TRANSPORT RESEARCH LABORATORY TITLE by Car ownership forecasts for low-income countries K Button, J L Nine and N Ngoe Overseas Centre Transport Research Laboratory Crowthorne Berkshire United Kingdom XA -31 BUTTON, K, J HINE and N NGOE, 1992. Car ownership forecasts for low-income countries. Traffic Engineering and Can trol, 33 (12), 666-7 1. Car ownership forecasts for low-income countries by Kenneth Button, Professor of Applied Economics and Transport, Loughborough University John Hine, Principal Scientific Officer, Transport Research Laboratory and Ndoh Ngoe, Research Associate, Department of Economics, Loughborough University Introduction. Considerable efforts have been expanded over the years to develop sophisticated forecasting models of car own- ership and use in developing countries' Efforts to look at the probable growth pat- terns of vehicle ownership in the very low in- come countries of the World have been much more limited. Traffic is. however, growing in many of the World's poorer nations, albeit at very variable rates, and this has implications for both the countries themselves and, via potential environmental implications, for the industrialised world. The implications for nations where growth is taking place are being felt not only in terms of pressures on the national road networks, but also through higher import bills for both vehicles and fuel 2.This growth in vehicle ownership is also continuing hand-in-hand with rapid urbanisation, and the strains on many nations' transport infrastructure are particularly severe in cities 3.Rising vehicle fleets also impose pressures on the vehicle maintenance facilities available and the ad- ministrative and planning structure required to police and regulate the road system. Since there are similar growth trends in commercial vehicle numbers and use, questions of road maintenance are compounded. More gener- ally, there are questions of social equity and mobility associated with rising levels of per- sonal car ownership and availability. It is important, therefore, for these sorts of reason that both those involved in transport infrastructure planning and those responsible for the development of tractable land-use and economic policies, more generally, have reasonable car ownership and use forecasts together with robust forecasts of future commercial vehicle levels. Further, the major aid-giving agencies, such as the World Bank and Asian Development Bank, in order to allocate funds over time find it useful to know whether similar trends in vehicle ownership patterns occur across a range of low-income countries or whether wide divergences exist. On a broader front, transport is a major contributor to many of the emissions which are associated with both trans-border envi- ronmental damage (e.g. with acid rain) and with global warming. International responses to what are perceived to be mounting prob- lems have tended to focus on the appropriate response in the industrialised countries (e.g. the fitting of catalytic converters to limit acid rain-associated emissions), but the parame- ters of these calculations are inevitably de- pendent, especially regarding the global warming issue, on the worldwide use of fossil fuels which, in turn, cannot be isolated from transport trends and developments in Third World nations. The objective here is to look at car owner- ship in low-income countries and to seek to develop a forecasting model which offers some guidance as to what we might expect to happen in these countries into the next cen- tury. There is no pretence that the model used is by any stretch of the imagination sophisti- cated by modern econometric standards, but rather it seeks to be fairly-general and robust with the minimum number of independent variables which themselves must be pro- jected. The current situation What one means by 'low income' is to a large extent arbitrary, but in this study we define low-income countries as countries which en- joyed per capita incomes of US$3 000 or less in 1986. We also exclude, for practical rea- sons of comparability, small island states which because of their geography tend to have individualistic policies and patterns of car ownership. The available data, even those reported to the major international aid agen- cies, on vehicle ownership aicross low- income countries are, to put it mildly, ex- tremely variable in their quality. In some cases they simply do not exist in a usable form which further reduces the number of countries examined or the observations for particular years are missing. The main data employed here draw upon published sources for various years supplied in publications of the World Bank, the Inter- national Monetary Fund, the United King- dom Society of Motor Manufacturers and Traders, the United States Department of En- ergy, and the International Road Federation. This is supplemented by data drawn from a variety of specific studies of trends in trans- port and the economies of low-income coun- tries. The full list of countries in the forecast- ing exercise are set out in Table I together with details of per capita income in 1986 in $US, per capita car ownership levels and total commercial vehicle parks. Much of the data used to develop the fore- casting models relate to individual time series for each country dating back as far as 1967. Previous studies in the various fields of eco- nomic development indicate that there are significant differences in the nature of the low-income countries' economies and, there- fore, a number of models subdividing the data-set are explored in the subsequent analy- sis'. This analysis itself also provides support for the stratification adopted. The countries examined are grouped into five categories which are themselves based upon car owner- ship rates and income levels in 1986. These are:- A: less than 0.002 cars per person; - B: less than 0.01 cars per person but more than 0.002, and a per capita GNP of less than $450; - C: less than 0.01 cars per person but more than 0.002, and a per capita GNP of more than $450; - D: less than 0.02 cars per person but more than 0.01I; and - E: more than 0.02 cars per person. Inspection of the raw time series data for the various countries indicates that countries in categories A and B, those with very low cur- rent levels of car ownership and also low in- comes, have experienced very limited growth in their car ownership per capita since the mid-I 960s. In contrast, those countries which already had a somewhat higher level of per capita car ownership, and generally income, have also experienced a relatively rapid in- crease in the level of their per capita owner- ship.This simple observation has implications for any forecasting model specification as well as highlighting the developments which are occurring in those lo.W-income countries where income is growing more rapidly. The implications for the transport systems and the physical environments of countries in groups C, D and E is compounded when it is remem- bered that not only are their per capita car ownership rates increasing rapidly, but in ad- dition their populations and cities are usually growing very rapidly as well. The forecasting framework Car ownership is normally seen as the main determinant of traffic growth in most low-in- come countries, although from an inter-urban road planning perspective freight transport can also be important. While the total car park is the key transport planning in this context, for technical reasons it is not the normal basis of forecasting models. Per capita car owner- ship is used as the dependent variable in the subsequent analysis both to permit easy com- parisons with other studies where helpful, and also to limit the statistical problems of heteroskedasticity which can arise when em- ploying data from countries of widely differ- ing populations. For car-park forecasts, per capita projections can easily be grossed 666 ~~~~~~~~~~~~~~~~~~~TRAFFIC ENGINEERING + CONTROL 666 Table I. Country Population (M) GDP (SDR) per head GNP per head Cars per head Commercial vehicles 1986 1980* 1986 ($U.S.) 1986 ('000) 1986 Ethiopia 43.5 84 120 0.00082 19.0 Burkina Faso 8.1 154 150 0.00198 14.0 Bangladesh 103.2 108 160 0.00048 35.0 Malawi 7.4 158 160 0.00189 17.5 Zaire 31.7 171 160 0.00284 80.0 Mali 7.6 180 0.00197 7.5 Burma 38.0 135 200 0.00163 45.0 Madagascar 10.6 269 230 0.00330 45.0 Uganda 15.2 1031 230 0.00211 15.0 Burundi 4.8 .181 240 0.00202 7.5 Tanzania 23.0 217 250 0.00202 50.0 Togo 3.1 325 250 0.00726 12.0 Niger 6.6 346 260 0.00364 22.0 Benin 4.2 251 270 0.00417 12.0 India 781.4 199 290 0.00154 1000.0 Rwanda 6.2 177 290 0.00119 11.0 Kenya 21.2 343 300 0.00604 110.0 Zambia 6.9 513 300 0.01377 50.0 Sierra Leone 3.8 287 310 0.00612 11.0 Haiti 6.1 225 330 0.00492 14.7 Pakistan 99.2 225 350 0.00489 183.2 Ghana 13.2 1059 390 0.00417 45.0 Sri Lanka 16.1 197 400 0.00932 132.0 Senegal 6.8 382 420 0.01264 23.8 Liberia 2.3 465 460 0.00469 6.2 Indonesia 166.4 388 409 0.00628 1133.3 Philippines 57.3 565 562.062502.0 Morocco 22.5 633 590 0.0242220. Zimbabwe 8.7 599 620 0.03046 80.0 Dominican Republic 6.6 956 710 0.01591 66.0 Papua New Guinea 3.4 703 720 0.00551 30.5 C6te d'lvoire 10.7 940 730 0.01822 90.0 Honduras 4.5 541 740 0.00862 52.7 Thailand 52.6 560 810 0.01088 824.0 El Salvador 4.9 560 820 0.01990 65.0 Botswana 1.1 914 840 0.01493 24.5 Cameroon 10.5 561 910 0.00827 65.0 Paraguay 3.8 1107 1000 0.01608 30.2 Peru 19.8 653 1090 0.01964 214.9 Turkey 51.5 868 1110 0.01910 553.1 Tunisia 7.3 1029 1140 0.03714 159.0 Mauritius 1.0 935 1200 0.03512 12.0 Colombia 29.0 939 1230 0.02899 391.4 Chile 12.2 1941 1320 0.05230 263.0 Costa Rica 2.6 1684 1480 0.03269 68.0 Jordan 3.6 869 1540 0.04414 57.0 Syrian Arab Republic 10.8 1177 1570 0.01019 160.0 Brazil 138.4 1303 1810 0.07225 2100.0 Malaysia 16.1 1367 1830 0.07491 310.0 South Africa 32.3 2295 1850 0.09402 1179.9 Mexico 80.2 2078 1860 0.06484 2250.0 Uruguay 3.0 2129 1900 0.06667 100.0 Hungary 10.6 1639 2020 0.14518 189.1 Portugal 10.2 1879 2070 0.12118 525.0 Yugoslavia 23.3 1864 2250 0.12498 301.2 Panama 2.2 1469 2300 0.06364 43.6 Argentina 31.0 4087 2330 0.12574 1434.7 South Korea 41.5 1182 2350 0.01601 627.2 Algeria 22.4 1476 2370 0.03237 458.0 Gabon 1.2 2962 2590 0.01496 17.5 Greece 10.0 2991 3080 0.13011 625.0 'The 1980 GODPfigures are expressed in the international unit of account, the.Special Drawing Right (SDR), used by the International Monetary Fund to reduce the effects of floating exchange rates in international transactions. In December 1980 1 SDR was equivalent to US$1.28 or £0.53. In December 1986 1 SDR was equivalent to U.S$1 .22 or £0.83. up by applying forecasts of demographic car ownership levels across various low- from the categorised data discussed above. changes. income countries, and also at differences be- A variety of possible sigmoid-shaped Developing a forecasting framework for tween them and industrialised nations, has functions could be applied to meet the gen- car ownership requires an initial analysis of been conducted in the past using linear and eral requirements of the model. The main the influences which affect ownership. The log-linear specifications 5.This type of speci- modelling thrust employed here uses the ag- slow growth in car ownership in the lowest- fication, though, tends to ignore the ultimate gregate quasi-logistic approach which has in income countries observed from the base saturation level towards which, following the past been used for both local and national data, coupled with the much more rapid rises standard product-life cycle theory 6, the con- level forecasting in industrialised countries. in those countries which exhibit both higher sumption of all commodities tends. A func- The model is easy to calibrate, flexible and existing levels of car ownership and of per tion which flattens out towards some asymp- relatively straightforward to interpret. It has capita income, suggests the use of a non-lin- tote is, therefore, preferred on these also proved useful in earlier work in less de- ear forecasting framework. A limited amount theoretical grounds in addition to the fact that veloped countries conducted at the national, of work employing aggregate data looking at it corresponds to the pattern which emerges case-study level 7. December 1992 667 If we take P as the probability of an indi- vidual owning a car, S as the ultimate satura- tion level of car ownership per capita, X,. X, . * .X,, as a set of socio-economic influences on ownership and a,b 1,hb ...b,, as parameters, then the model can be depicted as: P = ~~S 2 'I~~~~~~I Manipulating and converting Equation (I) into logarithmic form yields the operational model set out as Equation (2). Values of P for estimating purposes become the actual levels of per capita car ownership for each country. The values of the parameters can then be de- termined by linear regression procedures, i.e. applying least-squares techniques' to: While the saturation level has been the sub- ject of a number of different theoretical inter- pretations over the years, here it is seen as no more than a technical aid to improving the quality of the ultimate forecasts generated. Surveying the work which has been con- ducted in industrialised countries one finds levels of saturation ranging from 0.4 to 0.7. Before adopting these, or si mi lar figures, for low-income countries one should note that there is considerable evidence that different countries, and indeed regions within coun- tries, seem ultimately to be heading fordiffer- ent saturation levels9. When modelling car ownership for the five groups of countries various saturation levels are adopted. These are based on both exami- nation of the existing patterns of growth and on techniques for estimating saturation de- veloped by Tanner'('. Since the ultimate satu- ration level is some distance away. especially in the lowest-income countries, the level of uncertainty involved in estimating long-run saturation is large. Further, for forecasting one also needs to take into account the fact that the data suggest the saturation level moves up with time. Therefore, levels in the range 0.3 to 0.45 per capita are assumed for S and estimations of other parameters. together with subsequent forecasts, are based upon this range.Turning to the other explanatory variables, there are empirical reasons to suspect that the relationship between car ownership and the major causal variables is not stable through time. This is consistent with the findings, based on a large sample of developing coun- tries, of a shift in the linear relationship be- tween per capita car ownership and per capita GNP between 1965 and 19731 1. Strictly com- parable updatings of these calculations are not possible; however, taking the data-set for the countries in this study and repeating simi- lar linear regressions in constant 1980 prices for the years 1967. 1973 and 1983 provides evidence that this shift has continued. In tech- nical terms it suggests that some variable (or variables) is being omitted from the analysis. Identification and quantification of this vari- able poses intractable practical problems. In order to capture some of its effect, however, a time trend (T) is included in the forecasting models used here (with T=l1 for 1967). Cross-sectionally countries vary in many ways which are impossible to quantify. But even if one could isolate variables which cap- ture this national uniqueness, it is unlikely that their future values could be predicted and hence used directly in a car ownership fore- casting context. In. order to take account of these specific, national features a set of di- chotomous, dummy variables (D), one for each country, is included in the model. These take the values of unity if an observation re- lates to a country and zero otherwise. Having taken these factors into account, the resultant pooled equations relate per capita car ownership for all countries in the data-set and for all years to the GDP of the countries at constant prices (Y), a dummy variable (D) which takes up the effect of spe- cific national deviations from the general car ownership income relationship, and a time trend. The operational form of the model thus becomes: In ( P )=)=a Xb/Y i+Xckm-,,nDktlnT ... (3) The main independent variable influencing per capita vehicle ownership at the national level is income. There is some evidence that income changes may exert a lagged effect on vehicle ownership '~ but, given the problems of specifying the appropriate lag structure, current income is used in the model. Addi- tional variables which may influence vehicle ownership include: the price of fuel, the level of urbanisation and the degree of industriali- sation. The importance of such variables are explored, but there are inevitable problems of multicollinearity. They are, therefore, gener- ally excluded from the main forecasting models. ParametersThe income variable employed is adjusted to 1980 prices and standardised to allow for ex- change-rate effects. Current income levels are used and no lagged structures are in- cluded. A simple tabulation by category of per capita income against per capita car own- ership for all the countries in the study over the period (Table II) confirms the positive re- lationship which has been found in most other studies of car ownership be they in industrialised or in developing countries. Using per capita income in the pooled quasi-logistic model (i.e. Equation (3)) for the five categories of low-income countries defined above confirms the importance of in- come. The result equations are set out in Table III (excluding the set of values for the country-specific dummy variables. D,). The explanatory power of all the models is high, the coefficients are significant and all coeffi- cients take the expected sign. Further, com- parisons between the quasi-logistic specifi- cation and an alternative log-linear model indicates a considerable degree of similarity. This suggests that the saturation levels adopted are unlikely to be dominating the quasi-logistic results. Table IV provides details of the coeffi- cients associated with the D variable in the quasi-logistic specification. These indicate the degree to which the quasi-logistic rela- tionship differs between countries. For fore- casting purposes, when predicting per capita car ownership for any individual country the appropriate dummy is added to the constant term in the quasi-logistic model. In addition to income a variety of other variables may be thought to exercise some in- fluence on the magnitude and pattern of car ownership and some comments are offered regarding a number of these. The parameters estimated for the quasi- logistic equations set out in Tables III and IV include a time variable as well as income (with T=l1 in year 1967). The variable proves to be statistically significant at the 5 per cent level and to exert a positive effect on car own- ership, i.e. it conforms to the empirical find- ings of earlier Transport and Road Research Laboratory work and to the linear cross-sec- tion results cited above. The broad implica- tion of this, in the context of the specific spec- ification of the time trend effect included in the model, is that over time car ownership at each income level will rise, although the mar- ginal effect of this time factor decreases with the years. One should caution against Table II. Cars per capita Gross Domestic Product per capital (Special Drawing Rights) <101 101-250 251-500 501-750 751-1000 1001-1501 1501-2000 2001-2500 2501-5000 > 5001 >0.0220 39 29 99 97 39 114 1 4 0.0200-0.0100 22 80 56 27 1 0 4 1 4 0.0100-0.0067 22 67 44 1 2 1 5 2 3 0.0067-0.0050 9 54 26 9 1 3 1 1 3 0.0020-0.0030 37 23 1 6 1 6 0.0030-0.0025 33 1 5 2 1 4 0.0025-0.0020 38 4 2 4 1 0 2 0.0020-0.0013 4 53 3 2 1 0.0013-0.0001 7 46 1 <0.0001 1 0 29 1 668 TRAFFIC ENGINEERING +CONTROL Table Ill. Country grouping Constant (a) Income* (b) Time (T) Assumed saturation level Adjusted coefficient of multiple determination A -8.70 (0.982) 0.571 (0.191) 0.109 (0.026) 0.30 0.84 B -8.24 (0.431) 0.699 (0.076) 0.103 (0.015) 0.35 0.90 c -9.83 (0.039) 0.943 (0.066) 0.088 (0.000) 0.35 0.82 D -11.80 (0.887) 1.100 (0.138) 0.261 (0.032) 0.40 0.67 E -10.74 (0.741) 1.160 (0.112) 0.244 (0.023) 0.45 0.90 *Income is GDP per capita expressed in 1980 constant prices in SDRs for Time, T =1 in year 1967 Table IV. Group A Group B Group C Group D Group E Ethiopia 0.464 Malawi -0.564 Liberia 0.156 Zambia 0.910 Morocco -0.271 Burkina Faso 0.337 Zaire -0.351 Indonesia -0.436 Dominican Republic 0.420 Zimbabwe 0.059 Bangladesh -0.910 Madagascar 0.085 Philippines 0.046 Cbte D'Ivoire 0.320 Mauritius -0.544 Burundi 0.134 Uganda -1.950 Papua New Guinea -0.501 El Salvador 0.755 Chile -1.033 Burma 0.248 Tanzania -0.653 Honduras -0.23 Peru 0.995 Costa Rica -0.870 India 0.000 Togo -0.041 Thailand 0 Turkey 0.000 Brazil 0.000 Rwanda -0.123 Niger -1.040 Botswana -0.435 Tunisia 0.720 Malaysia -0.280 Benin -0. 148 Cameroon -0.296 Colombia 0.630 South Africa -0.150 Kenya 0.000 Paraguay -0.42 Jordan 0.875 Mexico -0.712 Sierra Leone 0.420 Syrian Arab Republic -0.745 Algeria 0.141 Uruguay -0.520 Haiti -0.110 Gabon -0.720 Hungary 0.030 Pakistan -0.473 Portugal 0.380 Ghana -1.060 Yugoslavia 0.063 Sri Lanka 0.603 Panama -0.310 Senegal 0.399 Argentina -0.660 reading more into the coefficients than this since the nature of the specification was se- lected for purely statistical reasons - i.e. to provide a good fit to the historic data - and the results are sensitive to the year chosen to be set equal to unity.. In the past a number of commentators have observed'thAt much of the growth in car own- ership in developing countries occurs in urban areas11. In~terms of the percentage of population living in urban areas, the statisti- cal analysis of low-.income countries reveals no consistent overall relationship with per capita car ownership. What one does find, however, is that for many individual coun- tries there does seem to exist a positive, linear relationship between car ownership and level of urbanisation. Technically it is possible to extrapolate, using these linear models, future levels of car ownership in relation to urban growth for countries which have in the recent past exhibited a consistent relationship be- tween the two. Caveats should, however, be attached if this is to be done. First, the link between car ownership and-urbanisation emerges as being very location-specific and, hence, questions of its generality and tempo- ral stability arise. Second, in higher-income countries there is evidence that urban areas exhibit lower car ownership levels than in rural areas, suggesting that a linear relation- ship may flatten out and possibly turn down as development progresses'". Third, levels of urbanisation are themselves extremely diffi- cult to predict in the medium and long term and they are highly correlated with other fac- tors such as income levels, which can lead to problems of multicollinearity in more ex- tended models of car ownership. The progressive shift away from agricul- ture to manufacturing and service industries is changing the aspirations and attitudes of many people in low-income countries. For modelling and forecasting purposes. how- ever, the high correlation of the industrial mix of the labour force with such variables as in- come and urbanisation makes it difficult to isolate its specific impact. Also, graphical ex- amination of the data-set suggests that in practice the correlation between car owner- ship and industrial mix is, in any case, in practical terms very small. ForecastsForecasting involves making use of the model developed above and feeding into it predicted values of the independent variates (i.e. income and time). Here, for reasons of tractability, we focus on only the forecasts for a selection of countries from the total data- base.The major exogenous variable in all forms of vehicle ownership forecasting is income. This is itself,-however, difficult to predict with any accuracy even for a short period. Bodies such as the World Bank, Asian Devel- opment-Bank and other agencies provide pe- riodic forecasts of the anticipated future growth rates for low-income countries, but, because they are regularly updated and modi- fied, no single projection represents a stable input for transport forecasting purposes. The approach adopted is, therefore, to em- ploy sensitivity analysis and to offer a range of projections of future growth in per capita car ownership based upon alternative scenar- ios of how income may change in the future. The assumed rates of growth in per capita in- come are 0 per cent. 1 per cent, 2 per cent, 3 per cent and 4 per cent. These cover the range of predictions made by the major interna- tional institutions in recent years, with figures in the 2 to 3 per cent range being regularly given. December 199269 VOLUME BINDERS Securely and conveniently retain a year's copies of Tra ffic Engineering + Control not by metal rods, but by bonded nylon cords of great strength and durability which allow the easy insertion or removal of individual issues. Each binder - covered in red Balacron PVC-coated board, and with the journal's name in gold foil on the spine - is supplied with a gold foil, self-adhesive abbrevia- tion of the Volume Number and Calendar Year - e.g. 33-92 for Volume 33, 1992; 34-93 for Volume 34, 1993; and so on - which is easily affixed to a special panel on the spine. It is essential that, when ordering, you specify the Calendar Year required. The binders cost £5.00 each, inclusive of VAT and inland/over- seas postage. Send your instruc- tions (remembering to specify a Calendar Year, and enclosing payment if possible) to Printerhall Ltd 29 Newman Street London W1 P 3PE 669 Car pe 00HasCrspr10 ed 1978 1986Yeau 1994 2.6 2 1.6 05 2003 ur~ ~ ~~............ .1987 1978 1976 1986Year 1994 2003 0 -- ..... L '' '' ' I 1987 1978 1986 1994 2003 Year Cars per 1000 Heads 100 80 -- - - - ----- -.- 80 0- 20 * 1987 1978 1986 1994 2003 Year Cars per 1000 Heads .nn 1978 1986Year 1994 2003 Cars Per 1000 Heeds 1967 1976 1986 1994 2003 Year Cart per 1000 Heads 20-~ 161- iol- 6, 0 1---L1907 inn 1978 1986Year 1994 2003 Cars per 1000 Heads 80 F- 60-t 40 20 19i7 1978 1985Year 1994 2003 Cars per 1000 Heads 160~ ~ ~~~~~~~~~~. 100 .. ..... .. . 7 5 -~~~~~~~~~~ £ .... ..... ...... .... ...... .... .. ...... .. .... .. 1967 1978 1986Year 1994 2003 Effective GOP Growth + OlGmwth 0 S% Gmwth C 1 %Growth 0 4% Growth - Actual 0 2 %Growth Effective GDP Growth .0 %GObp 1 %Growth 0 3 % GDP Growth 0 4 % GOP Growth Fig!1. Plots of actual. modelled andforecast levels of cars per I 000 head of population for these selected countries: - Burkina Faso - Rwanda - Haiti - Pakistan - Cameroon - Syrian Arab Republic - Algeria - Colombia - Mauritius - Malaysia Forecasts are based upon assumed growths in GDIP of!., 2. 3 and 4 per cent. The results for a selection of countries (in- cluding representatives from each of our five broad groupings) set out in terms of cars per thousand population are seen in the various components of Fig 1. Actual car ownership levels per thousand head of population for the period from 1967, together with the levels predicted by the model, are also shown for comparative purposes. Noting the scale on the vertical axes, the overall picture which emerges is that even on fairly conservative assumptions regarding income growth, many low-incbme countries in our groups C, D, and E will experience very considerable growth in per capita car ownership in the medium term. In contrast the very poorest countries, using similar as- sumptions regarding income growth, will 670 ~~~~~~~~~~~~~~~~~~TRAFFIC ENGINEERING + CONTROL 2.6 2 1.6 0.6 ..- .. ... .... t. . .. . ... l. .. . 1987 Cars per 1000 Head* --- ... .. ---- ---0 0 0 0_ - 0 0a 0 a 0 0v t f 0 * C. ......... C. ...... .....--- .. .. ........-- --- --... .. C .... ...--.. - in. a 6 4 2 0 0 _ _ _ - ~~~~~~~~~~~~~ 0 1986Year 1994 1967 2003 Care per 1000 Heads 26 ~ 20 16 4- - . - - .~~~~~~~~~~~~~~0' +. ... 2- CCC~~~~~~~~~~~~~~~~~'1.1 10 ~ 6, ...... ........ ... ...... ..... ..... -... ... .... .... ..... .. ........ .-.. -- -- --- -- - ..... .. O _ 00 0 -- -- ... ... ... .... ..... ~ ... ..----- ---- -- ... . 0000 000 ... .-- ..1..... ... ............... ..... ...... -- -- --- -- ------ -- .... --.111. ..... ........... ..................... *. 0 01 1- ----- - CC ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~ 80 60 --- ....... ---1.1 ............ .... .-_ --------- ......1...... .. ..........1.......... --------- --- ........1........ ... .-_------ ........ .. ...... ........................... 1......... ........ 1.11,11
..I.....I. 0 .0 0 - ..... ..... .... .. . -------- '.' : : r.t................. ,11 40t 20 -- - .- 0.000.. 0o0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.. 0....... ........... 1967 - Actual 0 2 1h Growth 3 ' .....
1 .......1 ........1 ...1 ....1 1 .... .......... Care per 1000 Heads -Care per 1000 Heads 6 4 2 121 670 Table V. Country Burkina Faso Rwanda Togo Haiti Pakistan Cameroon Gabon Algeria Mauritius Malaysia Assumed annual rate of per capita income growth 1 %4%1 %4%1 %4%1 %4%1%4%1 %4%1 %40/1 %4%1 %4%1 %4% have much less dramatic rises in per capita car ownership' 5. In terms of national car parks, projections of population levels need to be applied to the national per capita car ownership forecasts. Demographic forecasting for low-income countries is as difficult as car ownership fore- casting and a variety of forecasts could be adopted. Here we use World Bank predic- tions of populations for the years 2000 and 2025 to provide a feel for the actual implica- tions for the transport systems of the growth in vehicle numbers. The results of applying these projections to our models produce the car ownership forecasts, with 1986 as a base, for selected countries, shown in Table V. As can be seen, the countries with rela- tively high car ownership levels at present can expect, by historic standards, fairly rapid increases in their parks even if their income rises by quite modest amounts (i.e. I per cent per capita per annum) in the medium term. A substantial part of this trend comes from the projected increase in population levels which, in some cases, exert a stronger influ- ence on the car-park forecasts than does the predicted rise in per capita ownership. The projected tripling of the population of Gabon by the year 2025 and a more than tripling for Rwanda are examples of the importance of the demographic factor in national car park forecasting.While presenting the sample forecasts in index form offers insights into differing growth rates it does tend to obscure the actual magnitude of the pattern which emerges. If one makes the relatively conservative as- sumption of a I per cent annual growth in per capita income to the year 2000, for example, Malaysia will have an extra 0.59M cars, while if annual income per capita rises by 4 per cent, slightly above the Asian Develop- ment Bank's prediction, it will have l.63M extra cars. The Cameroon, on the same basis, would find its car park rising from 0.09M vehicles to 0. 14M on a I per cent annual income rise assumption and to 0.22M on a 4 per cent annual income rise assumption. Given their transport infrastructures, these are very significant numbers of cars for these countries to cope with, especially since their numbers are forecast to rise relatively rapidly. Index of total national car park (1986=100) 1986 2000 2025 100 143 286 100 183 394 100 177 499 100 224 951 100 149 335 100 205 751 100 96 161 100 128 351 100 148 340 100 196 739 100 162 440 100 255 1309 100 215 632 100 355 1922 100 141 341 100 232 1000 100 146 209 100 219 544 100 149 284 100 216 628 ConclusionsSimple examination of trends over the past 20 years shows clearly the rapid increase in car ownership which is emerging in many low- income countries as their prosperity gradu- ally begins to grow and the aspiration of their populations for cars grows even faster. While at the lowest income per capita vehicle own- ership is static or falling, at high income lev- els it appears to be following the classic sig- moid-shaped growth path which has been observed in industrial states. The increase in the overall car park moves ahead of the rate of increase in per capita car ownership as popu- lations expand. Surprisingly little detailed study has been made of the exact nature of the underlying re- lations influencing both vehicle numbers and their use in low-income countries. One of the major difficulties in this type of work is to es- tablish a reliable and consistent database, This has been attempted here, although it is clear that deficiencies exist in both the cover- age and quality of the data employed and this must be taken into account when assessing the forecasts.Notwithstanding the data limitations the study has generated model results which are both intuitively reasonable and cross-refer- ence well with the limited work conducted elsewhere. It indicates that as low-income countries become more prosperous there is an inevitable and rapid rise in their car owner- ship rates. Reinforcing this income effect there is also a separate temporal effect as car ownership levels at any given level of income rises over time. This may be due to a variety of factors and is not fully understood but, from a forecasting and transport planning perspective, it inevitably adds to the ultimate growth in traffic volume. The address of Professor Button and of Mr Ngoe: Applied Microeconomics Research Group. De- partment of Economics. Loughborough Univer- sity, Loughborough, Leicestershire LEJ 3TU; and of Mr Hines Overseas Unit, Transport Re- search Laboratory, Old Wokinghamn Road. Crowthorne. Berkshire RG.1J 6AU. ACKNOWLEDGMENTS The work described represents a component of a stud ' commissioned by the Transport Research Laboratory on behalf of the Overseas Develop- ment Administration. It benefited considerably, froin discussions with Dr M. Cundill of the TRL. The authors also gratefully acknowledge the assis- tance rendered by Mrs Reynier of the Overseas Development Administration 's Library; J. E. Davies of Lou ghborough University Library: the International Road Federation;.and Shell Interna- tional. All the normal disclaimers apply. REFERENCES AND NOTES 'For a survey see: BUn-ON, K. J:, A. D. PEARMAN and A. S. FoWKES. Car Ownership Modelling and Forecasting. Gower Press, Aldershot, 1982: 'SATHAYE, J., S. MEYERS. Transportand home en- ergy use in cities of the developing countries: A. review. Energy Journal, 8, 1987, pp. 85- 104. 'BAYLiss, D. One billion new city dwellers -How will they travel? Transportation. 10, 1981, pp. 311-343. 4BERLAGE, L_, and D. TERWEDUWE. The classifica- tion of countries by cluster and by factor analysis. World Developmient. 16, 1988, pp. 1527-45. 5See, for example, KHAN, M. A., and L. G. WILLUMS1EN. Modelling car ownership and use in developing countries. Traff. Engng. Control, 27, 1986, pp. 554-560; and SILBER- STON, A. Automobile use and the standard of living in the east and west. J. Transport Eco- nomnics and Policy. 1970,4,3-14. 6For instance, see: RINK, D. R., and J. E. SWANN. Product life-cycle research: A literature re- view. J. Business Res., 7, 1979, pp. 2 19-42. 7CUNDILL, M. A. Car ownership and use in Kenya. TRRL- Research Repnrt 48. Transport and Road Research Laboratory, Crowthome, 1986. There may be technical grounds for using rather more sophisticated methods of estimation, but given the quality of the data we are working with together with the objec- tive of developing a relatively simple frame- work for forecasting and previous evidence that maximum likelihood procedures, in any case, result in only marginal changes in para- meter values (e.g. in Button et al, Op Cit.), or- dinary least squares is adopted here. 9FowKEs, A. S. and K. J. BUTrON. An evaluation of car ownership forecasting techniques. Int. J. Transport Econ., 4,1977, pp. 11 5-43. 'oTrANNER, J. C. Long-term forecasting of vehicle ownership and road traffic. J. Royal Stat. Soc. 141A, 1978, pp. 14-63. For the actual estima- tion procedures and estimation equations used in this study, see: BUTTON, K. J., and N. NGOE. Vehicle ownership and use forecast- ing in low-income countries. TRPRL Contrac- tor Report 278, Transport and Road Research Laboratory, Crowthome, 199 1. 'TRANSPORT AND ROAD RESEARcH LABORATORY. Investigations into vehicle ownership in de- veloping countries. TRRL Leaflet LF758, Transport and Road Research Laboratory, Crowthome, 1979. 12TANNER, J. C. A lagged model for car ownership forecasting. TRRL Laboratory Report 1072, Transport and Road Research Laboratory, Crowthome, 1983. "3For instance, see: BARRETT, R. Urban transport in West Africa. In: Proc. Seminar H. 14th An- nual Meeting. PTRC, London, 1986; JACOBS, G. D., and P. R. FoURACRE. Urban transport in developing countries. Traff. Engng Con- trol, 15, 1974 pp. 926-928, 93 1; and THOM- SON, J. M. Towards Better Urban Transport Planning in Developing Countries. World Bank, Washington., DC, 1983. '4PEARMAN, A. D., and K. J. BUTTON. Regional variations in car ownership. App. Econ., 8, 1976, pp.231-233. "5The forecasts shown are based upon the quasi. logistic models, but the same general conclu- sions emerge if log-linear specifications are employed. December 199267 671