238 PREDICTION OF PULP PRICES - 1 REVIEW TWO YEARS AFTER Jérgen Randers Norwegian School of Management, ‘BekKestua, (Oslo), Norway apsreact During the sumer of 1962 the author made predictions of wood pulp prices for the period 1982 to 1986. The predictions were part of 2 decision on wh ‘pulping plant in Norway. the predictions’ and the sani on which they were made. Next, the predictions are compared with actual Gata for the period 1962 to 1984, ‘The price predictions were based on the ansunption that the inter- national market for wood pulp is characterized by inventory gycle theory. According to thie theory price oscillates with © 36 year wavelength primarily caused by the “production rate —-> inventory => production rate" loop: ‘Thhe paper reviews the existing Literature on inventory cycle and extends the theoretical work to cover inventory oscillations © the industry level in some detail The resulting theory turns out to be supported by tine series of quarterly data for international pulp production, pulp inventories and pulp prices over the 25 years prior to 1982- ‘TThe time series for pulp price, once corrected for inflation, denonstrate a clear S-year fluctuation around an exponentially @eclining trend. The declining trend e probably due to technical advance in the production of wood pulp.” The S-year cycle i= avan ore ‘conspicuous in inventory seatiatice. ‘The price predictions were made by ining: 4) continued technical advance - i.e. a continuation Of the exponentially declining trend in prices 44) continued inventory fluctuation - 1.e- a continuation Of the 5-year oscillation in prices around the trend. Comparison with actual date denonstrates that thie (simple) method Of prediction ylelde good results. Interestingly, the ‘unavoidable noise (due to random, unpredictable events in the market) appears too weak to hide the systematic festures in the Sevelopment of prices on which this method depend. PREDICTION OF PULP PRICES - 1 REVIEW THO YEARS APTER conresers BerRopueTToN Inventory Cycle Theory Purpose of this Paper ‘RIEORETICNL EXTENSION Baste Hechanions Elaboration MEASUREMENT OF PAST DEVELOPMENTS Procedure Phecuseton PREDICTION OF FUTURE DEVELOPMENTS Procedure PLecuasion COMPARISON OF PREDICTED AND ACTUAL. DEVELOPMENT. Reaulte Discussion ‘sowcLUBTON239 PREDICTION OF WOOD PULP PRICES - ‘A REVIEW TWO YEARS APTER wrRoDUCTION INVENTORY CYCLE THEORY Inventory cycles 1s an old topic in the systen dynamice Atterature, Periodical fluctuations in production, inventories and price are frequently observed in the economy, and have attracted the interest of system dynaniciats since Forrester's original Work on 3 to 6 year long cycles at the company level (Forrester, 1961). cycles of a 2 to 6 year wavelength can be observed both at the company, industry and economy level. In order to observe the cycles, one does need a longer time horizon than is common in economic affairs. To observe a pattern of waves where each wave te 3 to 6 years one must have knowledge of, or time series for, & period of at 1 insufficient, - because they only give 3 to'6 observations per wave. t 10 to 15 years. Further, yearly data is Quarterly data, att least, are neces FY. Because of these Feasons, inventory cycles are hard to observe and hence much less Fecognized than should be expected, given their prevalence. Sti11, inventory cycles have been given extensive treatment in the system dynamics Literature. common to these treatments ie the belief that inventory cycles are endogenously generated; they ‘Fe not seen as the result of some exogenous rhytmic driving force (2tke sun opots, vars, or pariianentary decisions). In hie original work in Industriel pynanics (Forrester, 1961) demonstrated how @ecteion rules (policies) in the company cause regular fluctuations An the level of activity with a wavelength of some years. me Ante: ting point ie that fluctuations arise fron policies that are + Forrester shove that. ‘common to most manufacturing buain« ‘the common tendency to £111 orders in proportion to the sire of ‘the order backlog - or equivalently to increase the utilization Of installed capacity when there 1s = amall inventory of finished goods - te sufficient to generate cyclicality vhen the company operates in an environment characterized by noise (e.g. random variations in denand, delivery delays, input prices or the like). The central role of order bakcloge - or product inventories - in this process is the basis for the name “inventory cycle ‘The theme of inventory cycles wae picked up by Meadows (weagovs, 1970) at the industry level in hie study of the U.8. Pork industry. Of course the cyclicality of the hog industry had “ ‘and commonly been used as an example of the cobweb theorem in been known for deca (Beekiel. 1938) for the clasete expositon) textbook micro-economics. However, Meadove did describe the detailed decision process of the farmer. He demonstrated how "rational* decisions by individual farmers concerning breeding ‘and slaughtering of pigs, ad4 up to strong and unintended fluctuations An the industry, with periods of boom and depression following each other in endless succession. ng Meadows’ work was extended by NaL1I et al. (wai11,1973) who made.3 240 the t1es to the real world of the hog industry more obvious by by incorporating exogeno enand for pork. effects - Like seasonality in the final 1A weakness of all studies this far vas a lack of statistical evidence of the inventory cycle. The studi Af any, time series iilustrating cycles in past developments. Hilgh quality tine series were not, published until 1979 (Rgeteland, 1979)1 not for the hog thdustry, but for the Norwegian suiphite were based on scant, Pulp industry. Hgateland gave a lucid illustration of the reality of inventory cycles (see figure 1 a) shoving waves in production, ant Anventory and price over a 20 year period. me variables éid Anteract in a pattern that was are slovly reduced when inventories are large. Production dose respond to prices, and ultimately production rates are reduced below sais + Inventories start to contract. Production is not increased again, hovever, lunti1i inventories fa11 so low that prices pushed upwards by worried customers - signal a new period of and boom in the industry. Ngateland stressed - ae did Majani and Makridakis (Hajant and Makridakis, 1977) in an independent analysis outelde the yyeten Aymanics tradition - the need to smooth the weekly or monthly data ‘that ie commonly used in the industry. Holse in the data obscures cyclical trends in production and sales and my explain the very slow sdjustment tovards market equilibrium. Production and salet statietion are so olay that industry decision makers hardly can be expected to discover new market trends untill mich time has passed. At the sane time, Randers (Randers, 1978) published several other indications of the 3 to 6 year cycle. Me distinguished the short inventory cycle from other periodicities that can be An various observed. He argued that the 3 to 6 year cycl industries are sufficiently strongly coupled to move in phat ‘204 up to “the business cycle" at the national level. Another excellent illustration of the 3 to 6 year cycle at ‘the industry level was assembled by Antun (Antun, 1961) in hie study of the ferro alloy industry of Norway (see figure 1b). Outside the system dynamics tradition, the 3 to 6 year cycle [At the level of the economy is probably best known ae “the business cycle", The business cycle 1a rarely viewed ae an however, the perspective of the ‘endogenous phenonenom. ‘This 4a, fiat aystem aynamice study of the 2 to 6 year cycle at the level 1975). as 2 further generalization of the inventory cycle studies at the ‘study can be een Of the economy by Mase (Ha defends the view that the 3 company and industry Level to 6 year fluctuation in the economy is rooted in adjustmente in the use of labor (and not, as 1 nore often hypothesized, in the wae of capital). He argues that the adjustment of production rates to sales in the short run takes place primarily through adjustments in the utilisation of existing capital, by alteration of ouch factors ae the number of shifte, the average Length of the work week, the extent of temporary layoffs atc. Furthermore, ‘he argues that these adjustments are the primary cause of the short term busine cycle. Finally, it 4s worth mentioning the work of Low (Lov, 1980), where he tries to bridge the gap between the eysten dynanice Perspective on the short term business cycle and the more conventional macroeconomic perapactive on economic dynamics. Starting with241 ‘simple Keynesian multiplier-accelerator model Low makes the fevest possible extensions to make the model dynamically complet. ova that 3 to 6 year cycles do arise in the extended model, caused by variations in the utilization of the existing capital stock. Forrester (Forrester, 1962), using advanced nathenatical techniques, hae extended Lov'e work, adding mich support to the hypothesis that adjustments in capital utilization, rather than Variations in capital investment rates, is the prime force behing the short term business cycle. PURPOSE OF THIS PAPER One purpose of thie paper is to:present a detailed theory of Anventory cycles at the industry level, using sulphate pulp a ‘the exampl Another purpose is to fill in» missing piece in the line of work that was discussed above: Statistical work supporting the extetence of the inventory cycle at the industry level - in an Anternational perspective. This far, statistical vork on inventory cycles hae been reeteioted to one nations in epite of the fact that available inventories probably impact on the level of price: ‘and on industry optimiam and production rates, regardless of where the stocks are held. In this paper the sulphate pulp industry Le seen in fan international perspective. The sulphate pulp industry te Adenlly suited for the study of inventory cycles, because sulphate been produced and Pulp te & stable, homogenous comodity that hi traded internationally for several decades. The industry is dominated By Canada, USA, Sweden, Finland and Norvay. statistical tine 25 ‘A third purpose of the paper is to try and exploit the ‘opportunity to use inventory oycles for the purpose of predicting future developments One such prediction of future price was wade in 1962, The method of prediction 18 described and the result Compared with actual developments over the ensuing two years (4, the period 1982 to 1984).2e_ THEORETICAL EXTENSION 242 BASIC MECHANTEHS 12. on the work of the various authors described in section 1, At 1 poneibie to propose the folloving view of inventory cycles at the industry level, and their underlying causes, This view - or theory - forme the fundamental hypothesis which the’ present paper 10 example. socks to test. The sulphate pulp industry te used ‘The prine characteristic of the 3 to 6 year cycle 1s the following: in periods vhen inventories are larger than normal, tend to fall. When inventories are lover than normal, ‘This relation between inventories and price is easily Both buyers and sellers know vhen there te much pulp ‘the opportunity £0 oppositely, when inventories are depleted, understandable. An the market, and buyers understandably negotiate price containing enough pulp only to cover demand for a short period of 1¢ prices. And buyers time, producers see a possibility to incr: have to follow suit, for fear of not being able to feed a ‘of polp to theis paper machin continuows strs But depleted inventories and high prices furthermore tempt. Ultimately production Pulp producers to increase production rati Fates exceod sales, and inventories once more begin to accumulat have to be lowered vhen inventoris Inversely, production rats exceed storage capacity, or vhen prices fall below production cost! ‘The Link between inventory size and production rate closes the Regative feedback loop portrayed in figure 2. this 1oop 18 hypo- thenined as the basic mechaniom behing the 3 to 6 year inventory cycle at the industry Level. To some observers the wavelength of the inventory cycle ope the "inventory-production rate*-loop is capable of generating ‘8 surprisingly long. Many find it hard to believe that measuring 3 to 6 years between peake. ‘The explanation reste with the many delays in the pattern of behavior over one cycle. when tnventort are evelling and prices start to fall, Producers start to reduce production rates. The process, however, time. First of all inventory statistics have to be gathered, not only nationally, but internationally. statistical evidence and Andustry gossip have to convince management that the ongoing Anventory butidup 48 not only temporary. Falling prices in the spot market adds weight to gossip, but even faced vith falling Prices manags nt often hesitate, hoping for a imminent reversal Of negative trends. continue to accumulate, the: te finally no vay around the bitter decision to change production Plane, reduce the labor force (or at Last the hours worked), ~ 4 to reduce utilization of installed capacity. Of adjustment takes place in ali pulp ore oF less simultaneously, given that they G11 respond to the same signal: international price and Anventory levele. As ar mult, aggregate production of pulp finally fale below aggregate consumption of pulp, and inventories of finiened Pulp start to contract. With time inventories decrea 10 much that pulp buyers no longer fee1 certain that there will continue to be a surplus of243 Feadily availiable pulp. They increase their orders, securing rav materials for additional days of production - “Just in ‘Af & consequence there appears a stronger tendency in the market, and prices stabilise. Given the higher demand for pulp, existing inventories appear smaller (inventory “size” is best expressed that can be supported by the Anventory). Ae inventory coverage declines, buyers become increasingly as the number of weeks of eal Worried about insufficient availability. orders are increased, ‘and with time buyers also become willing to pay higher prices to secure detivert Tt takes more tine, however, before increasing sales force pulp producers to up their production raté Anventories are sufficiently deplete After some nonthe, however, producer: that plane Implementing the plane also 18 production rat time, hovever, and before production actually exceeds sales, PULP inventories have become miniscule, the market hectic, and Prices are forced upwards. The “good times" lure the pulp companies to maintain high production rates, and ass consequence, inventories fonce more start to accumalate. Gradually inventories build to “normai* level, prices stagnate, and the industry 1e back to Where it started some years earlier. It takes tine, however, defore groving inventories actually cause & new decline in prices. But finaliy price drops occur, first in the real price adjusted for inflation and later even in the nominal price per ton, thie closes the cycle, and the pattern of behavior repeats itself. Experience shows that one full inventory cycle last between 3 and 6 years. Roughly speaking, it takes a firet yaar of low Production rates to reduce inventories from thelr peak to “normal” 10 Jevela. ‘Then 1t takes a second year while inventories continue to Gecline, before prices increase sufficiently to call forth pro- duction rates that exceed sales. Next, it takes a third year of high production rates to move inventoris back up to “normal* Jevels. And, finally, a fourth year before steadily incressing Anventorie o result in price declines which in turn force Producers to reduce their capacity utilisation, thereby halting the accumulation of inventori ‘The reference mode is summarized in figure 3. ELABORATION ‘Te theory of the inventory cycle at the industry evel cried above, can be supplemented with the folloving hypothe 1+ Me price of pulp wil be influenced by the total, international Aaventory of pulp, Pulp is reasonably imple to transport. tence, Prices will not increase much if, say, Norwegian inventories are ' Gepteted, a8 Long as foreign producers (or consumers) have bulging inventors 2+ Producers vill tend to give discounts to get ria of pulp when thet inventors are evelling. Consequently, achieved price 4s not the aame as Listed price. The price actually received Py producers Le lover than Listed price (the formally agreed price quotations) vhen the market ia falling. Conversely, when pulp 4 acarce, suppliers my well charge above the Listedaa 244 Price of the industry. minimum level of prices in the 3 to 6 year cycis vill be Eelated to variable costs of production in those plants that stil1 operate at the bottom of the price cycle. In fact, the great majority of pulping plants are sti11 operating when prices fare at their lowest. The pulp industry te "cyclical" industry. Sti11, 4¢ 18 primarily inventories’ and, hence, prices, that Production rates fluctuate significantly during the cyel tend to be more stable, moving perhaps 158 above and below the long term average. In other words, capacity utilization varies between approximately 70 and 1008. As ® consequence, the lowest price in the price cycle must be related to the lovest Price for which a majority of producers are still willing to Produce. ‘This minimum price 1s probably Linked to an average Of variable production costs in the majority of existing plants. Plants that have economic difficulties during price troughs Anclude the oldest plante (having high variable costs due to oldfashtoned technological standard) and the nevest plants (having high capital coats because of high debt loads compared with industry average). ‘The price peaks have no relation to production costs. The maximum price reached during the cycle 4e determined by the ‘scarcity of pulp when inventories are lov. The extent to which a2 Pulp becomes @ scarce ~ and hence highly valued - commodity random events during inventory lows depends on unpredictabt: Like transient production problems, atrik shortages of raw materiale or other inputs, transportation bottlenecks, temporary surges in demand and so on. ‘The occurence of euch events 2 However, random events Like those mentioned are impossible to predict, as ie the maximum price 1 they will generat Ihave strong effects when inventories are low, even though they would not have been noticeable 1f inventories were large. Business cycling in the sectors that consune pulp — primarily aper, newsprint and cardboard production - tends to ampli the inventory cycle. The business cycle produces an exogenous 3 to 6 your wavelength in the consumption of pulp. This ‘exogenous driving force will tend to resonate with the “inventory --> production rate"-loop, to generate stronger even” 245 2:_MIASURENENT oF PAST DEVELOPHEITS ‘PRoceDuRE In order to test the theory and the various hypoth ed in section 2, time series date ware collected for the bleached sulphate pulp industry, covering the period 1957 to 1984. 1) quarterly data were collected for production, United states, Anventories and Listed price, covering Cana ‘Sweden, Pinland and Norway. These countries Of bleached sulphate pulp in the Western world and completely are the major producers Goninate the international trade in this commodity. ‘Their relative site 1s indicated by the following production figures for 1980. Production Market share (mi1i-tons/year) canaaa 6.0 “ Usa, 4.0 29 Sweden 2a as Pintand a Norway a Sum "BORSCAN" 13.7 100 Over the Last 25 years the ranking has remained as in the table, but the relative importance of Sveden and Norway has decreased. 1) Date were collected in 1962 for the period 1957 to 1962, and updated An 1984. 4 Canada, Sweden and Finland are the large exporters: Canada covers the US deficits and Sveden and Finland sell to the European markets. During market dovn turns, North America typically tries to sell excess pulp in European markets. To arrive at tine series for total production rate in the “HORSCAN" countries (see figure 4), monthly production data for the five producers were added up to quarterly production figure ‘These in turn were multiplied by 4 to arrive at production rates expressed in million tons per year. As can be seen from figure 4, Production grew fast until 1970, and sonevhat slower in the following decade. Fluctuations around the long term trend were severe, particularly during the period of slower grovth. The tine Dleached sulphate pulp held by producers sm of Inventories in the five producer countries at the and of each ries for total inventories ( figure 5) include ‘The plot shows the quarter. No emoothing has been applied to the data. As can be seen from figure 5, inventories have fluctuated strongly over the last 25 yeare - with a wavelength of some 5 years. thie regularity te nore obvious vhen inventories are expressed as # number of deys « ‘the amount of pulp held in inventory). To arrive at tine series measured as the number of days it would take to produce for this variable (eee figure 6), total inventory at the end of ‘tech quarter (from figure 5) vas divided by the production rate Of that quarter (from figure 4), and the result sultipiied by 365 daye/your. Finally, time ries for the price of bleached sulphate pulp were gathered (#00 figure 7). Firat the quarterly listed price (4n Britten £ per ton from 1950 through 1969, and in US § per ton fromas 246 2970 on) was converted into Norwegian kroner per ton using quarterly ‘averages for the exchange rate between £, § and kroner. Then the Norwegian consumer price index at the end of to convert the ich quarter wae used ries into constant 1979-kroner per ton of pulp. ‘As can be seen from figure 7, the price of pulp has fallen Gradually since 1950, vith pric fluctuating around the long term trend. Pigure 7 represents “international” prices os they Norwegian exporter of pulp. ‘The starting point for figures 4 through 7 were industrial appear from the point of view of Branch statistics obtained from the forest products industry. DtscussioN Im general figures 4 through 7 support the theoretical perei yetive and many of the hypoth 1 Presented in section 2. Starting with the general theory of the inventory cycle, over the last 25 years the auiphate pulp industry has been atrongly cyclical. ‘The cycle can be most clearly seen in inventory statistice, and can leo be traced in price devieopments. the cycle ts much harder to ee in time wries for the production rate. The phase relationship between the variables are 1° when and prices fail vnen inventories are above "normal". “Normal” appears to be between 3 and 4 weeks of coverage. The phase relations of the production Fate te more complex, and is discussed belov. Although these conclusions follow easily from figures 4 to 7, they are not at ai1 apparent in nost time series describing the 16 industry. In order to aleplay the desired resuite, time series must be sufficiently long to cover several wavelengths of the inventory cycle. Leas than 12 yeare of data will not suffice. Secondly, yearly data are too coarse to portray a 3 to 6 year cycle: ‘at least quarterly data are necessary. On the other hand, ‘excessive resolution in time also disguises the inventory cycle. If monthly or, worme, weekly data are used, the 3 to 6 year periodicity Thirdly, is hard to see because of seasonal effects and notec. current prices can not be used because general price inflation nid Fourthly, both the downvard trend and the 3 to 6 year cycle in price ‘the data must cover a sufficiently large fraction of the total market. Individual country (or producer) inventories often move ‘out of phace with the total, and hence may not be a good indicator of total inventoris Next, figures 4 through 7 seem to support the following conctusiona: 1. There is a clear inventory cycle in the bleached sulphate pulp Andustry. The cycle is most easily observable in tine series for total inventories, but can be perceived in eeris for real price ae well. expres 2. Ef one number hes to be picked, the inventory cycle in bleached sulphate pulp should be described as a "5" year cycle. The 14 somewhat. ‘The best indicator seena to be total inventory 1d in days (eee figure 6). wavelength of the cycle appears to have incre over the 25 years since 1957. Inventories have fluctuated with a clear 5 year wavelength since 1968. Prices havea followed the same cycle from 1958 or, possibly, 1953. 247 Inventories appear to cycle around a “normal” value of sone 3 20.4 weeks of production. The total inventory held by producers Farely falle below 2 weeks. Maximum values of 5 weeks are common, ‘and peaks of 10 weeks hava been observed. Since 1970 inventorie have fluctuated around a “normal” value of some 3 to 4 weeks, from troughs of 2 weeke to peaks of 7 weeks. Inventory trough values vary less than inventory peak value ‘The decline in inventories are halted vhen the producers find ie neces mis FY to boost production rates above sal @ecision appears to be triggered simultaneously in a11 producing companies once inventory coverage falle tovarde two weeks. Inventory peak values, on the other hand, are determined by a large number of independent factors. The extent to which produc- ere are willing to hold large inventories ie influenced by expect- ‘ationa about future sales, by interest rat by current social actitades cover lay-offe, by available storage space and 10 fon. Even less predictable was the state subsidy “for extraordinary Anventory accumulation” that was offered in the three Scandinavian countries during the middle 1970's (in order to avoid ley-offe in the puip industry). The subsidy covered much of the inte cost of inventories and probably contributed significantly to the Large inventory peak in 1975-77. continued in 1977. ‘The subsidy wae die~ The long ters trend in the price of bleached sulphate pulp in constant dollars are to be falling at a rate of 2.8 per year. in the price of pulp has sunk at a rate of approximately 26 on] During the last decades, the long term trend per year. This is probably because of increased plant size, material and general increases in productivity. cheaper ‘As a consequence, the minimum price in one cycle may be up to 10% below the minimum price of the preceding cycle. The Fate of decline in the long term trend appeara to be slover during periods of low grovth in capacity, i.e. when fev new plants come on stream. ‘There is a "S*-year cycle in the real price of pulp sround Ate long term trend. Ae for inventories the wavelength varies somewhat over time, but 5 years ie a fair approximation if fone number has to picked. Price troughs have been 3 - 168 below the long term trend, with an average’ of 118. The peaks in pulp prices lie 3 ~ 438 above the minimum price ‘ of the preceding cycle, with an average of 198. the exception ‘was the boom of 1951, when the real price of pulp peaked more ‘than 1008 above the preceding minimum, The price of pulp moves 900 shead of inventory in the cycle. ‘The time series show that prices increase when inventories are Righer than normal. More precisely, inventories have to fallto. re 248 below 3 to 4 weeks of coverage before the industry increa ‘the Listed price (in real terms). And conversely, Anventories have to grow beyond 3 to 4 weeks before the Lated real price actually 1s reduced in real terms. The Fesult Le that the price wave Lies 90° (or 1.25 years) ahead of the inventory wave. The Listed price appears to lag behind the achieved price. Evan though it takes tine before the Listed price is reduced, spot price and company income may well decline rapidly hen inventories start to accumulate. Most likely Producers are willing to give discounts (in order to get Fld of excens pulp) long before the Listed price is lowered through formal agreements. conversely, spot prices and company income may Well increase faster than Listed price, when inventories are depleted and pulp ie getting scarce. ae ‘This conclusion does not follov directly from figures 4 through 7, aince we have no tine series for achieved price, only for lated price. sense and by fgsteland's data ( ‘The conclusion is supported, however, by common figure 1). Hie date indicate that the achieved price responds rapidly to inventory chang Listed price may lag half a year behind achieved price. The level of price fedictable than the Level of price troughs ‘the pressure in the market at the time of minimum inventoris 8 ore: 10s ‘The peak value of prices Le determines by Accidental shortages may send prices soaring - at least the 20 jure in the market at the time of minimum inventori the pri Accidental shortages may send prices soaring - at least the spot price, but even the Listed price if supply probleme endure. The price minima develop in a more predictable fashion, Probably due to their relation to the slowly decreasing real costs of production in the industry. Achieved prices (spot Price corrected for discounts and expressed in local currency) probably reflect actual production costs vhen inventories bulge. jer to press prices tovarde the producer's variable cos! In periods of exce inventories, pulp buyers find it Producer coste are typically expenditures in local currency. ‘Thus 9 Norvegian seller my 1 have to accept @ § price below the Listed § price in times when high exchange rates give hin ‘a larger azount of kroner per $. Pulp enles appear to move 180° behind inventories in the cycle. In other words, pulp sales are large vnen inventories are email and vice versa. This conclusion ia derived by combining the Anventory statistics (figure 5) and the production figures (figure 4). this indirect method 1s used, since data for Pulp eales are not available. From pure logics it follows that Af inventories grow (¢a11), jos must be lover (higher) than the production rate. Applying this algorithm, one obtains the rough outline of historial Ae con be seen by comparing figures 6 and 8, low (high) sales coincide wien high (low) Anventort dorated by Hpstelind's data (figure 1). sales shown in Figure 8. |. This conclusion te alse corro-a2. a 249 But Why ahould sales be large vhen producers’ inventories ‘re email? And vice versa? The Likely explanation is that Anventoris are large at the ive production, Anventories gradu both producers* and consumers! ‘Throughout the period of exc same tine, ot only producers’, but also coneunere’ ally £111, Te result te a non-causal, but strong, correlation between large producers' inventories and emai Because large inventories also coincide with tin falling prices, one has the surprising conclusion (at 1. theoretical micro economics) that sales are small vhen prices ‘Fe failing. And conversely, booming sales coexist vith increasing prices at other points, in the taventory cycle. ‘Thue consumers will need less pulp (because their inventories re f0i1) Just when the producers would want to sel more pulp (because their inventories are full). ‘There Le no simple phase relation between the of pulp sates. Production rate on the one hand and inventory oF price on the other. Delow normal and vhen pric relationship ie unambiguous. Production rates tend to be high when inventories are are above the trend. put neither Tt 4a better to view pulp production as a lagged version of ‘Le conclusion Pulp sales, with a lag of some 6 to 12 month: 4s also supported by Hgsteland’s data (figure 1). ‘Thie relation of sales to production 4» caused by the coupling through Producers’ inventory, which is kept within rather strict Limite, Fry 13. The ian eulphite pulp industry and the international sulphate pulp industry move in phase, there is an extra: ordinary degree of correlation between Norwegian suiphite Pulp inventories (in figure 1) and HORSCAN sulphate pulp joventories (in figure 5). the degree of covariation is 11 the wore surprising given the vast differences between the two products, the two producer groups and the tonnages Anvolved. the covariation in prices 1s not less impressive. ‘Tis ie but another example of the fact that the 3 to 6 Year inventory cycle in most industries are in phase, Presunably because increa 4n production rates in one sector Bullen tendenct £0 epread throughout the economy, a8 do be require output from other sectors. tena ish (Randere, 1978). ‘Te foregoing results and conclusions concerning past develop- ments An the aulphate pulp industry are summarized in figures 9 and 10. Figure 9 portrays the phase relationships that appear to exist ‘ Detween central variables in the inventory cycle. the figure ‘Wo important ‘should be viewed ag an extension of figure 3. Perepectives follow from figure 9: 4) Production can be mB AS a lagged version of sales. when sales f011, production ts pulled after - because the ‘two are couped through inventories. Inventories would SOW out of hand 1f production were not stopped. when sales increase, production must follow, lest. inventories23 250 shall be depleted. In shorts production rates are primarily governed by inventory sti Higher prices a 1A) Price dose have some effect, however. but only ‘cause production to ineres Tf enles are declining of other reasons, production may fa11 even vhen prices are growing. Figure 10 portrays the system structure that appears to underlie the inventory cycle at the industry level. The structure devia from the structures proposed by earlier authors (Meadows, Hgsteland, Antun) in that production and sales depend more on inventory sie and Me ‘The conclusion 1s derived primarily from the ph relationships in the tine series data reviewed (eulphite pulp, fon price. ferro silicon, sulphate pulp). ‘These phase relationships are not anetly explained by the traditional structures. The proposed structure does, however, require for the atudy. 24 PREDICTION OF FUTURE DEVELOPMENTS PROCEDURE Prediction of future price 1s of great interest to decision makers in the pulp industry. Production plans, sales atratesy, ‘and investment policy - all depend on opinions shout the future price of pulp. Scientifically speaking, price prediction in socto-economle systems in often impossible, except with wide margins of uncertainty. ‘The system structure and the amount of noise often cause enough uncertainty in the forecast to make it usele for practical pur Poses. (This of course does not exclude the possibility of doing crystal ball guesswork - without indication of the margin of uncertainty). ‘The preceding sections reveal that the sulphate pulp industry Ao characterized by a system structure and ® level of noise vhich make useful price predictions possible ons 1 - 5 year time horizon. The system structure creates @ stable inventory cycle with a 5 year wavelength in inventories and price. The noise characteristic are such that quarterly data are well behaved and Tt Se Amportant to add that this conclusion does not necessarily hold for price predictions on @ shorter or on a longer Im both cae other factors than those responsible for the inventory cycle. time hortzon. Price developments are determined by mh factora constitute oystens vith different atabiiity and nol25 characteristics, vhich may vell make useful (1.8. nigh certainty) Predictions impossible. ‘Tha recommended procedure for 1-5 year prediction of pulp prices, based on inventory cycle theory, 18 as follows: ee rae et eae Pree dT cesca wo ae tempers eottante ee eer areas Fe ocr eet ates eater [2] eae Preceding decades. 111) atnusota [Be *6.sio(t0h oe eresea co ene querterty tine series of the total inventory of pulp, expressed An days of production. Te sinusoid 1s used to determine tne wavetensth [T'] in tne price development. 40) Turthermore, the sinusoid 9 used to locate the position Of future troudhe and peaks in the price of pulp by 90° anead of inventories in the juning that price mover cyete, te, that price ta proportional to sin(2rt 4p -). ¥) The amplitude of the price wave ie determined fron a time series of the ratio between the original price and the ratiog = pricer/ Ae™*l-t+) ), tre anpiivuce[C] te decived by titting « alnusotd wien tne rcorrece® phase [l +Cen(2nt ep-2)] exponential trend (4. 251 26 to thie tine series. el) Me extenuton of the stousold Ls aded tothe exteneton of the exponentia2, giving the most 1ikely prediction ei h of pete [ae at), (1-4 Cain (2, 2-B))]. Vi) A maroin of uncertainty 18 added, based on the standard deviations in the estimated values of the rate of Gecline, the wavelength and the axpitua brscossrow ‘The fitting of the exponential and the sinusoid to historical ata can be done quantitatively, using standard regression techniques, or manually, by hand in graphical plots. The last Procedure does not necessarily give inferior resulta; but it is ese reproducible. ‘The manual procedure does have the savantage of being simple, fast and transparent, even for ‘ Ron-mathematiciar ‘The crucial point in the recommended procedure is to use inventory series to determine the wavelength and to locate the position of troughs and peaks. ‘The inventory fluctuations are much more accentuated - and easier to observe - than fluctuation in other variables like price, production and sal inventory statioticn are easily avatiable, and they show Little seasonal fluctuations given that they are true levels or accumulations. ‘The position of price peaks and price troughs are found by shifting7 252 the inventory vave one quarter wavelength ahead. Rather than relying on this phase relationship, one may choose to fit a ‘sinusoid with the right wavelenght to the price series in order to determine the phase relation between price and Anventort ven 1f inventories play a crucial role in the recomended method of prediction, one should of course use all other available tine series to check consistency and to renove ambiguities in the location of trougha and peaks: 28 5: _COMPARIGON OF PREDICTED AND ACTUAL DEVELOPMENT Figure 11 shows the result of applying the manual method of Prediction, based on time series for the period 1952 to 1982. ‘The graph 1s copied from an consulting report by the author dated September 27, 1902. Figure 11 also shows how prides actually developed over the frat two years of the prediction period. Figurt 4 through 7 aia Aieplay that actual developments over these two years were in accordance with inventory cycle theory. The inventory cycle hae continued in the sulphate pulp industry in the 2 yeare following 1982, am it 41d in the 25 years preceding 1982. In retrospect it i ey to aee that the price prediction was auffictently precise for ite intended purpose: to determine vnether pulp prices would atart to increase within the next few years, creating a more favourable climate for the planned sale of a large pulping plant. ‘The con- sulting report concluded that pulp prices would halt their decline in the 4th quarter of 1983 (42 quarters) and reach a new peak in the Let quarter of 1986 (42 quarters). Hence the aale of the pulping plant ought to be postponed to 1984/85. In fact, prices aid reach a minimum in the Let quarter of 1983. Aftervards prices have inched upwards, vhila inventories have Geclined. The most recent available statistics (May 1964) show in the process of failing tovarde the magical Limit Of sone 3 to 4 vacks of coverage vnich formerly has signalled significant price snor ‘The spirit of the sulphate pulp2 Andustey 18 once wore rising. One has come @ tong way from the 253 1962. The price of pulphing plants are substantially higher - and increasing. prscussio# Im principle the total, global inventories of pulp influence ‘the price of pulp. All finished pulp - resting with the producer, in transit to consumers, in consumers’ taventories and elsewhere - 4s in principie available and thus acts to press price, ‘The current Work seems to indicate, hovever, that the producers‘ inventories give a good indication of market conditions. This 1 advantageous from a practical point of view, since producers’ inventories are the most readily available inventory statistics ‘The individual supplier of pulp is most interested in the price the discounts he has to give he can achieve, i.e. tated price 1 in order to se11 his pulp. The achieved price is closer to spot Price than to Listed price. But time series for spot prices are hard to define as are tine series for achieved price. Hence it ‘appears more practical to base price predictions on tine series of Ldated price, Keeping in mind that achieved price may well deviate from the Listed price - increasing before {t does and declining ahead of At. Finally, using Listed price makes it unnecessary to ‘smooth the price series, since the listed price 1s itself = “running average" of spot prices. The Listed price ie quoted roughly once every quarter and is representative of the current going rate. Even simpler than the manual method of prediction, te to 30 establish a long, quarterly time series of inventories and use this ‘curve as a rough indicator of one's present position in the Anventory cycle. Inventory figures in tone, from one major supplier of pulp, may suffice for thie purpose. But quarterly data mensured in days of production covering a significant fraction of the international market, gives a much better indication Of the state of affairs.a 254 &_covcwerox ‘Thho paper has elaborated on the inventory cycle theory and propor @ detailed theory for inventory cycles at the induetry level, me tailed theory Le supported by tine series obtained for the sulphate pulp industry, covering some 25 yest ‘A method for prediction of pulp prices, based on the detatied theory, was proposed. The method vas used in 1962 to forecast the Price of sulphate pulp. Comparison with actual developments over te. ‘the two ensuing years shoved that the method gave good r ‘The method of prediction {# most Likely applicable in other industries producing comodition. 2 anton, Ps, ichelor's theste, Noi ‘SF Management, Bekkestua (Oslo), 1981 Erekiel, M. “me cobweb Theorem, Quarterly Journal of Economice, Vol 52, ps 255, 1938 Industrial Dynamics, MIT Pri Forrester, J.W., + Combridge, Me Forrester, N.B., Wgsteland, J., "Stock Fluctuations in the Pulp Industry" in Upnnstedt, L., and Randers, J., (editors) Wood Resource ics_in the Scandinavian Forest Low, 6, me Multiplier-Accelerator Model of Business cycles Interpreted from 8 Syatem Dynanice Perspect in Randers, J., (ed.) Elements of the Sytem oe sched, MIT Prees, Cambridge, Hast-, Majani,B. ang “can Recessions be Predicted?", in Longe! Range Planning Mairidakis ‘S-, Vol 10, April 1977 Mave, Wa. Meadows, D.L., 108 of Commodity Production Cycles, Wright-Allen acre es “Dynamic Modelling as 2 Tool for Managerial Planning: A Case Study of the US Hog Industry", Proceedings of the 1973 Summer Computer Simulation Gonference, hontreal, 17S “om den skonomisk utvikling pA 10-15 A (“Econontc Development ina 10 - 15 Bergen Bank Kvartalsskeift, No 4, 1978 Naitl, Ros i NeJ.,'Randare Je, Simpon Woks. Randers, 3-5 atx iF Perapective"),u 2 255 SALES RATE +f INVENToRY PRopuction : RATE e-) \ * - PRICE CAPACITY UTILIZATION Figure 2. The basic mechaniem behind the inventory cycle. Pup INVENTORY’ (tons) as os io so Figure 1b. The inventory cycle in the Norwegian ferré silicon Andustry.” (From Antun, 1981. Exports from the A/S FESIL # Co. gx running averages of monthly data. hieved" price in constant’ 1977-kroner.) group of ‘compant Quarterly data for Figure 3. The reference mode of the inventory cycle theory.(inthovsand tons par year ) i i i E 1000] a ae mr ‘ate ae ‘me ‘we Figure 4. Production rate. Quarterly data for the NORSCAN countries ~ production of bleached sulphate pulp. 9sz Cin theosand. bons) oe g 8 & F F = ‘hee a ‘wre me ‘ie ‘ae Pigure 5. Inventories in tons. Quarterly data for the NORSCAN countries ~ producers’ inventories of bleached sulphate pulp.Cin days) w) | oy 50) 7 1 30} ' 20 0 aa a co mm ms ia Sage coed mcam reais ats ePreeeaees ar 8 8 (in 19 Wo parton) ‘Sone tise to0e i i im i im in a Garterly data for the “listed price" of bleached sulphate pulp expressed in constant kr/ton.258 40 Figure 9. Sumary of phase relationships in the inventory cycle. *eamnB73 wor3onpo26 pur AZ0WweAUT SuFUTUCS Aq PeATIEP ‘SoTI3UMCO NYOSYON ea WOT} SOTeS 30 NOONE reve sores poatzeg -@ exnsra om ow su ow au om, ou a9 Cook and t00y pootcons 1)a [PRooucens} Consumeas 1 >< IWENTORIES @ @ t ' a) H ; ‘, \ A ? “i - ; RATE (Un) b= ~~ [Provuctio a oo (m) ii Figure 10. Summary of the system structure behind the inventory cycle. wo3/x0u034-2861 NI aNd ZIYBZINS aaHVaTE 4O soTWa Price of bleached sulphate pulp made in velopment in price. red with actual de (2. quarter 1984 provisonal.) Figure 11, Prediction of the real September 1962