that import prices and output from the Norwegian economy move in different directions. The lagged crosscorrelations for the HP series are stronger than the contemporaneous correlations. To sum up this section, both export and import prices indicate that supply side shocks are decisive in order to explain the counter-cyclical movements between prices and output in Norway from WWI until present. During the years prior to the war, i.e., 1830 – 1913, we do not find this counter-cyclical tendency and the effect of external supply shocks on short-term output is not clear for this period.

6. Conclusions This paper offers an empirical investigation of the short-term correspondence of historical output versus prices 1830 – 2006 for the small open raw material based economy of Norway. The examination is done by applying three different filters (HP, BK and CF). Except for the times before WWI, where there is a weak, but basically positive significant correlation between the two sets of variables, we find chiefly significant negative correlations. Thus, prices do not seem to reflect short-term movements in the economy. This is contrary to the typical Keynesian view, i.e., demand is decisive for the short-term economic performance. In consequence, prices should mirror short-term swings in the economy. However, they did not for Norway during most of the twentieth century until present days. On the contrary, shortterm prices and output more often move in opposite directions. On the basis of the chiefly counter-cyclical historical movements between general output and prices and similar results for export and import prices versus output, we conclude that domestic and in particular external supply side shocks offer important information in order to understand historical short-term output movements. We have tested for the importance of the external sector for the negative relationship between observed prices and output. The results indicate that supply side shocks effecting both exports 22

and imports are decisive in order to explain the counter-cyclical movements between prices and output in Norway from the World War I until present times. During the years prior to the war, i.e., before 1914, we do not find a counter-cyclical tendency between foreign trade prices and output. Hence, the effect of external supply shocks on short-term output is not clear for this period. However, rather positive than negative.

23

7. References Baxter, Marianne and Robert G. King 1999, “Measuring Business Cycles: Approximate Band-Pass Filters for Economic Time Series”, The Review of Economics and Statistics, Vol. 81, no. 4, 575--593. Bergh, Trond et al 1983, Norway from an underdeveloped to an industrial economy, Oslo: Gyldendal. Bjørnland, Hilde Christiane 2000, “Detrending methods and stylized facts of business cycles in Norway – an international comparision”, Empirical Economics, Vol. 25, 369 – 392. Brautaset, Camilla 2002, Norwegian Exports 1830-1865: in perspective of historical national accounts, Bergen: NHH. Canova, Fabio 2007, Methods for Applied Macroeconomic Research, Princeton University Press. Chatfield, Chris 2004, The Analysis of Time Series. An Introduction, Chapman & Hall/CRC. Christiano, Lawrence J. and Terry J. Fitzgerald 2003, ”The Band Pass Filter”, International Economic Review, Vol. 44, No. 2, 435--465. Cogley, Timothy and James N. Nason 1995. “Effects of the Hodrick-Prescott filter on trend and difference stationary time series. Implications for business cycle research.”, Journal of Economic Dynamics and Control, Vol. 19, 253 – 278. Cogley, Timothy 2008. ”Data filters.” The New Palgrave Dictionary of Economics. Second Edition. Eds. Steven N. Durlauf and Lawrence E. Blume. Palgrave Macmillan. Cooley, Thomas F. and Lee E. Ohanian 1991. “The cyclical behavior or prices”, Journal of Monetary Economics, Vol. 28, 25 – 60. Eitrheim, Øyvind et al 2004, Historical Monetary Statistics for Norway 1819-2003, Oslo: Norges Bank. Eitrheim, Øyvind et al 2007. Historical Monetary Statistics for Norway, part 2, Oslo: Norges Bank. Furre, Berge 1991, Vårt hundreår: Norsk historie 1905--1990, Oslo: Samlaget.

24

Grytten, Ola Honningdal 2004a, ”A consumer price index for Norway 1516-2003”, Eitrheim, Øyvind et al (eds), Historical Monetary Statistics for Norway 1819-2003, Oslo: Norges Bank, 47-98. Grytten, Ola Honningdal 2004b, ”The gross domestic product for Norway 1830-2003”, Eitrheim, Øyvind et al (eds), Historical Monetary Statistics for Norway 1819-2003, Oslo: Norges Bank, 241-288. Grytten, Ola Honningdal 2004c, ”A Norwegian consumer price index 1819-1913 in a Scandinavian perspective”, European Review of Economic History, 8, 47-76. Haan, Wouter J. den 2000, ”The comovement between output and prices”, Journal of Monetary Economics, 46, 3-30. Haan, Wouter J. den and Steven W. Sumner 2004. ”The comovement between real activity and prices in the G7”, European Economic Review, 48, 1333 – 1347. Hamilton, James D. 1994, Time Series Analysis, Princeton University Press. Hanisch, Tore Jørgen 1996, Om valget av det gode samfunn, Kristiansand: Høyskoleforlaget. Hanisch, Tore Jørgen et al 1999, Norsk økonomisk politikk i det 20. århundre, Kristiansand: Høyskoleforlaget. Hodne, Fritz 1983, The Norwegian Economy 1920--1980, London: Croom-Helm. Hodne, Fritz and Ola Honningdal Grytten 2000, Norsk økonomi i det 19. århundre, Oslo: Fagbokforlaget. Hodne, Fritz and Ola Honningdal Grytten 2002, Norsk økonomi i det 20. århundre, Oslo: Fagbokforlaget. Hodne; Fritz 1983, The Norwegian Economy 1920-1980, London: Croom-Helm. Hodrick, R. and E. Prescott (1997): “Post-War Business Cycles: An Empirical Investigation.” Journal of Money, Credit and Banking, Vol. 29, no. 1, 1 – 16. Husebø, Tore Anders and Bjørn-Roger Wilhelmsen 2005, ”Norwegian Business Cycles 19822003”, Staff Memo, 2/2005, Economics Department, Norges Bank, 1-23.

25

Juglar, Clement 1916, A Brief History of Panics and Their Periodical Occurrence in the United States, 3. ed, Gutenberg project. Keynes, John Maynard 1936, The general theory of employment, interest and money, Cambridge: Macmillan. King, R. and S.T. Rebelo 2000, ”Resucitating real business cycles”, Working Paper, 7534, NBER. Kitchin, Joseph 1923, ”Cycles and trends in economic factors”, Review of Economics and Statistics, 5, 10-16. Klovland, Jan Tore, 1998, ”Monetary policy and business cycles in the interwar years: the Scandinavian Experience”, European Review of Economic History, 2, 309-344. Kondratiev, Nikolai 1926, ”Die langen Wellen der Konjunktur”, Archiv für Sozialwissenschaft und Sozialpolitik, 56: 573-609. Nelson, Charles R. “trend/cycle decomposition.” The New Palgrave Dictionary of Economics. Second Edition. Eds. Steven N. Durlauf and Lawrence E. Blume. Palgrave Macmillan, 2008. The New Palgrave Dictionary of Economics Online. Palgrave Macmillan. 07 April 2010 doi:10.1057/ 9780230226203.1739 NOS C. 188 1994, Historical Statistics 1994, Oslo: Statistics Norway. NOS XII. 291 1978, Historical Statistics 1978, Oslo: Statistics Norway. NOS. 163 1965, National Accounts 1865-1960, Oslo: Statistics Norway. Pakko, Michael R. 2000. “The Cyclical Relationship between Output and Prices: An Analysis in the Frequency Domain,” Journal of Money, Credit, and Banking, Vol. 32, No. 3, 382 – 399. Parker, Randall E. 2005. “A Historical Examination of the Comovement between Output and Prices.” Unpublished manuscript East Carolina University. Ramstad, Jan 1982, Kvinnelønn og pengeøkonomi, Bergen: NHH.

26

Schumpeter, Joseph 1939, Business cycles: a theoretical, historical and statistical analysis of the Capitalist process, New York: McGraw-Hill. Sejersted, Francis 1973, Historisk introduksjon til økonomien, Oslo: Cappelen. Sircular, 4th Royal Norwegian Ministry, January 20th 1816. Smith, R. Todd 1992. “The Cyclical Behavior of Prices”, Journal of Money, Credit, and Banking, Vol. 24, No. 4, 413 – 430. Stock, J.H and M.W. Watson 1998, ”Business cycle fluctuations in US macroeconomic time series”, Working Paper, 6528, NBER. Søilen, Espen 1998, From Frischianism to Keynesianism, Bergen: NHH. Sørensen, Petter Birch and Hans Jørgen Whitta-Jacobsen, Introductin Advanced Macroeconomics: Growth and Business Cycles. McGraw-Hill. Wedervang Archive, files W272, W276 and W383. www.ssb.no/emner/09/01/nr/index.html

27