Fuel markets in urban Bangladesh

IJjOC?~lJ.~Sh

Prrgamon

53.lHl +

Journals

O.(H) Ltd.

Markets in Urban Bangladesh M. J. PRIOR*

Consdmnt,

P. E. E. Ltd.. London, UK

Summary. - The results of a survq into fuel u\c In urban Bangl,ldesh arc presatcd. The fuds considcrcd include \\ood. gas. kcroscnc. electricity nnd other biomass. The survey sho\r> :I dominclncc of biomass use csccpt in town> whcrr natural gar and LPG art’ available. Thr: arc dkcus~cd. conscqucnccs ol’ thk for cncrgy plannin, u in Bangladoh

1. INTRODUCTION The fuel economy of Bangladesh is dominated by various forms of biomass. Recent work suggests that as much as 76% of a total fuel consumption in IYSJ of 730 PJ (17.4 mtoe) is provided by biomass. The major part of this is used in rural areas nnd is a mixture of various types of agricultural residue (rice-stri\w. cowdung i\nd jute-stick ;\rr the main forms) ancl wood. The supply of such fuel is subject to scason:\l tluctu;\tion as well as to annual variation as ;\ result of harvest conditions and cropping patterns. The bulk is obtained by gatherin g nnd ofher processes outside monetary exchange. Various studies have been of rural fuel use and a detailed analysis of these was undertaken as part of the Bangladesh Energy Plan.’ Fuel use in cities, particularly in households, has in contrast been largely ignored and treated as residual after industrial and commercial consumption of fossil fuels has been deducted. Urban domestic fuel use is, however. likely to become a much more important issue in the future as cities in Bangladesh are growing faster than the gener;\l population. It is generally accepted that whilst total population will grow at a rate of 2-7,.5’X/ year over the next 20 years, urban population will grow at a rate of 64% annually. Thus if, as is almost certain. the population reaches 1-WI50 million by 2000. from its current level of about 100 million. the cities will contain 45-60 million of these. In effect the cities, which at present form only about lS”/u of total population, will absorb most of the projected increase. This large shift and growth in population is the key planning problem in Bangludesh and it has important implications for energy as well as any other area. If, at one estreme, the fuel consumption accompanying this growth is largely accommodated by hydrocarbons then import

levels of kerosene are likely to rise sharply unless a major increase in the supply of domestic natural pas c:\n be sustained. At the other extreme. if the main fuel used is biontass then great strains ;\re likely to be placed upon wood supplies with repurcussions throughout the economy. The urban context is different from the rural in its lack of flesibility with respect to seasonnl variation. transport and storage requirements and. perhaps most important. because fuel needs are supplied within commercial markets. The change from a nonmonetary to a commercial market structure mei\ns that the role of fuel substitution based upon price-relativities may be much more important than in a rural context. Any forecast of future needs must rest upon an understanding of the current urban situation and this was clearly deficient in Bangladesh as indeed it seems to be in nlost countries with a heavy reliance 011 biomiiss use. Various investigations were made Lvhich included two detailed surveys as well as discussions with those involved ;n urban fuel supply. This paper reports the results of these suneys and comments upon their implications.

This work was done for the Bangladr~h Planning Commission as part of the Bangladesh Energy Planning Project. financed by the Asian Devclopmcnt Bank and the United Nations Development Plan and undcrtakcn by a consortium of consultants Icd by Sir William Halcrow & Partners. Thr: content of this paper is entirely the responsibility of the aulhor und doss not represent any other views. It has housvcr been influcnccd by many pcoplc associutcd with the project. too many to cnumcratr. Particular thanks are due to Dr. D. Wright. Project hlanagcr. for his support and advice. IO Dr. N. Islam for his offering of detailed knowledge ahout the Baryladesh energy situation and to Andrea Barnctt for hts helpful comments on an carlicr draft of this paper.

WORLD

x66

_. ? SUR\‘Ey

DE\‘ELOPMEZT

CONTENT

The surveys were done in blay/June lY8-l and consisted of a Household Survey (HS) and a LVood Merchant Surtry (W&IS).

The HS consist of IY27 interviebvs lvith urban households based upon a questionnare lvhich contained some 25 sections relating to consumption behavior in the preceding month. The survey was stratified with respect to income using an income distribution based upon the lY7-l-75 National Household Survey the latest available. It was felt that although this needed to be updated with respect to monetary values. it provided a reasonable guide to current income distribution. The distribution used was: Income group’ ($/month) O-29 ‘Y-42 42-7 1 71-144 >I44

Proportion (‘X) 15.3 20.3

30.8 24.5, 8.6

The survey was undertaken in towns throughout Bangladesh. chosen to include all geographical areas and according to their populations. The only major bias known to be incorporated in to the survey is that by taking the city of Sylhet as representative of the Northeast region, it overemphasizes the proportion of towns which have access to natural gas. In 1Y84. these included only the capital, Dhaka, in adition to Sylhet. This situation is changing rapidly however. The second largest city. Chittagong. is now connected to gas supplies and other urban centres such as Mymensingh are likely to follow shortly. The dwellings of respondents varied widely; slightly less than 23% were classified as being constructed wholly of brick or concrete and about 33% had earth floors with wood, straw or corrugated-iron walls and roofs. included in the latter were 13% of the total with earth floors and bamboo or straw walls and roofs. It is not known whether this represents an accurate sample of urban housing types but care was taken to include temporary and informal dwellings as well as permanent structures. It is not easy to place general confidence limits on the accuracy of a survey of this kind. The size of the sample is sufficient to ensure reasonable statistical accuracy with respect to the total population, but any survey in which some of the

responses are hassd upon memory (that is purchases of fuel and rice over the previous month) ib open to other bources of error. In some cases. (for example. aLerape per capita rice consumption and family hize) independent data exist to validate the sur\e)’ rwdt, but in its central conclusions about fuel use there is very little alternative data. The assumption is made that no systemaric biases exist in the responses and that therefore the central value3 are reasonablv accurate. However it would tre unsafe to assign a sreater accuracy than about 10% to these. It shouid also be noted that qurbtions were asked about purchases and not consumption. The survey \v;I~ carried out in one month (hlny 19S-1) and errors could therefore be introduced if households had any tendency to stock more than a month’s consumption of either rice or fuel. General observation suggests that this is unlikeI> in the urban contest but this cannot be demonstrated. It is also possible that the quantity of fuel obtained informally is underreported despite the efforts to include proper representation of lowincome respondents. The main results of the HS with respect to cooking fuel are shown in Table 1. The urban areas have been broken into three types: those with access to natural 2”s. those with ready access to LPG.’ and those \vith no significant gas fuel. The HS showed that the fuels used for lighting were restricted to electricity or kerosene and that, nationwide, 31.3% used onlv kerosene, 1Y .4X used only electricity whilst 49:2X claimed to use both in an undisclosed ratio. This probably reflects the frequent power cuts and the continued..use of kerosene as a backup lighting source. The mean nationwide consumption of firewood amongst wood-using households was 175 kg/month though this showed significant regional variation. In the capital, Dhaka, mean use was 127 kg/month whilst in towns closer to the main forest areas, principally Khulna and Chittagong, the mean use was 170-180 kg/month. These figures refer to those who claimed to use only wood for cooking. 23.4% of respondents claimed to have used some other form of biomass in the preceding month and. for these. the mean wood consumption was 95 kg/month. The mean per capita consumption was 291 kg/year with an average family size of 7.2. This is somewhat less than that reported from limited data by Foley (1984) which suggested a range of 400-1000 kg per capita/year in other countries. It is possibly significant that most of the data used by Foley is from areas where some of the wood is obtained without monetary payment. The effect

FUEL

Table

Fuel available

Fuel used Firewood Other biomass* Wood plus other biomass Total: biomass only Kerosene and biomass Kerosene Gas+ Electricity Other combinutionsl

I.

MARKETS

IN

URBAN

Fuel used for cooking

867

BANGLADESH

(percerrr of households) LPG (Chittapong)

Natural gas (Dhaka and Sylhet)

31 -t I1 16 -I 3 43 0 -I

49 1-t 7 70 I 0 18 0 1

No gas fuel

51 -I 21 76 s It -_

3 LV

‘Including sawdust, dry leaves, cowdung and other residues. tNatural gas or LPG. SSmall amounts of LPG are available. Xombinations of gas. kerosene. electricity and biomass.

of relying upon commercial supply may be to drive consumption down to minimum requirements. The various types of biomass consumed have different properties and can only be approximately aggregated. However, assuming equal calorific content, wood forms 83% of the total urban biomass consumption. It is probable that other forms have a slightly lower calorific content and that wood, in energy terms. constitutes perhaps M-86% of the total. It is possibly significant that the number of “other biomass users” was dominated by the Western and Central regions which are remote from forest areas and have no access to natural gas. (The possibility noted above of the underreporting of scavenged biomass should however be recognised.) When expressed in energy-equivalent terms, the mean household consumption of total biomass was about 35 GJlhouseholdlyear (0.83 ton-oil equivalent). There was considerable fluctuation about this figure, partly because of real variation due to family income and size but also because of unknown respondent errors and possible stocking effects. However it is interesting to note that, on general grounds.’ an average family would require 3WO GJlyear to provide subsistence cooking-fuel needs. It would appear that actual urban use corresponds quite closely to this subsistence level. (i)

Income

rwinfion

It might be expected

that fuel consumption would be income-dependent and the HS does show such variation. 1lowever the differences are not a straightforward income effect as the family

size shows a marked increase with income and larger families use more cooking fuel. The average wood consumption for households of all income groups of IO-12 members vvas 240 kg/ month whilst families of 5-6 used only 1-E kg/month. There is also a tendency - though not a marked one for higher income families to consume a larger amount of rice. Table 2 summarizes various income-group results. The per capita wood consumption does appear to rise slightly with income from 20.2 kg/month for the lowest group to 21.9 kg/month for the highest, but when this is corrected for the small rise in per capita rice consumption, the difference is small and certainly without statistical significance. It would appear therefore that the consumption of wood for all income groups is close to a subsistence need to cook rice of 1.8-2.0 kg wood/kg rice. A necessary corollary to this is that the burden of fuel purchase is very much higher, proportionately, for the poor. An exact estimate is not possible as only income groups were collected but the proportion of income spent on wood in the lowest range is probablv X-25%, a level which diminishes to below 5% for the highest range. The survey also suggests that the poorest pay higher prices for wood; the average expenditure in this group is $1.43 compared with %1.2-1.3/kg paid by all other groups. This may be a consequence of the structure of wood marketing; in most towns. secondary markets exist which sell small amounts of wood at prices well above the larger unit sizes obtainable at the main markets. Natural gas is much cheaper than biomass as a domestic fuel; a single-burner stove costs only

The

bur\cv

was slratlfisd

I IT.3 Cr’Y (a)

by income

31 iollow5: IlrUJrtrr grorcp

2 20.3 2Y-4’

3 31r.s 11-71

2 6.6

3 7.2

-1 24 Y 71

i X.6 >I&

proportion of ample S/month income

(‘1;))

size of IWUS~/lOf~~ Il!conlr

1 6.1 Survey (b)

mean:

7.2

gruup

4 7.Y

5 s.s

persons

perwns

Rice conscmpriou Irrcovw group I

67.2 II.0 Survey mean: (c)

88.9

2 73.8 II.’ k$hsrimonth

Ffw/wootl ~wchwtl (The analysis is confined

or

3 SY.6 12.5 12.3 kg/per

4 1UI.S 12,s capita/month

IO those clrumin g to use only wood

65.: 7.4

for cooking.

kg/hse/month kg per capita/month

a sample

of 837 households)

I’OlW!W

Sample

mean:

I 119.7 20.2 1.8

174.9

kg/hsr/month

or Z-I.5 kg/per capitdmont! Irrcotnr g’oU[’

1

3

1X.8 20.S

174.2 21.4

I .9

2.0

4

5

215.6 21.9 1.7

276.4 2S.O 3.8

kg/hsc/month kg per capita/month kg per capita/month/kg

rice

E.rpcv~di/wo Sample

mean:

I 3.9 0.66

5.0 $Ihsclmonth

2 4.1 0.61

or 0.7

j,‘pcr capitdmonth /rrCwtre grrJrr,r

3 5.0 0.69

-I 3.7 0.72

7.: 0.89

(d) Gus IIS(’ A total of-135 housheolds claimed to USC’gas for cooking. This shows the distribution It1conw gro1cp 1 2 3 In gas sample 10.3 13.5 31.1) 30,s 2U:3 In total sample 15.3

%Z.J/month for unlimited use. It is not surprising therefore that in towns where it is available it is in great demand. In Dhaka, 40.8% of respondents claim to use gas for cooking a proportion which is higher than suggested by official statistics which record only about one-third of homes as connected in 1983. It is probable that this reflects a degree of illegal connections, a fact which would pass unnoticed in consumption statistics which record domestic consumption as a residual after metered use in industry and commerce are deducted. The proportion of various income groups which claimed CO use gas is shown in Table 2 and although there is some bias to the higher groups it is not large. This includes households using LPG and, when these are

$/hsc/month S per capita/month

of such housheolds by income 4 32.0 21.9

5 IO.l”/o 8.6%

removed, there is very little income bias recorded. This is surprising given widespread anecdotal belief that gas supplies are provided mainly to higher income areas and the fact that the gas authorities are reluctant to provide gas to informal structures. It is possible that the explanation for the discrepancy lies in illegal connections.

The HS did not directly enquire about fuel price but the combination of quantity and expenditure data enables an average “price” to be derived. The national average for wood was 3.5 c/kg (or about 2.4 c/GJ) with a statistically

FCEL

blXRI\;ETS

IN

L’RB.\%

significant regional variation from a high of 4.S c.‘kp in Dhaka to 2.7 c in Chittagong which is close to the forested hill region of Bangladesh. The apparent price of --other biomass” where this was not free was 2.2. c/kg. Only 58% of the users of “other biomass” reported any espenditure at all: the remainder may be presumed to be scavenged in some way. As noted above. there is some indication that lower income groups paid more for their wood. (iii) Keroserze corrswnption The results for kerosene consumption are difficult to interpret as this fuel is used for lighting, often in conjuction with electricity, and. in a small number of cases, for cooking. However by making reasonable assumptions about the cooking use. the following conclusions could be drawn: -

-

the mean use of kerosene for lighting was 4.3 l/month including partial use with electricity; the mean use of kerosene for cooking was 3S.X l/month or about 15.5 GJ/year. This is based upon a small sample of only 30 households.

(b)

A total of 217 interviews were undertaken with urban wood merchants throughout Bangladesh.

p&e

Bqing

price (clkg)

No listing exists for merchants so it is not possible to know if the sample was representative. It is likely that it was biased against Dhaka in favor of other towns and in favor of the larger merchants. The size range of the respondents varied from those who claimed to sell below X tons annually (15.8%) to above 3750 tons (1%) but most, 55%. fell in the range 38-375 tons annually. Table 3 summarizes the main results of the W&IS both nationally and for the main regions. It can be seen from Table 3 that Dhaka. a city remote from forests. has the highest prices and the longest distances from which wood is brought. Other regions show a corresponding rela&onship depending upon their access to wooded areas. The average margin claimed by merchants. that is the difference between buying and selling price, was 17%. This is not all profit as the merchant incurs some selling costs but, as these are not likely to be large (transport costs are not included). the result suggests that the firewood trade is quite profitable.

3. COMMENTS

AND

CONCLUSIONS

The most important conclusion is obvious: the extent to which urban areas in Bangladesh are dominated by wood fuel and, to a lesser extent, by other biomass, principally leaves, sawdust and cowdung. Although corroborated by other indirect evidences this conclusion runs counter to

Wood Mercitatlt Swvey

Sri&g

869

B.-\SGLXDESH

Chittagong

Central

Western

Sundcrban

(cikg) 2.5

4.7

2.3

3.6

3.0

2.7

1.7

3.2

l.J

2.3

2.0

1.8

128 253 3.0 ma.’

86 107 5.4 4.8

104 98 4.5 23

98 49 10.9 20

61 64 4.5 n.a.*

Tratlsporr disrance (km) Truck Boat Pushcart Bullock curt

92 72 5.9 19

Transport COSI (clktn) Truck Boat Pushcart Bullock cart

0.50 0.31 0.21 0.38

Type of sules (%) Domestic Restaurant Industry ‘Not

available

56 32 11 because

of small sample.

870

\VORLD

DEC’ELOPMENT

previous work’ Lvhich has focused on biomass as a rural fuel and ignored its importance as an urban fuel. Such a division into rural and urban reinforces the USC of categories such as “traditionul” or “noncommercial” for biomass fuel and has tended to split energy analysis and planning into two compartments. one concerned with fossil fuels in the towns and in industry. the other with biomass in the country. Such distinction has its relevance in situations where biomass fuel supply is either associated with agricultural residues or with wood taken without charge from forests. However it is clear that in urban areas in particular. and in rural areas to some extent. such a situation does not exist. Fuelwood is a commercial good which competes. more or less successfully. with alternative fuels. It is selected in preference to kerosene, which is always available. but is widely rejected in preference to gas where this is available either as natural gas or LPG. In this selection process the dominant factor is the relative fuel price taking into account the heat values of the fuel. Although by no means exact. survey results suggest that domestic consumers have rather good perceptions about the relative values of fuels. It was first noted by Leach (IYSJ) that the price of fuelwood in Bangladesh bore a remarkably constant relationship to the price of kerosene. When adjusted for their relative heat values, the wood price has been about 30-4(1’% that of kerosene with a long-term average of 33%. (There is some regional variation in this. Dhaka prices tend to be closer to 40%.) This holds over the decade I97FS-l. a period which encompasses two major rises in the real price of oil which were passed on. with some lag. to the domestic consumer. Kerosene is not subsidized to any significant extent in Bangladesh. The current production costs of fuelwood arc not known, in part because illegal acyui>ition from state forests involves costs such as bribes which are not widely published. However it is unlikely that they have tracked the price of kerosene so precisely over the past decade. What is more likely to be happening is that merchants are adjusting prices of fuelwood to maximize their profits whilst retaining the competitive edge of wood. An obvious explanation of their ability to do this is that urban wood markets are highly cartelized. something which could be connected with the large volumes of wood which are known to be illegally taken from state reserves. There is some anecdotal evidence for the extent to which organized elements undertake the bribing of the forest rangers who are. in principle, responsible for the conservation of state forests. However a

firm conclusion would require a much more detailed examination of the market processes underlying urban fuelwood supply than was undertaken. A more straightforward conclusion is that the fuelwood market is well integrated with that for fossil fuels and that consumers are well aware of the value in use of the different fuels. Wood is unable to maintain a competitive position with respect to gas either piped or LPG. The fact that significant device costs are associated with both forms of gas suggests that, even for the poorest income groups. it is not the initial costs of purchasing stoves which inhibits consumers from switchin fuels but the relative fuel prices. It is interestmg that the Chittagong, a city without piped gas until 19S5. LPG was able to take nearly 28% of the domestic market in only two or three years. despite large initial costs for stoves and for bottles. Bangladesh has a comparatively low degree of urbanization, 15% in 1985. and may have a rather larger use of wood fuel than other countries though the data on this latter point is scanty. Foley (198-l) has reviewed the published information and suggests that generally in “major urban areas” of other countries. 50-80% of households use fossils fuels whereas even in Dhaka only 46% do so. Only two studies have been located which deal with urban energy use in other cities in the Indian subcontinent. A survey undertaken in Hyderabad in 1981 by Alam et 01. concludes that JI’%, of household cooking energy was supplied by wood, 31% by kerosene. 24% by LPG and 5% by charcoal. This survey excluded an estimated 19% of total households living in temporary houses; if these all used biomass, then the biomass proportion could rise to 50%. However it is still noticeable that kerosene has made much greater inroads into Hyderahnd domestic energy than in Bangladesh. The reason is likely to be that, as reported by Alam, the price of kerosene as well as LPG. when available, was subsidized and was below that of wood when expressed in useful-energy terms which account of relative combustion efficiencies. (The price ratios were wood: LPG: kerosene/l76:126: 100.) There has been a steep decline in the proportion of wood used as a domestic fuel from the 72% reported in 1966. A survey carried out in Bangalore in 1979-81 (Reddy, 1983) found that fossil fuels. primarily kerosene, were used by 65% of the households for both cooking and water-heating and that only 17% used biomass fuels for both functions. It seems inevitable in Bangladesh that the dominance of biomass in the domestic sector cannot continue much longer. The projection of

FUEL

MARKETS

IN

URBA-\N

energy demands is complex but. if it is assumed that urban population grows at 8% per annum and if gas penetrates to 6O’!L of available households. then urban biomass demand kvould increase from the current annual level of about 3.7 million tons to 13.1 million tons by 3000. The scale of such consumption can be gauged from the fact that if the whole of the Sunderban forest the major forest reserve outside the hill regions in the Southeast were to be clearfelled. it could supply about 13.5 million tons of wood (Byron. 198-L). In contrast. the rural areas can be forecast to show little or no growth in demand for biomass given a fairly slow-trowing population and a slow penetration of Foss11 fuels. The rate of migration to the cities suggests that the net rural population is indeed likely to remain fairly constant. If there were to be a significant shift towards kerosene in addition to a growth use of gas, then there would be major trade implications for Bangladesh. The substitution of 25% of the projected year XNt) wood demand would require about 1 million tons of kerosene. In 1983. total petroleum needs were only I.2 million tons. The point at which major substitution effects could begin to bc seen is not clear as little is known yuantitativcly about deforestation levels. The wood merchants ‘may. at the moment. be obtaining substantial excess profit over production costs and these could be trimmed if the underlying costs began to rise as transport distances rose and as the cost of gaining ndmission to “protected” reserves increased. A personal and tentative estimate is that the process could begin in about 1990 when the supply of urban fuelwood could become difficult to sustain below the price of kerosine in the major centers, particularly Dhaka. There are likely to be complex economic balances between the various planning strategies which might be adopted to cope with the increasing demand for wood in the cities. One might be

ST1

Br\NGLr\DESIi

to allow the market to determine the issue by setting wood prices at whatever level produced the necessary supply. At some point. prices would presumably rise to the point where kerosene became competitive and this would take a larger market share and ease prc’ssure on wood resources. The fuel mix would approach that observed elsewhere when kerosene is cheaper than wood. The disadvantage with this would be that it could lead to a rapid increase in the import of kerosene and. in the period u hen wood prices were adjusting. there could be acute effects in rural areas deprived of traditional wood resources in favor of urban demand. An alternative would be to increase the suppI> of wood by encouraging plantation growth For Fuel. This appears to be an economic proposition even at current prices (Byron. 19%) but there are various obstacles to farmers taking up this option, not least the acute shortage of land. However. various types of reforestation programs, including the use of marginal land in the reserve Forests and the tea-estates and comactive munity wood-plantations. are under consideration. One problem with assessing these is that there are virtually no data available. either for Bangladesh or elsewhere. on the realresource costs of fuelwood. particularly From plantations. as against kerosene. Such analysis would require agricultural as well as energ? expertise. given that one important component of the assessment would be the value of the land in alternative uses. It might also be possible to increase the penetration of gas beyond the 60% level assumed in the above projections. This would be a huge technical operation however as even the gas levels assumed require that annual connections increase more than I&Fold in 15 years from the current level of about 30,000. One thing is clear; the issue of fuelwood cannot be separated from other fuels. either statistically or in planning terms.

NOTES 1. The rural energy study was undcrtakcn by Dr. N. Islam and Dr. F. U. Mahtab and will be reported separately.

distributed in that city in any significant quantity. LPG recovery from natural gas may increase supplies in the future.

2. The USIS will be used throughout, though surveys were, of course, done in local currency, the taka. US161 = 2-l taka in mid-1984. No attempt was made to correct for purchasing power parity as no direct data on this is currently available for Bangladesh.

4. This figure is based on the assumptions that about 0.73 GJ per capita/year of useful energy is required to cook a subsistence diet. that the average family size in Bangladesh is 5.8 and that wood cooking stoves have an efficiency of IO-12%. The heroic nature of the calculation should be clear.

3. LPG quantities

has recently became from the Chittagong

available refinery.

in limited It is only

5.

This includes the Household

Expenditure

Surve!

and the e~idencr camot atisfy rhs twn. 6.

For exampI<

hlimeo. (Washington. D.C.: Drpartrn~ntoiG~upraphy. Osmania University. 19Sl). Byron. S.. FAO Adviser on Forcstry’ to Bangladeh Planning Commission. Personal Communication (1984). Foley. G.. “The impact of substitution hp conventional ftwls on woodfuel demand.” (London: Earthscan Information Programmc:. Earthscan. 19S-1). Leach. G., “lkxachold energy USC and prices: Bangladesh. India. Pakistan and Sri Lanka.” Xlimco

(London: International Instttutc for Enwrunment and Drwlopmrnt. 1%-l). Rural and R~ncw,lble hlanibog. F.. “Bangladesh: Energy iswe and projects,” Energy Dqxirtmrnt Paper No. .i (Washington. D.C.: Th< R’orlJ Bmk. l%? )