W O R K I I V G P A P E R
INCOKE AND PRlCE ElASrJUrTlCj
OF
DEMANDH)R PAPER
ANDBOARD:
CONSIDERATIONSIN
NUMERICAL EslWATlONEsko Uutela
December 19B4 WP-84- 102
I n t e r n a t i o n a l I n s t i t u t e for Applied Systems Analysis
NOT FOR QUOTATION WITHOUT PERMISSION OF
THE
AUTHORINCOHF,
AND
PRICE E~~ OF DEMANDFDR
PAPER AND BOARD: CONSIDERATIONS IN NUMERICAL ESIDUTIONEsko Uutela
December 1984 WP-84- 102
Working Papers a r e interim reports on work of the International Institute for Applied Systems Analysis and have received only limited review. Views or opinions expressed herein do not necessarily .represent those of t h e Institute or of its National Member Organizations.
INTERNATJONAL INSTITUTE FOR APPLIED SYSTEMS ANALYSIS 2361 Laxenburg, Austria
The objective of the Forest Sector Project a t IIASA is to study long- term development alternatives for the forest sector on a global basis.
The emphasis in t h e Project is on issues of major relevance t o industrial and governmental policy makers in different =egions of the world who are responsible for forestry policy, forest industrial strategy, a n d related trade policies.
The key elements of structural change in the forest industry are related to a variety of issues concerning demand, supply, and interna- tional trade of wood products. Such issues include the development of the global economy and population, new wood products and substitution for wood products, future supply of roundwood and alternative fiber sources, technology development for forestry and industry, pollution regulations, cost competitiveness, tariffs and non-tariff trade barriers, etc. The aim of the Project is to analyze t h e consequences of future expectations and assumptions concerning such substantive issues..
The research program of the Project includes an aggregated analysis of long-term development of international trade in wood pro- ducts, and thereby analysis of the development of wood resources, forest industrial production and demand in different world regions. This article studies alternative approaches for forecasting demand for paper and board. Some numerical results and comparison with earlier work is given. The paper resulted from the authors participation in the Young Scientist Summer Program in 1984 a t IIASk
Markku Kallio Project Leader
Forest Sector Project
CONTENTS
1.
INTRODUCTION
2.
MODELS USED FOR ESTIMATING INCOME AND PRlCE ELASTICITIES OF DEMAND
3.
DATA PROBLEMS FACING ELASTICITY MEASUREMENT
4.
SOME EXPERIENCES FROM WORK DONE AT A CONSULTING COMPANY
4.1B a c k g r o u n d
4.2
D a t a a n d Models U s e d
4.3S t a t i s t i c a l T e s t s Used
4.4E s t i m a t i o n R e s u l t s
5.
COMPARISON OF RESULTS WITH OTHER STUDIES
5.1B a c k g r o u n d
5.2
I n c o m e E l a s t i c i t i e s
5.3Own-Price E l a s t i c i t i e s
5.4C r o s s - P r i c e E l a s t i c i t i e s
6.CONCLUSIONS
REFERENCES
APPENDIX
INCOlE AND PRICE ELASIlCITIES OF DEWWO POR PAPER
AND BOARD. CONSIDERATIONS IN NUKERICAL
lCSlTKATION
Esko Uutela
1. INTRODUCIION
The Global Trade Model
(GTM)
developed by t h e Forest Sector P r o j e c t(FSP)
a t IIASA for analyzing global production, consumption, a n d international t r a d e in forest products h a s four basic forces driving t h e model (see Kirjasniemi e t al. 1983, Dykstra a n d Kallio 1984):(i) Demands for e n d products.
(ii) Competitiveness of industries (technology, production costs).
(iii) Available resources.
(iv) Differences in t r a n s p o r t a t i o n costs.
Roughly speaking, t h e demands for e n d products in different regions d e t e r m i n e t h e absolute level of global production of forest products, a n d e l e m e n t s (ii), (iii), a n d (iv) d e t e r m i n e how production i s d i s t r i b u t e d between t h e regions.
The preliminary Global Trade Model (GTM-1) includes forest p r o d u c t s consumption as a p a r t of t h e objective function. The demand module of GTM-1 is given a s a price (or inverse consumption) function showing for e a c h region i a n d product k t h e relation between price
nil,
a n d c o n s u m p tion ci, (Dykstra and Kallio 1984):Consumption ci
=
( c q ) refers t o t h e demand for end products which a r e consumed outside t h e forest sector. The demand for i n t e r m e d i a r y pro- ducts, i.e., logs, pulpwood, pulp and recycled paper, can be derived from t h e technical input coefficient matrix4 =
(&,) and t h e o u t p u t levels of e n d products, assuming t h a t consumption of intermediary p r o d u c t s out- side t h e forest s e c t o r is negligible. The price function ofGTM-1
for e n d products is:where -l/yik i s t h e price elasticity of demand and )CJc t h e level p a r a m e - t e r for t h e demand curve. As can be seen from equations (1) a n d (Z),
GTM-1
is a s t a t i c model with simplified demand functions where all important d e m a n d shifters a r e exogenous and t h e i r effects on c o n s u m p tion levels a r e investigated through separate analyses.The purpose of t h i s paper is t o p r e s e n t some estimation r e s u l t s of paper and board consumption functions with special e m p h a s i s on t h e long-term price and income elasticities of demand. The r e s u l t s described h e r e a r e based largely on practical forecasting work in a f o r e s t industry consulting company. Data problems r e l a t e d t o elasticity m e a s - u r e m e n t a r e discussed a n d estimation r e s u l t s compared with t h o s e from consumption studies by o t h e r researchers. Attention is also paid t o t h e dynamic aspects of elasticities, i.e., changes in t h e n u m e r i c a l values of elasticities over t i m e a n d income level, a s well a s t o t h e i r variation across countries and different paper grades.
2. MODElS
USED MIR ESI'IMATING
INCOMEAND PRICE
ELASrJCITlES OF DE36ANDMost of t h e r e c e n t international long-term consumption s t u d i e s made in t h e pulp a n d paper sector have been based on pooling cross- sectional a n d time-series d a t a instead of using t h e traditional country- wise time-series analysis. As Buongiorno (1978) h a s s t a t e d , pooling alleviates problems posed by s h o r t time-series, small variability in t h e data, a n d high collinearity between explanatory variables m e a s u r e d in value t e r m s .
Both s t a t i c a n d dynamic models have been used for estimating con- sumption functions. One of t h e m o s t i m p o r t a n t s t a t i c functions, widely used in g e n e r a l d e m a n d studies, i s (Houthakker 1965):
=
A . #'jPC1 dr
&..Cijf J
,
st$ iJt ~ j f (3)where a , j , a n d t refer t o a specific country, product, a n d y e a r , respec- tively. C i s consumption p e r capita,
Y
i s income (orGDP)
p e r capita,P
is t h e price of t h e p r o d u c t i n question, and
P'
is t h e price of t h e m o s t direct substitute.A
dynamic version of t h e model (3) based on Nerlove's partial adjust- m e n t model i s (see Nerlove 1958):where t h e only difference to equation (3) is the introduction of a lagged dependent variable CijVt-] (= consumption in the previous year) into the model. The quantity demanded in period
t
depends on, among other things, on t h e quantity demand in previous periods due t o a habit forma- tion process nrhich is characteristic of human behavior (Houthakker and Taylor 1966). In the case of paper and board, the habitual n a t u r e of con- sumption is evident; t h i s is stated, e.g., by Aberg (1968).Nerlove's dynamic theory allows a distinction between short-term and long-term demand elasticities (Nerlove and Addison 1958). Nerlove postulates that equation (3) represents, in fact, t h e long-term demand if demand adjusted immediately t o changes in explanatory variables:
log C,h.
=
o+
b - logYit +
cj logPijt +
d j log P I i j t+
rijt1 I (5)
where
c ~ ; ~
is t h e long-term equilibrium of consumption which would be observed if all explanatory variables remained a t a fixed level for a sufficiently long time. Since Yit ,Pijt,
andPeijt
change continuously, CGt is never observed. To measure the ratio between Cijt and C$ the hypothesis used is t h a t the ratio (c&/ Cijt) "ill be closer to unity than the ratio (Ci;t/ Cijef because there nil1 tend to be greater coincidence between short- and long-term demand in year t than between short- and long-term demand in successive years (Nerlove and Addison 1958, see also Koutsoyiannis 1973). This implies thatwhere h measures t h e velocity of adjustment to demand rigidities. By substituting (6) into (5)
s o t h a t
=
m(l-A)-$(l-A) ACijt j Cij ,t -1 'ijt
we obtain t h e short-term consumption function, from which it is possible t o estimate both the short- a n d long-term elasticities by computing from the regression and of t h e relationships:
where capital l e t t e r s r e f e r t o short-term elasticities and t h e lower case t o long-term elasticities. The closer X is to unity the faster t h e velocity
of adjustment, t h e static model being an ultimate case where t h e adjust- m e n t of consumption is completed in one year ( A
=
1).In several international studies either a static (3) or a dynamic (4) model, or both, has been used for estimating forest products consump- tion functions (with or without some mo&fications); for example, among others, by Buongiorno (1977, 1978, and 1979), FA0 (1977a), Baudin and Lundberg (1984) and Suhonen (1984). In t h e FA0 (1977a) study, an addi- tional explanatory variable, literacy r a t e
4f,
was introduced into the s t a t i c consumption model for newsprint and printing and writing papers in t h e developing countries, the model t h u s being:In his work, Wibe (1984) used a simplified "dynamic" function where time
T
represented (as a yearly index) substitution effects over time:where T is time measured in years and aj is interpreted as t h e yearly r a t e of substitution for forest products.
In several earlier paper demand studies, income (or GDP/capita) was t h e only explanatory variable in t h e cross-sectional consumption models (e.g., F A 0 1960, FA0 1963b, Sundelin 1970, 1976). The relationship between paper consumption and income per capita in these models was assumed to be an S-shaped curve based on Engel's law1 of diminishing marginal utility with increasing incorrie. Different mathematical func- tions have b'een used to describe t h e cross-sectional S c u r v e in year
t ,
of which one of the most complicated is (FA0 1960):where
Y
is GDP per capita. S, t h e saturation level of consumption (kg/capita) defined n priori, and p and p parameters of the log-normal function. A simpler formulation was used by Sundelin (1976):which is, in logarithmic form, a second-degree curve.
All t h e above consumption functions a r e implicitly based on t h e con- s u m e r demand theory. However, t h e demand for forest products has only indirect connections t o this theory.
A
relatively small fraction of forest products is consumed a s such directly by c o m m e r s (e.g., sawn- wood for do-it-yourself purposes, household and sanitary papers); t h e bulk comprises intermediate-
or r a t h e r complementary-
products related t o t h e output of other commodities. Additionally, t h e value of forest products is normally very small compared with the total costs of t h e consumer or producer. According t o Aberg (196B), paper and board costs in Sweden accounted for only 1-2 percent of t h e total consumption value in t h e industry and trade, as well a s in private households. Theonly exception in the paper and board sector is in printing houses; the costs of newsprint might be up to 30% of t h e total costs of newspaper publishing. Therefore, i t has been &scussed whether traditional demand theory, using relative prices of substituting commodities as exogenous variables, is able to explain the variation in consumption levels.
Although the product characteristics of forest products are t h e rea- son for assuming their demand t o be d e r i v e d d e m a n d r a t h e r than f i n d c o n s u m e r d e m a n d , the estimation of demand relationships from produc- tion or cost functions is not as straightforward as for many capital goods or rawT materials. Especially in t h e paper a n d board sector, t h e consump- tion is widespread among numerous branches of t h e economy indicating t h a t t h e r e a r e also several different production or cost functions t o be identified. The complexity of processes or systems, to which forest pro- ducts consumption in these branches is tied, might also contribute difficulties to t h e interpretation of estimation results. For these rea- sons, and t h e limited availability of sectoral input-output data, t h e r e a r e only a few studies h o w n t o the present a u t h o r which .apply derived demand theory to forest products. Doran and Williams (1982) used t h e Diewert cost function (or a generalized Leontief production function, see Diewert 1971), based on Shephard's duality theorem (Shephard 1953) of t h e equivalency of t h e production Function, and minimum cost function t o analyze t h e demand for domestically produced sawmn.ood in Australia in t h e period 1957-1977. Own-price elasticity for sawnwood and cross- price elasticity for five substitutes/complements were estimated.
Frenger (1983) used the same generalized Leontief production function approach for estimating price elasticities in six manufacturing sectors of Norway covering t h e period 1962-1980. Both short-term and long-term elasticities were estimated for four variable inputs: material, energy, labor, and capital.
I t
is evident t h a t t h e r e a r e too few experiences of t h e use of t h e derived demand approach in the forest s e c t o r to make any conclusions as t o its suitability for consumption analysis of forest products. The two above-mentioned national studies only provide examples of results from two specific cases, which a r e not to be generalized to the global level. On t h e other hand, t h e r e is little evidence t h a t models based on t h e tradi- tional consumer demand theory, in spite of their theoretical weaknesses, would in practice produce forecasts of a completely false magnitude. For example, t h e paper consumption forecast prepared by FA0 (1963a) in t h e early nineteensixties for the whole of Latin America u p to 1975 used a simple income-related model and differed by only 1.8% from t h e a c t u a l consumption in 1975 (1 974-76 three-year average, see Uutela 1979). For these reasons, t h e r e s t of this paper deals with elasticity estimation for paper and board obtained from traditional demand models.3.
DATA PROBLEMS FACINGELASI?CITT I i E S U R w
When using international data for cross-sectional or pooled cross- sectional a n d time-series analysis, t h e r e arise some data problems which may have a drastic effect on t h e estimation results. These problems include, among others:
(i) Availability of valid and reliable price d a t a directly comparable with similar d a t a series from other countries
(ii) Choice of an appropriate deflation method for all value d a t a (iii) Exchange r a t e variation when converting value d a t a from
national currencies into a common currency.
Availability o f suitable price data (2) is a problem because value d a t a on domestic consumption a r e normally not available. The FA0 yearbook statistics include only i m p o r t and export values of commodities t r a d e d internationally. Recent efforts by
FA0
t o collect domestic price s e r i e s for forest products have yielded relative1)- s h o r t time s e r i e s for a few countries, with varying product classification. Only a few developed countries systematically publish domestic price series for specific paper and board grades. For building a global model, t h e contribution of t h e s e price series is insufficient. Therefore, t h e most suitable available infor- mation on price levels for many countries i s t h e u n i t value of i m p o r t s o r exports. FA0 (1977a) a n d Buongiorno (1978) in t h e i r studies u s e d aver- age export (FOB) values for n e t exporting countries and average i m p o r t (CIF) values for n e t importing countries. This approximation s e e m s justified for c o u n t r i e s which a r e e i t h e r major importers or m a j o r export- e r s of paper a n d board grades. Problems arise when i m p o r t e d or . exported quantities a r e small and/or there a r e clear quality/gradedifferences in comparison with domestic consumption, i.e., e x p o r t or import prices do n o t r e p r e s e n t t h e average prices of domestic consump- tion of a c e r t a i n product group. Fortunately, this heterogeneity h a s seri- ous effects only on a few countries and products; in many c a s e s domestic prices follow closely t h e world m a r k e t prices, with allowance for differences in transportation a n d o t h e r trading costs. Buongiorno (1978) made a comparison between wholesale domestic price a n d u n i t values of foreign t r a d e of newsprint for five countries, resulting in correlation coefficients varying f r o m 0.90
(FRG)
t o 0.99 (US). However, t h i s does n o t hold for all countries a n d products. Before any statistical analysis t h e u n i t value d a t a s e r i e s should be checked and distorted price s e r i e s replaced by domestic price estimates.Deflation of value data (ii) should always be m a d e using national deflators a n d national currencies. When using FA0 u n i t value d a t a for product prices, prices a r e directly expressed i n c u r r e n t US dollars.
These figures include both t h e US inflation a n d changes in e x c h a n g e r a t e s . Since consumption i s r a t h e r a function of real income a n d yea1 prices t h a n a function of money income and nominal prices, o n e h a s t o
adjust t h e u n i t value s e r i e s from t h e US inflation. In some s t u d i e s (e.g., FA0 1977a, Wibe 1984, Suhonen 1984) the prices in c u r r e n t US dollars have simply been divided by t h e US wholesale price index, which closely follows t h e index of export prices of manufactured goods. Then i t i s assumed t h a t c h a n g e s in official exchange r a t e s sufficiently reflect differences i n cost developments between t h e USA a n d o t h e r c o u n t r i e s . Some r e s e a r c h e r s have u s e d US nominal prices a s s u c h without deflating t h e n a t all (see e.g., Buongiorno 1978). In these cases i t i s evident t h a t t h e n u m e r i c a l values of the e s t i m a t e d elasticities a r e b s t o r t e d by t h e likely correlation between price and t i m e (or inflation).
&change r a t e varinfions (iii) cause comparability problems between countries. As stated in an earlier paper by the author (Uutela 1983), t h e choice of base year and currency used in cross-sectional s t u h e s may affect rankings between countries with respect to GDP per capita and product prices.
I t
has been shown t h a t when using the same per capita income and per capita paper consumption data-
but base year 1978 instead of 1970 for t h e US dollar exchange r a t e for GDP figures-
t h ecountry rankings changed and the variation in t h e data increased, t h u s reducing t h e overall
fit
of the cross-sectional model remarkably(JPI
1982). Large fluctuations in the value of t h e US dollar since t h e begin- ning of t h e period of Aexible exchange r a t e s i n the early 1970s make reli- able comparisons of purchasing power and commodity prices over time and between different countries difficult.The effect of different deflators and exchange r a t e variations on price series a r e illustrated in I .
I t
can be seen t h a t price series develop r a t h e r differently depending on t h e data manipulation made, and evidently t h e r e would be remarkable differences in the parameter esti- mates of consumption models based on these data. To see these differences, t h r e e consumption models for two products, total paper and board and newsprint, and one country, t h e Federal Republic of Germany, were built of using data for 1964-02 based on FA0 (1977b, 1984) andIMF
(1982). The models were:
log
Cjt
= a j + b j logYjt
+ c j logPjt +
cjt (14)log
C j , =
log A+
a, T+
b j logY j t +
c j logPjt +
cjt (16)of which (14) and (15) a r e t h e logarithmic forms of models (3) and (4), but without prices of substitutes a s explanatory variables, and (16) is the logarithmic form of (11) used by Wibe (1984). The price variables (all CIF import prices) used were chosen a s follows:
A) Nominal price in US dollars.
B)
Constant price in 1975 US dollars, deflated by the US wholesale price index.C) Constant price in 1975 US dollars, deflated by the German industrial products price index.
D)
Constant price in 1975 US dollars, deflated by t h e US wholesale price index and adjusted for exchange r a t e fluctuations by using t h e average r a t e between t h e US dollar and German m a r k for t h e period 1964-02.E)
Constant price i n 1975 US dollars, deflated by t h e German industrial products price index a n d adjusted for exchange r a t e fluctuations by using t h e average exchange r a t e for t h e period 1964-82.GNP
was measured in all cases a t constant 1975 prices, deflated by t h e GermanGNP
deflator, and converted into US dollars by using the average exchange rate of 1964-1982. The results of a total of 30r 1 T o u l paper and b a r d Price (USSftonl
b) Newsprint Price (USSftonI
Explanations tor A, 6. D, m d E: see text
FIGURE 1. An example of the effects of different deflators and varying exchange rates on the price series of paper products in the Federal Republic of Germany 19M-82. (Explanations for A,B,D, and E given in the text).
regression equations a r e presented in Appendixes 1 and 2. Table 1 sum- marizes t h e results of all t h e static and four dynamic models.
The measured price elasticities varied from -0.15 t o -0.40 for total paper and board, and from -0.14 to -0.29 for newsprint, depending on t h e price series used. Surprisingly, variation in income elasticities, measured from different models, was even more important; from 0.81 to 1.78 for total paper and board, and from 0.97 to 2.00 for newsprint. The s t a t i c models were statistically better than dynamic models including time or lagged consumption. Parameter estimates i n t h e dynamic models h a d negative signs in 19 of 20 dynamic regressions, t h u s violating the assumptions of Nerlove's partial adjustment theory on which the model was based. Only in four equations with time (years) as t h e expla- natory variable is the parameter estimate statisticaly different from zero a t t h e 0.99 confidence level.
High multicollinearity (0.94-0.98) between GNP a n d time/lagged consumption was t h e main reason for t h e failure with dynamic models.
When introducing a t i m e variable in consumption models, t h e standard e r r o r of income elasticity a t least doubled and
R2
did not increase remarkably; in the case of Ci,l it even decreased. Nominal prices also correlated relatively closely with time (0.77-0.95) a s well as with GNP (0.82-0.92); in static models, however. standard errors of parameters were small. The rejection of dynamic models because of t h e observed multicollinearity and t h u s indeterminate parameter estimates narrows t h e variation of t h e mzasured elasticities, but the range in s t a t i c models is still confusing:Income elasticity Price elasticity Total paper and board 0.81 to 1.37 -0.15 t o -0.40
Newsprint 1.02 to 1.64 -0.16 t o -0.27
As a result of this exercise, i t seems t h a t more attention should be paid t o data preparation work before elas ticity estimation. F u r t h e r conclu- sions from the results a r e prevented by the limited d a t a material used, but apriori, the following comments can be made:
1) Domestic deflators should be preferred t o common deflators because change in exchange rates may not strictly follow m e r e n t cost developments between countries
2) Exchange r a t e variations should be excluded from t h e price series if most of consumption is domestically produced and domestic pricing is n o t affected by foreign currencies. If the bulk is imported, i t would be logical to include exchange r a t e changes, which afTect world market prices, i n the price series to be used.
TABLE 1. Estimetion results of different paper consumption models for the Federal Republic of Germany 1964-82.
EQ.
E s t i m a t i o n r e s u l t sModel type
no.
TimeGNP
P r i c e3 D-W
a ) Total paper and board
1. Static model, nominal
-
1.369 -0.152 0.958 1.909price (0.1 19)*** (0.039)***
2. Static model, constant
-
1.102 -0.225 0.951 1.908price, US deflator (0.071)*** (0.069)***
3. Static model, constant 1.291 -0.227 0.959 2.040 price, German deflator (0.099)*** (0.057)***
4. Static model, constant
-
0.813 -0.176 0.927 1.608price, US deflator (O.llO)*** (0.122)
adjusted for exchange rates
5. Static model, constant
-
0.977 -0.399 0.953 2.364 price, German deflator (0.051)*** (0.114)***adjusted for exchange rates
7. Dynamic model, constant -0.022 1.779 -0.235 0.972 2.161 price, US deflator (0.006)*** (0.196)*** (0.052)***
8. Dynamic model, constant -0.029 1.599 -0.229 0.959 2.083 price, US deflator (0.008)*** (0.229)*** (0.097)*
adjusted for exchange rates b) Newsprint
9 . Static model, nomind
-
1.643 -0.159 0.976 2.284price (0.092)*** (0.028)***
10. Static model, constant
-
1.369 -0.220 0.967 1.889price, US deflator (0.067)*** (0.052)***
11. Static model, constant
-
1.515 -0.198 0.972 2.184 price, German defiator (0.082)*** (0.040)***12. Static model, constant
-
1.016 -0.275 0.953 1.450price, US deflator (0.090)*** (0.099)**
adjusted for exchange rates
13. Static model, constant
-
1.242 -0.274 0.965 1.915. price. German deflator (0.056)*** (0.069)***
adjusted for exchange rates
15. Dynamic model, constant -0.022 2.000 -0.179 0.979 2.486 price, US deflator (0.007)*** (0.211)*** (0.044)***
16. Dynamic model, constant -0.025 1.815 -0.209 0.968 1.805 price, US deflator (0.009)** (0.284)
***
(0.085)**adjusted for exchange rates
Notes: The figures in t h e parenthesis are standard e r r o r s of t h e coefficients.
+**
+*, and indicate coefficients significantly differ from zero a t the 0.99, 0.95, and 0.90 confidence levels, respectively. is the adjusted coefficient of determination and D-W t h e computed Durbin-Watson statis- tic.
4. SOME EXPEFUENCES
FRDM WORK DONE AT A CONSULTING COMPANY
4.1 Background
Until t h e mid-1970s, t h e r e was little interest in studying t h e effects of price c h a n g e s on paper and board consumption because:
-
Real prices did not change m u c h from t h e early 1950s up t o 1973.-
Differently from many o t h e r industrial products, paper and board do n o t have suitable or acceptable substitutes a t a cheap price (FA0 1977a).-
Paper and board a r e complementary products whose s h a r e of t h e total price of products t o which they a r e r e l a t e d i s very small and, therefore, even large price increases n~ould not m u c h affect t h e i r consumption (USDA 1973).-
Price variables include s h o r t - t e r m variation which c a n reduce even t h e coefficient of determination in long-term consumption models (FA0 1960).-
The consumption of paper a n d board is habitual and t h u s the effects of price movements a r e weak ( b e r g 1968).-
The u s e of price a s an exogenous variable in a practical fore- casting situation would r e q u i r e t h a t reliable price forecasts a r e available; forecasting price developments may be even m o r e difficult t h a n forecasting f u t u r e demand levels.Rapid i n c r e a s e s in real prices of paper and board i n 1973-1977 caused a growing i n t e r e s t i n studying t h e i r effects on consumption, which led t o studies (FA0 1977a, Buongiorno 1978) in which price param- e t e r s had values t h a t significantly differed from zero. However, in t h e l a t e 1970s r e a l prices began t o decline again, more or less r e t u r n i n g t o t h e s a m e development path where they were before t h e oil crisis (see Figure 1). Therefore, it was decided by Jaakko Poyry Companies t o sug- gest a s t u d y (Suhonen 1984) of t h e importance of price variables in con- sumption models by using d a t a material t h a t also included t h e l a t e s t price developments.
4.2
Data
a n d Models UsedThe models used were of type (3) and (4). By using a stepwise regression procedure, i t was possible to study t h e effects of t h e introduc- tion of a lagged endogenous variable on t h e statistical properties of t h e consumption model. Separate models were constructed for newsprint, printing a n d writing papers, wrapping and packaging papers, a s well a s for sack paper. Substitute prices used were printing and writing paper prices for newsprint and vice versa, and low density polyethylene (LDPE) prices for wrapping and packaging papers and sack paper.
The d a t a material included 40 countries and t h e years 1965-80 for newsprint a n d printing and writing papers, and 14 c o u n t r i e s and t h e years 1970-80 for m a p p i n g and packaging paper a n d s a c k paper. In t h e l a t t e r case t h e n u m b e r of countries was limited by t h e availability of relevant price series for
LDPE.
Twenty-three of t h e 40 c o u n t r i e s weredeveloped countries and 17 newly industrialized or developing c o u n t r i e s . Data were collected mostly from Jaakko Poyry's Forest P r o d u c t s Market a n d Price Data Banks, FA0 Yearbooks, and
LDPE
s e r i e s from several pro- fessional publications, national statistics, and earlier studies (e.g.,UNIDO
1981, SRI 1981). GDP figures were deflated by national GDP deflators and converted i n t o US dollars by using t h e 1970 exchange r a t e . P r i c e s were deflated by t h e US wholesale price index. No a d j u s t m e n t s for fluctuating exchange r a t e s were made.
All models were estimated from pooled cross-sectional time-series d a t a t o broaden t h e variation range of observations. A dummy variable was s e t for each country t o absorb t h e specific variation between coun- t r i e s n o t explained by differences in income, prices, and prices of substi- t u t e s in p a s t consumption. Therefore, t h e e s t i m a t e d model was:
log C,
=
a+
b logCi,f +
c logYit +
d log Pi2k -1
+
e log PIit+
zi Dik+
E'k =l
where k r e f e r s to t h e number of countries, for every product j in ques- tion. The numerical values of dummy variables tell how m u c h a c o u n t r y deviates from t h e average level of estimates calculated from t h e pooled data.
Additionally, t h e data material was grouped, accordmg t o t h e level of GDP per capita in 1980 and t h e period of observation, into c l u s t e r s t o analyze whether elasticities varied systematically between different income classes and over time. A s e p a r a t e consumption model was built for t h e
EEC
region, which is composed of countries with a relatively simi- l a r economic and cultural background.4.3 Statistical Tests Used
In addition to t h e traditional calculation of adjusted coefficients of p a r a m e t e r s and Student's t-statistic, some o t h e r statistical t e s t s were also made. Autocorrelation of residuals were t e s t e d using t h e Durbin- Watson s t a t i s t i c calculated both from t h e original country-wise d a t a a n d d a t a s o r t e d into a rising o r d e r according to t h e GDP per capita. Hornos- cedasticity was tested with t h e help of t h e Goldfeld-Quandt t e s t based on splitting one regression residual i n t o two subsamples, one with low and . t h e o t h e r with high values of explanatory variable, and t h e n calculating a n F-ratio of t h e two variances (see Goldfeld a n d Quandt 1965). Since t h e consumption models included a lagged end.ogenous variable, special a t t e n t i o n was paid t o detecting t h e consequences of possible multicol- linearity in them. The approach adapted h e r e was Frisch's Confluence Analysis Frisch 1934, see also Koutsoyiannis 1973) based on t h e value
-L
r z i z j : s .
R
a n d a stepwise p r o c e d u r e t o study t h e effects of new variables on t h e values of t h e estimates of p a r a m e t e r s a n d standard e r r o r s intro- duced first in the consumption models. The importance of differences between different data c l u s t e r s were t e s t e d by using analysis of variance.4.4 Estimation Results 4.4.1 Newsprint
The results of t h e regression analyses and statistical t e s t s for news- print are presented in Appendix 3.
I t
should be noted t h a t the coefficients Y t , Pt, and PIf refer to t h e short-term elasticities. Based on Nerlove's partial adjustment theory (Nerlove 1958) it was possible t o cal- culate t h e long-term elasticities presented in Table 2. The coefficient of adjustment (A) measures the velocity of adjustment. Although i t is impossible to calculate the number of periods required for a complete adjustment ( t h e function used is asymptotic with regard t o time), i t is possible t o choose some arbitrary percentage of adjustment and calcu- late t h e number of periods required t o reach it, or t o calculate t h e per- centage of adjustment after a certain number of periods. According t o Nerlove and Adhson (1958), the n u m b e r of periods(N)
required for adjustment t o within, e.g., 5% of t h e long-term equilibrium level may be determined by t h e formulawhere A i s t h e coefficient of adjustment (or 1 minus t h e coefficient of t h e Cf-, variable) and N the number of periods required. This method was l a t e r been used by, among others, Yadav (1975) for calculating elastici- t i e s for the following periods.
TABLE 2. Estimated long-term elasticities for newsprint.
Long-term elasticities
Coefficient Percentageof Data group
Income Own- Cross- of adjustment adjustment
price price (A) after 3 years
- % - All data 1965-80 +0.84 -0.30 +0.06 0.64 9 5
Years 1965-72 +l.13 -0.18 +0.03 0.77 9 9
Years1973-80 +1.04 -0.17 +0.11 0.84 100
GDPpercapita* +1.02 -0.28 +0.09
< USD
1000 0.58GDP
per capita +1.19 -0.73 +0.02USD
1001-3000 0.36GDP
per capita +0.59 -0.04 +0.01> USD
3000EEC region +0.75 -0.45 +0.36 0.57 9 2
GDP per capita in year 1980, measured in US dollars at 1970 prices.
The estimated p a r a m e t e r s for Ct-l and Yt &ffered in all c a s e s significantly from zero even a t the 0.99 confidence level. Own price- variable was i m p o r t a n t a t t h e 0.90 confidence level in all o t h e r c a s e s except for t h e income g r o u p GDP p e r capita
>
USD 3000, whereas cross- price elasticity was i m p o r t a n t only for t h e years 1973-80 and t h e EEC region (see Appendix 3).In g e n e r a l , income elasticity tended t o decrease over t i m e and with increasing GDP per capita, although t h e highest income elasticity was measured in t h e group CDP per capita USD 1001-3000. Own-price elasti- cities varied between -0.04 (richest countries) and -0.73 (medium- income countries). The l a t t e r figure depends heavily on t h e low coefficient of a d j u s t m e n t (short-term elasticity only -0.26), which indi- cates a strong dependency on past consumption (or t h e habitual n a t u r e of consumption) r a t h e r than a high price sensitivity. The EEC region.
which predominantly imports newsprint, s e e m s t o b e m o r e price- sensitive t h a n t h e average of countries.
The choice of t h e period for estimation affected t h e n u m e r i c a l values of estimates. Over t h e e n t i r e period 1965-80 t h e price variable appears t o have varied m o r e t h a n in 1965-72 or 1973-80. Correspond- ingly, t h e value of income elasticity for the complete d a t a s e t was lower.
4.4.2 Printmg and writing papers
The r e s u l t s for printing and writing papers a r e p r e s e n t e d i n Appen- dix 4, a n d t h e corresponding long-term elasticities in Table 3. The price variable u s e d t o r e p r e s e n t prices of substitutes f o r printing a n d writing papers was t h e price of newsprint.
TBBLE 3. Estimated long-term elasticities for printing and writing papers.
Long-term elasticities
Coefficient P e r c e n t a g e of -
Data group
Own- Cross- of a d j u s t m e n t a d j u s t m e n t Income
price price (A) a f t e r 3 y e a r s All d a t a 1965-80
+
1.52 -0.00 -0.14 0.61Years 1965-72 +1.25 -0.24 +0.53 0.62 Years 1973-80 +1.31 +0.16 -0.27 0.79
GDP <
per capita+
1.47 +0.08 -0.15USD 1000 0.62
GDP
per capita +1.42 -0.21 +0.08USD 1001-3000 0.52
GDP
per capita + 1-43 -0.01 -0.1B>
USD 3000 0.87EEC
region +1.56 +0.02 -0.11 0.75 98Again, t h e estimated parameters for lagged consumption and income were important in all t h e models. In contrast n i t h this, own price and substitute price received unexpected signs in several equa- tions, and when they had t h e expected signs, the coefficients differed significantly from zero in only one case (years 1965-72, s e e Appendix 4).
One reason for this inelasticity may be found in the price series used.
The breakthrough of coated printing paper grades which have higher prices made t h e internal grade s t r u c t u r e of printing and writing papers in t h e 1970s more heterogeneous. The average price, and also t h e rela- tive price, of printing and writing papers increased; meanwhile t h e con- sumption grew strongly due to advantageous developments in end uses and production technology (JP 1984). Additionally, t h e use of newsprint price as substitute for printing and writing paper was a somewhat arbi- t r a r y choice made in t h e lack of better variables. Actually, newsprint and other printing papers compete strongly only in a few end uses.
The income elasticities measured were notably higher than for newsprint (between 1.25 and 1.56 compared n i t h 0.59 t o 1.19 for nenrs- print). There was no clear tendency in respect of GDP per capita. or time;
income elasticities were in most cases between 1.3 and 1.5, t h e highest numerical value being for the
EEC
region.Since the values of own-price and cross-price elasticities were largely meaningless, t h e effects of multicollinearity were cautiously t e s t e d With t h e help of Frisch's Confluence Analysis and a stepwrise regression procedure i t was noticed t h a t multicollinearity did n o t aflect t h e regressions; price elasticities did not change m u c h regardless of t h e set of variables used in t h e regressions.
4.4.3 Wrapping and packaging papers and boards
The data material for wrapping and packaging papers and sack paper was more concise t h a n for newsprint and printing and writing paper, con- sisting of 14 developed countries and 11 years. The estimation results a r e presented in Appendix 5 and the long-term elasticities in Table 4.
Lagged consumption and income were important variables i n explaining consumption except for the highest income group, whereas own-price and substitute-price variables were not important a t all, hav- ing either wrong signs or too large a standard e r r o r (see Appendix 5).
The almost complete inelasticity of consumption t o price changes m a y be a consequence of t h e low value of packaging m a t e r i a l s in relation t o t h e total value of t h e final products t o be packed, or t h e total costs of t h e whole distribution system. No single variable could be found which unambiguously explained the complete substitution mechanism; prices of packaging materials may be of minor importance compared with, e.g., traditional packaging and/or t h e distribution system replaced by a new alternative system.
The use of t h e price of
LDPE
as substitute price also has some disad- vantages. First, t h e substitution effects of LDPE are limited t o a number of products (mainly wrapping papers) only. In t h e case of cartonboards, LDPE can even be considered as a complementary product, because boards used for packaging liquids a r e normally coated with LDPE.TABLE 4. Estimated long-term elasticities lor wrapping and packaging papers and boards.
Long-term elasticities
Percentage of Data group
Own- Cross- of adjustment adjustment Income price
price
(A)
after 3 years-
0. /4-
All data 1970-80 +0.62 -0.10 -0.03 0.71 98
Years 1970-75 +0.93 -0.03 -0.16 0.53 90
Years 1976-80 +0.82 +0.19 -0.09 0.65 96
GDP
per capita +0.28 -0.02 +0.05>
USD 3000 0.8 1EEC region +0.62 -0.07 -0.18 0.71 98
Second, t h e technological development in the manufacture of shrink foils and other films from LDPE was not taken into account when con- structing price series. Today it is possible to produce many times more plastic film from t h e same quantity of LDPE than in t h e early 1970s (Vol- p e r t 1982).
Income elasticity for wrapping and packaging papers a n d boards behaved a s expected; i t decreased with increasing GDP per capita and also over time. Compared with cultural papers, the income elasticities were even somewhat lower than those measured for newsprint.
4.4.4
Sack
paperThe results of regressions for sack paper a r e shown in Appenbx 6 a n d t h e corresponding long-term elasticities in Table 5. It can be seen t h a t in t h i s case short-term a n d long-term elasticities were quite similar due to the high values of t h e coefficient of adjustment.
In most cases t h e estimates of parameters for lagged consumption were significantly different from zero a t t h e 0.90 confidence level. The price of LDPE h a d a positive sign in all equations, but t h e standard errors were too large to make i t
an
important explanatory variable. Income elasticities for t h e highest income group and t h e EEC region were nega- tive, indicating t h a t sack paper has already passed t h e saturation phase of its life cycle and is t h u s a n inferior commodity on high income levels.Price elasticities varied between -0.26 and -0.56, which m e a n s t h a t sack paper was t h e most price-sensitive of the four paper grades investi- gated here. It was also the bulkiest product of those studied here. Simi- larly t o newsprint, t h e
EEC
region is a major importer of sack paper, andit
accounted for the most negative value.TABLE 5 . Estimated long-term elasticities for sack paper.
Data group
Long-term elasticities
Coefficient Percentage of Own- Cross- of adjustment adjustment lncome
price price
(A)
after 3 years - % --~ - - -
All data 1970-80 +0.12 -0.46 +0.21 0.61 94
Years 1970-75 +0.60 -0.36 +0.15 0.80 99
Years 1976-80 +0.18 -0.26 +0.15 0.67 96
G D P p e r c a p i t a +0.69 -0.53 +0.02
< USD
3000 0.64 95GDP
per capita -0.07 -0.26 +0.15> USD
3000 0.78 9 9EEC
region -0.37 -0.56 +0.13 0.96 1005. COMPARISON OF
RESULTS
WITH OTHER SI'UDIES5.1 Background
There a r e n o t very many international studies dealing with elasti- city measurement from paper consumption models. The studies referred to here include FA0 (1960), FA0 (1977a), Buongiorno (1978), and Wibe (1984), whose results a r e compared with the results of Suhonen (1984) discussed earlier in this paper. It should be remembered t h a t s t r i c t comparisons between different studies may sometimes be confusing because of different composition of countries, different deflation methods and exchange r a t e manipulations, different time periods for observations, a s well as diflerent explanatory variables used i n t h e con- sumption models. From t h e Global Trade Model's point of view it is, how- ever, important to compile t h e existing results from earlier works t o be used as references establishing relationships between c o n s u m p tion, prices, substitute prices, income, and o h e r possible demand shifters for t h e final
GTM.
5.2 lncome Elasticities
Income elasticities had been measured already in an early work by FA0 (1960), where a clear tendency of decreasing income elasticities with increasing GDP/GNP per capita was found. Elasticities were measured by using time-series data of both individual countries and cross-sectional comparisons between countries from a model, where income per capita was the only explanatory variable. in a later study by
FA0
(1977a) i t was noted that for cultural papers the highest income elasticities were t o be found in high and medium income countries, whereas for industrial paper grades they were found in the low and very low income countries.When price effects and literacy level were taken into account in t h e m.ode1, the income elasticities measured for developing countries drasti- cally decreased.
A s u m m a r y of income elasticities m e a s u r e d in different s t u d i e s is presented in Table 6. For comparison. all
GNP/GDP
per capita figures were converted into c o n s t a n t US$ 1975 prices. It should be noted t h a t t h e results from Buongiorno (1978) and Suhonen (1984) refer to long- t e r m elasticities, whereas t h e income elasticities of other s t u h e s a r e t o be interpreted as short-term elasticities.TABLE 6. Comparison of income elasticities measured iri different paper con- sumption studies.
Income elasticity
- -
Explanatory
Study/lncome variables Time ru'ews- Printing and Other paper and Group included i n period p r i n t writing papers board /packaging
the model p a p e ~ and board
FA0 (1960)
&
GEiP per capite*
> USS 3000 US0 1500-3000 US0 750-1500
< USS 750
FAO (]!ma)
?
Jt GDP per capita> USS 3000 USt1500-3000 US$ 700- 1500
< US6 700
Buongiorno (1978) q , c t - l , GDP per capita Pt ,PIt
Average (for culture2
> USS2600 pepers only)
c USS 2600
JP/Suhonen (1884)
? .
CtGDP per capita Pt .PIt
Average
> USt 4750 U S 1800-4750 C US6 1600
Wibe (1884)
%
,pt *GDP per capita Time Aver y e
> USS2500
US6600-2500
c USS 600
before 1960
0.4-0.8 0.B-1.1 1.1-1.5 1.5-2.9 1963-73
1.0 0.8 1.1 1965-80
(cultural papers) 0.8 1970-80 0.6 (other 1.2 grades) 1.0
1970-79 r
at US8 1975 prices
For newsprint and industrial grades, t h e r e seems to be a falling trend in income elasticity with rising income, whereas in t h e case of printing a n d writing papers t h e highest elasticities seem to be in medium a n d high income countries. It is also important t o include price and/or other variables in t h e models for developing countries, otherwise the income variable absorbs variation which in reality does n o t belong t o it. These variables include t h e literacy rate used by FA0 (19??a), and a supply availability index, which was found to be an important explana- tory variable first by Gregory (1966) and then in a later study by Uutela (1979).
There have also been discussions a s t o whether the developing coun- tries will follow t h e same per capita consumption patterns a s industrial- ized countries with increasing income. In another study Wibe (1983) argues t h a t the developing countries do not follow the path set by t h e already industrialized nations, but have a lower consumption of paper products because of t h e availability of "new" technologies (e.g., radio, television, plastics, etc.). This was also t h e a priori expectation of a study by Uutela (1979). However, t h e results were not as expected, but showed instead t h a t t h e countries reaching a certain
GDP
per capita level in t h e 1970s consumed more paper and board per capita t h a n those countries t h a t had reached t h e s a m eGDP
level in the 1950s or 1960s.The interpretation of t h i s surprising result may be t h a t although t h e r e a r e nowadays more substitutes for paper and board than, e.g., in t h e 1950s, t h e r e a r e , on t h e o t h e r hand, also many more end-use applica- tions (and industries) for paper and board (e.g., computer print-outs, consumer packages).
A conclusion of the different elasticities presented in Table 6 and other experiences gained from practical work is given in Table 7. The assessment is partly subjective a n d applies only to countries with aver- age economic, cultural, social, etc. conditions. In extreme cases t h e values of elasticities may considerably differ from those i n Table 7. How- ever, they may give t h e r e a d e r some indication of the magnitude of income elasticities in different product groups and income classes.
5.3 Own-Price Elasticities
Price elasticities t e n d t o vary quite a lot depending on t h e product, country group, time period for m e a s u r e m e n t and, a s was shown in Sec- tion 3, t h e way the price variable is valued. Therefore, t h e price elasti- city estimates compiled in Table 8 m u s t be interpreted very carefully.
FA0
(1977a) and Buongiorno (1978) concluded that in low-income coun- tries, which normally also import most of their paper products, t h e con- sumption is seriously affected by price increases. Suhonen (1984) found only very small correlations with income levels; t h e price variable was not significant a t all for printing and writing papers and wrapping and packaging papers, a n d only slightly significant for newsprint and sack paper.Any conclusions as t o t h e numerical level of price elasticities of Table 8 a r e difficult to make. However, newsprint seems to be t h e m o s t price-sensitive and printing and writing paper may be t h e less price-
'fbgLE 7 . Conclusion of the numerical values of income elasticities based on ear- lier studies and practical experiences from forecasting work.
lncome Newsprint Printing and Other Paper and
group writing papers board/packaging
paper and board
High income 0.4-0.8 1.0-1.5 0.3-1.2
(GDP per capita
>
US$3000)Medium income 0.6-1.2 1.2-1.8 1.0-1.6
(GDP per capita USb1500-3000)
Lou income 0.7-1.5 0.7-1.5 1.4-2.0
(GDP per capita
<
US$ 1500)TdBLE 8. Comparison of own-price elasticities measured in different paper con- sumption studies.
Study/lncome Group
FA0 (lone) GNP per capita
> USS 9000
U S 1500-3000
USZ 750-1500
< USS 750
Own-price elasticity Explanatory
variables Time News- Printing m d Other paper and
included in period print mi- papers board /packaging
t h e model paper and board
yt opt 1963-73
Buongiorno(lQ78) Yt.Ct-l, 1963--73 GDP per capita Pt ,PIt
Average (for culturd -0.7 -0.5 -0.7
> USS 2800 papem only) -0.6 -0.2 -0.3
< USS 2800 -0.B -0.7 -0.8
IP/Suhonen (lo&) Yt,Ct-l, 1965-80 GDP per capita Pt ,PIt (cul t u r d
Average papers) -0.3 -0.0 -0.1
> USS 4750 1970-80 -0.3 +O. 1 -0.1
USS160C-4750 (other -0.7 -0.2 -0.0
< USs 1600 grades) -0.0 -0.0
. .
Wibe (1QEM) yt
3
1970-79GDP per capita Time
Average -1.1 -0.8 -0.9
> USS 2500 -2.6 -0.4 -1.3
USS 600-2500 -0.7 -0.5 -0.3
< USS 600 -0.6 -1.1 -1.4
sensitive product. The studies by
FA0
(1977a) and Suhonen (1984). where the same deflation method but different time periods were used, pro- duced results of largely the same magnitude. Wibe's (1984) study prices resulted in the most negative price elasticities.5.4 Cross-Price Elasticities
The results from the few paper consumption studies which include substitute prices as an explanatory variable a r e not very encouraging.
In FAO's (1977a) study even the sign of the cross-price elasticity was, against t h e a priori expectation. negative in most equations. Newsprint accounts for t h e best results; for industrial grades, there is only one international study known to the author t h a t deals with substitute prices. The results from three studies a r e presented in Table 9. The prices of substitutes used in all three studies were printing and writing paper price for newsprint and vice versa, and
LDPE
price for packaging paper a n d board in t h e study by Suhonen (1984).The poor results a r e partly explained by t h e choice of substitute variables. As discussed earlier in this paper, prices of different materials a s such m a y not be decisive factors for buying or consumption decisions for paper and board; there are many intervening variables such as labor intensity and costs, flexibility of use, or product performance, which together determine the ranks between different alternatives.
I t
is a question of s y s t e m s u b s t i t u h n rather than product o r price s u b s f i t u - tion. For this reason, the traditional price theory may not be able t o explain paper and board consumption.The effects of real substitutes for paper and board, such a s new elec- tronic information media or plastic-based packaging systems, a r e extremely difficult to quantify and include a s explanatory variables in consumption models. There a r e hardly any statistics which could meas- ure these relationships. Even t h e national input-output statistics nor- mally have too rough a classification of products and industries for paper and board substitution analysis. The best applicable method might be a market research approach; it would require product by product a thorough analysis of the most important end-use sectors to understand their decision-making patterns and buying practices. This would not be possible without extensive field work based on interviews and deep dis- cussions with people in the relevant branches.
TABLE 9. Comparison of cross-price elasticities measured in different p a p e r con- sumption s t u b e s .
Cross-price elasticity Explanatory
Study/lncome variables Time News- Printing and Other paper a n d Group included in period print writing papers boerd/packaging
t h e model paper and board
FA0 (1 Dna)
5
n P t 1963-73 GNP per capite PIt,> USt3000 literacy r a t e +O. 1 -0.6
. .
USL 150&3000 -0.0 -0.3
. .
USL 75&1500 +0.3 -0.6
. .
< USS750
+o.
1 -0.0. .
Baongiorno (1878)
5 , ct ,,
, 1963-73GDP per capite Pt ,PIt
Average
+o.
1+
0.3. .
> USS 2600
+o.o +
0.2. .
c ~ ~ $ 2 6 0 0 +0.4
+
0.3. .
J'P/Suhonen (IDB4)
5 ,
Ct 1965-80 GDP per capita Pt ,PIt (culturalAverage papers) + O . 1 -0.1
+o.o
> USt4750 1970-80 +O.O -0.2 +O. 1
USS 1600-4750 (other +0.0 +0.1 -0.3
c USS 1600 grades) +O.1 -0.2
. .
6. CONCLUSIONS
The basic aim of this paper was to compile some practical results from income and price elasticity measurement. The following conclu- sions can be drawn:
-
The numerical values of income and price elasticities are sensi- tive t o tbe explanatory variables included, deflation methods and exchange r a t e t r e a t m e n t used for converting income and price variables into constant prices and a common currency, a s well a s t h e t i m e period for observations in international models. So far. too little attention bas been paid t o these data manipulations before undertaking t h e numerical estimation.-
Short-term and long-term elasticities should be distinguised;the consumption of many paper and board grades is habitual and does n o t r e a c t to changes in income and/or prices immedi- ately, i.e., within one year. Thus the use of dynamic models (e.g., t h e partial adjustment model) for measuring elasticities would be preferable.
-
The velocity of adjustment t o changes in income and prices is faster in high income countries than in low income countries.The adjustment processes seem also t o accelerate with time, which may be an indication of growing flexibility because of increasing supply of commodities and tightening competition i n a society.
-
The use of country-specific dummy variables in pooled cross- sectional a n d time-series models is essential; otherwise t h e measured price and income elasticities may absorb some of t h e variations caused by other variables omitted from t h e regres- sion, which results in serial correlation of t h e regression resi- duals; i.e., the numerical values of elasticities will be meaning- less. In some earlier studies, the statistical properties of con- sumption models a r e not discussed in detail, leaving t h e reader dubious as to t h e validity of the results.-
Elasticities t e n d t o change with income levels a n d over time.The use of constant elasticities for a period longer t h a n 10 years may lead t o unrealistic forecasts.