Income tax incidence in a developing country

Income tax incidence in a developing country

Journal of Development Economics 7 (1980) 247-262. @ North-Holland Publishing Company INCQPME TAX INCIDENCE IN A DEVELOPI?&; COUNTRY The Case of Gr...

1MB Sizes 3 Downloads 59 Views

Journal of Development

Economics 7 (1980) 247-262. @ North-Holland

Publishing Company


of Athms,



Received December 1978, final version received April 1379 The objective of this study is to throw some light on the question of direct tax incidence in a developing country like Greece. A two-sector mo+l is built up which emphasises a number of duality aspects. In that respect the model deviates from the standard Harberger type models. Appropriate shifting formulas are then con f~*~.~&d. __ Making use of a set of plausible parameter values our findings suggest that the taxed input may avoid some portion of the tax burden even under a perfectly competitive environment.

1. Introduction

Income tajc incidence theory has recently been tacklec” in a general equilibrium framework which, no doubt, represents a far more adequate trertment compared with the traditional piecemeal analysis. Most studies of tax-shifting in advanced countries use simple two-sector models, with fixed factor supplies e.g., Harberger (1962), Mieszkowski (1967, 1969). and McLure (1970, 1975) while they have appeared in :he literature models with variable supply elasticities but in the restrictive context of one-sector models [Feldstein (1974a,b)]. This paper is concerned with tax incidence in a developir+, ecopgmy. For this purpose, a two-sector, or dual, modtl seenas highly suitable; the assumption of’fixed factor supplies, however, is inappropriate. In view of this, the paper develops a model containing both features discussed above. The analytical framework thus developed provides the background for a detailed investigation of the nncidence eff’ects of partial factor taxes which are typical in the Greek tax structure. Nevertheless, we have to admit that, in order to keep the analysis within manageable dimensions from an incidence viewpoint, we are constrained to using a fairly simple model. Incidence effects are analysed within a comparative static framework in the absence of any dynamic element. Our major aim is to identify those forces which have a bearing on the final determination of incidence. ““This paper is based Essex.1 wish!o thank

on a Ph.D. thesis [Provopoulos (1977)-j submitted to the Universitya: my supervisor, Professor A.B. Atkinson, as weil as M. Wickens and R. Junankar for their many useful comments. Thanks are &a due to the Greek State Scholarships Foundation for financial assistance.


GA. Prcruopoltlus, Income tax incidence in a developing country

‘:‘his study is basically divided into sr:ven sections. In the second a simple model is presented, wl1ema.sin the th!ird the way taxes enter into the analysis is briefly explained. Then, sections 4 and 5 deal with the comparative static properties of the system regarding a. lar,jour and a capital tax. In order to examine the sensitiveness of our indices ;Nhen the various parameters involved take on iarbitrary, yet plausible, values specific numerical paradigms are offered in sectkm 6. Finally, sectian 7 summarises the conclusions, 2. The moCeI The two-sector model described below draws its inspiration both from the Harberger model wiclely used in tax incidence studies and from the dual economy models of the development kiterature. The latter, stemming from the early work of Lewis, have characterised the two sectors as agricultural and manufacturing. A major neoclass!cal element here is the hypothesis of a positive labour productivity in agriculture. This relates of course to the existence of ‘disguised’ unemployment which does not seem to be present in Greece.’ Agricultural labour lorce surplus is therefore ruled out. In addition, labour is represented in terms of man-hours rather than labour force. Since price flexibility is also assumed the model automatically implies a full-employment equ’librium. To put it differently, all the available labour force is engaged in productive work though the man-hours actually worked are endogenously dett:rmined. F’erfectly competitive behaviour is assumed to prevail in both the product andi the factor markets. Moreover, for the sake of simplicity, the model pertains to a clorsed economy. Considering an open economy setting would certainly introduce additional complications which are avoided by the closed economy assumption. It it, thus assumed that labour and capital are perfectly immobile2 at an international level and that imernal trade is only taking pla 3% Consider now the following formulation. Superscripts refer to the relevant sector, i.e., A for the farm and A4 for the manufacturing sector, and sulscripts denote the variables with respect to which dilferentiation is carried ou . !Support br such a stand is offered by the findings of two independent iuquiries, one carried ou. hy Pepelasis and Yotopoulos (1962) and another published as a ‘Report on the Long-Run Peqectives of the Greek Economy’ (1367, p. 164). So far as capital inflow is concerned it may bt: noted that its relative contribution in financing gross iuvezrment has been reduced during the last years. Foreign capital contributed, on the average, by about 11 percent to total gross domestic asset formation during the last d::ade. Capital outllow is not of a sizeable scale and can practically be ignored. Furthermore, nk em&r&ion whic.:;~in the past did take place at a large scale has now become a negative mqnitude though cff negligible dimensions.

G.A. Provopoulos, Income tax incidence in Q deoeloping countq


:igricultural production function

XA =.P(LA,IP),


where X represents output, L labour services and K capital sei vices. Manufacturing production function

XM =F”(LM, KM).


Both production functions are assumed to be linearly homogenous possessing the following properties: F{, Fi > 0,

F’L, FiK < 0, FhL > 0,


Agricultural real wage determination w’=F;,


where w is the wage rate. The farm good serves as the r,umeraire, so that its price is set equal to one. Manufacturing real wage determination wM= PFF,


where P is the price of the industrial good expressed in terms of the farm good, i.e., the terms of trade. Agricultural real capital price

rA= I;&


where r is the price of capital. Manufacturing real capital pric.: rM= PF,M. Frice determination P = P(X”, XA ),




G.A. Provopoulos, Income tax incidtvzce in a deueloping country

Eq. (7) implies that P depends on the quantities

produced in the two

:;ectors.3 The formulation of the labour supply schedules to follov. needs some explanation. Empirical research carried out in developed nations tends to establish, though inconclusively, a more or less inelastic 1abou.r s~>ply curve for supply of hours by adult males. But how far can these c,Jnclusions be applied by simple analogy to a less developed country like Greece? There are strong indications here to believe that the man-hours supplied are sensitive to wage rate movements. The most persuasive reason for such a view is a relatively slow rate of increase of the labour force during the last two decades accolmpanied by a much higher increase in the number of mandays employed. At the same time, this period is characterised by a substantial rise of the ovtrall and sectoral productivities and wages. Such evidence points to positively sloped labosr supply schedules. ,4gricultural labour supply

Labour migration is supposed to take place in one direction only, i.e., :From the rural to the urban sector. This might be explained partly by ‘economic and partly by social factors. Once the existing relationship between the industrial and the farm wage is disturbed migration may accelerate or :slow down. In other words, a rise in We will reduce labour offered in the rural sector, while an increase in wA will act toward more work elI’ort. In the first case migration becomes more, and in the second, less attractive. Manufacturing labour supply .

Eq. (9) shows th;at the labour supply in the industrial sector depends ou manufacturing, and farm wages, respectively. ‘The farm wage rate enters (9) in a lagged form on the ground that it takes time to migrate !;om the rural to the urban sectclr when changes in sectoral wages induce such i: migration. If, for instance, W’ goes down then farmers might migrate, brit t:‘~, .:bour supply in the m;anufacturin.g sector may only rise in period t + 1. ._ :vhat follows we are concerned with the impact of taxation in the current ie.kiod, that is, farm wages in (9) are considered as a predetermined variable. In a steady state sense :farm wages are also importart. Such a case is houever examined by Provopoulos. ‘A similar prices dett,rmination equation is used by Harris and Todaro (1970, p. 128) stated as f’=:IVX”/XA), P’cO. Our formulation aligebraic manipukxtions..

here is, however. more general to simplify subsequent

G.A. Brouopoulos, Income


inciderrce rn


tierekrpir~p coumr



Agricultural capital supply


.KA, 10.


Manufacturing capital supply

(11) Capital supply scheduies are non-decreasing functions of their owrr respective capital prices. Such a feature unde&es the assumption that capital employed in either sector is different in typz from the other, in view of the widely dinering capital requirements. As a consequence, no intersectoral capital mobility is allowed. 3. Introduction of taxes

The model just developed may be used 90 analyse the effects of some major direct taxes. Specifically, both a capital and a labour income ta.x can be introduced. Their main feature, however, is that they are thought to be partial rather than general factor taxes, in the sense that they apply to industrial labour or capital only. Such a treatment is certainiv consrstent with the country’s tax structure. Introducing a labour-income or payroll tax mc,Mies (4) to (12).

W”(l +rL)=fFf,


whereas a profits-tax modifies (6) to !13), (13: where tL and t” represent a proportional labour-Gncome and profits-tax rate respectively. In what follows the comparative static properties of introducing taxaiion are explored. Before, however, proceeding the way incidence effects are worked out should briefly be mentioned. Incidence is a fairly complicated phcnomenoq and relates to both ir put (sources of income) and commodity (uses &f income) price c Imposition of a p tial factor tax causes relative input prices in th rium in the taxed-Input market sector to change. y, since the net J-e will occur at a low iallen. Harberser t from the sector concerned. Instead, in pa-t of the taxed i our model such intersectoral movements are not taking place. Economic


GA. Provopoulos,

Income tax incidence in n developirig country

duality is here interpreted to mean that the links between the two sectors are rather wea.k. As an irnmediate consequence, the effects on the sources-of-income side are prim,arily lworked out through factor-price changes in the taxed sector alone. Only effects on the uses-of-income side might have a bearing 01: the after-tax income distribution on a nationwide scale. irlowever, commodity price movements have traditionally been left out from the analysis. I‘Le standard assumption made, and adopted by us as well, is that all Income groups involved display identical spending propensities with regard to the two goods in question. What is actually needed here is therefore an explicit formulation which hopefully quantifies the impact of taxation on ahe sources of income side. Such an irncidence measure for the taxed input has been put forward by Feldstein in the form of a ratio, the numerator of which indicates the factor’s income loss while its denominator being the change in total tax revenue. Let us first an employment-income tax as a representative case. A distinction should be drawn between alterations in employment-income due, OS;the onle hand, to changes in the wage rate with hours of work fixed and, on the other, chang;es m working hours at a fixed wage rate. To illustrate, the alteration in manufacturing labour income is anal ysed to (1.4) If both components are included in the computation of labor’s tzw pos:tion, the gains or losses from resulting changes in leisure are disregarded. For &at :reason ;it may appear theoretically superior: to identify income-loss with the second term of the right-hand side! of 1J4) only. Furthermore, posMating a tax function of the simple forrr, TL=p(wMLM), where ‘TL represents tax receipts, our incidence index may be expressed as .r,M(dwl”ldtL) w”LMi- tL(wM(dI?/dfi) + L"(dwM/dtL

sf =:-_-_



On a sitiilar basis, an .index for industrial capital in the case of a profitstax .is expressed as

116) r

‘o investigate the effects of a tax on iadusrial

emplo;yment income we

G.A. Provopoulos, Income tux incidence in a developing country


totally differentiate eqs. (1) through (ll}, where (4) is now replaced by (12). We thus get a system of eleven equations. Eq. (15) indicates that what is crucial in determining S is the sc!ution of that system for dw”/dti. Nonetheless, such an approach would not be much informative since dti,‘dti depends upon L&, L$ Lg, K’j and KE which are not unit-free magnitudes. An alternative and more useful approach is to make incidence a fu’unction of pure magnitudes such as elastZties. The fc!lowing aefinitions are employed for that purpose. We start by defining the factor-supply elasticities in both sectors as

The price determination demand equations, namely,




also provides

the two commodity


where nXtMjis the own-price elasticity of demand for the jndustrial Froduct, and %(A) is the cross price elasticity of demand for the agricultural good with respect to the price of the manufactured commodity. It might be noticed that the hypuihesis PA>0 implies that the two goods are gross substitutes. It also prove: to be useful to define each factor’s contribution to totA output of each sector. Co~~sidering t e case of an em loyment tax and earing in mind that linearly omogenous production functions have been


GA. Provopordo~, Income tax incidence in a develllping country

postulated jwe have4


Factor shares are not to lx considered as necessarily constant as a CobbDouglas version of the production-function might suggest. In fact, u’s describe ;T local property of these functions whenever infinitesimal tax alterations occur.. In a similar way, the local elasticities of substitution are provided by (18) for the farm and (19) for the industrial sector, respectively5 (#=






a”rM a”b~_--..-._=_ PK “F&





(1 -LEA)WA -7 GAF$,

(l-u”)wM(l+tL) PL”FM LL --

(1 - a”“)#(1 + tL) == [email protected] ’ = PL”FFL Kl.


We now proceed to a more detailed invcstlgation of incidence. Making use of the above definitions one is able to express SF in term;i of the factorsupply and product-demand elasticities as well as the technological properties of the scct.orai production functions. Eq. (15) may, of course, capture the situations of XI either incremental or a new tax. None:heIess, we focus attention to the second case only. It bas been show that, .when infinitesimal tax changes are studied, it is possible that the stem of the net income losses of the taxed and the untaxed factors exceed % section 5 where the mcidence of a tax on industrial capkal is examined (17~) is to be replaced by aM= w”k”/PXM and (17d) by 1 -a”=tif{l -t-rK)K”/FXM. % the case of a capital tax (19) is to be replaced by

G.A. Provopoulos, Income’tax incidence in a developing country


the incremental tax revenue.‘j To avoid such a complication we therefore consider the impositio.? of new taxes, thereby putting tL=O. It may be shown, after a series of manipulations, that whel a labour tax is introduced, incidence is given by SC L






and aME B=l+oM-

(1 -


*X(M) We are now interested in finding out the conditions under which industrial labour bears the burden entirely, i.e., SF = 1. Eq. (20) implies that for this to be so it must be the case that the numerator of the ratio in the right-hand side is equal to zero. Such a condition is met when EF =0 and &&) -0. Moreover, fullilment of these requirements leads to that result irrespective of the exact values taken by all other parameters. It can generally be shown that industrial labour bears at most the entire burden, since the denominator of the ratio in the right-hand side of (20) is always greater or equal to one, and the numerator less or equal to zero. Another interesting result is that the ratio of industrial labour’s share to th:1 total tax burden decreases as the value of EF and E&~) tend to increase. This is so, because (-A) is a decreasing function of E? and &&M).In other words, SF tends to be reduced for higher values of these parameters. 5. A. tax on manufacturing capital Similar procedures are followed here to analyse the effects (>f a tax on industrial capital. Now our incidence index is provided by

sKM-I+ -(/l;Z)


6Ahhotigh Feldstein (1974a) has shown algebraically and Musgrave and Mgsgrave (1973) graphically that this may be so .in a one-sector economy, it is expected thct a generalisation holds here for the same reasons cited by these authors.

G.A. Provopoulos, Income tax incidence irt a developing country


for the case of a newly imposed tax, i.e., tK= 0, where

and --_---I



nXW) A simple inspection of (121)reveals that for industrial capital to bear the full burden it is sufficient that its supply elasticity alone equals zero, i.e., $=O. This finding resembles the standard partial equilibrium conclusion, despite the fact that a widely differing analytical framework has been employed here. We also notice that SF = 1, when ef=O, no matter what values the other parameters. assume. As a general rule, capital will bear no more than a hundred percent of the nominal burden and in most cases below that limit. Finally, it can be seen that Z is an increasing function of $. therefore, thz: ratio Z/- (n+Z) Is negative and increases in absolute terms wh,en E: takes on successively higher val.ues. Under such circumstances SF tends to decrease. Deriving explicit formulas for both manufacturing lnbour and capital income tax incidence is an jmportant step toward understanding the bas:lc aspects of tthe whole &~ce(js. Nevertheless, beyond some statements of qualita.tive nature nothing, more can be added. No quantification of our findings can bl: established by applying econometric techniques. Suck an approar,h which would prov:de empirically determined parameters is unfortun;ltelly impossible in the light of the presant availability of statistical information ri4ne might, however, gain some insight by using a somewhat arbitrary, though not implausible, set of parameter values. The following section helps to illuminate that issue.

6. Numerical paradigms We first report the magnitude

of certain parameters resorting to findings provided by a number of authors for the Greek economy. Estimates of the elilsticity of substitution in manufacture were furnished by both cross-section and time-se:rie:s analyses with alternative specification G for the production function. Two of the earliest studies [Dracatos (19&) and KoutsoyianniMokklova (196r)] imply an elasticity equal to unity by the mere face of employing the Cobb-Douglas production framework. Further tests have also been provided by a couple of studies conducted by Mir tis (1973) and Lianos (1975), but wic:LeTy differing in their findings. The forme,, author’s results rend

GA. Pro~opo~l~s, Income tax incidence in u developing cotrntr)


support to an elasticity measure well below unity. On the other hand, Lianos using similarly a CES version of the production function, but &&rent definitions of variables and estimation techniques, obt ins a value for D generally greater than unity. We thus find it convenient t, simply use # = 1 for our purposes, a finding supported by Provopoulos’ econometric results too. Unlike manufacture, there is very scarce evidence for the farm sector. AS far as we know, there is only one study made by Yotopoulos (1967). Its objective was to fit a Cobb-Douglas function on the basis of a sample of farms in a specific area of the country. The author’s findings were pretty satisfactory. We therefore employ cA = 1. Labour’s shares to the aggregate output of each sector are represented by CIA =0.9 and a”=0.6, which reflects the differing degrees of capital-intensi1.e techniques in either sector. Slightly different values for uA affect very little (20) and (21). Furthermore, uM has been approximated by using National Accounts statistics. In the following tables concrete numerical examples are provided on the basis of the above information and a plausible range of’ values for the rest of our parameters. The latter, in the absence of any guiding information, were assigned three different values. Our benchmark case, which certainly cannot be defended on empirical grounds, is characterised by the parameter combination E&,) = - 0.5, s&A)= E; = nX(A)= 0.5, E$= 0.0, ELM = 1.O and YI~,~)= - 1.5. The own-price elasticity nXIM,is set to be greater than one (in absolute terms) reflecting the reasonable assumption of an elastic demand. Moreover, the assumption E; =O is not critical, in the sense that even somewhat higher \.alues have an insignificant effect on SM. Tables 1 to 12 present incidence values (in percentage terms) for an either labour (tables 1 to 6) or a capital (tables 7 to 12) tax, when two parameters are allowed to vary each time holding all others constant. As can be seen, the tendency is for the taxed factor to bear less than the full burden in most cases. These conclusions offer a useful guide for economic policy to the extent that realistic parameter values have in tact been employed.

S,” under alternative

0.00 - 0.50 -1.00

Table 1 combinations

100.00 52.71 35.79

71.66 43.62 31.35

of&f and ~;j,,+~,.

55.84 37.20 27.89


G.AL Prawopoulos, Zncome tax incidence in a developing country

Tabic 2 SE under alterilative combinations of 6: and E& Y 9 tiLIAl





51.49 52.71 53.87

42.18 43.62 44.41

36.59 37.20 37.77

0.50 1.00

Table 3 SF under alternative combinations of et and E$.






52.71 51.49 50.84

43.62 42.78 42.34

37.20 36.59 36.27

0.50 1.00

Table 4 SF under alternative combinations of ELM and EF.


G! c




0.00 0.53 1.00

53.84 52.71 52.12

44.30 43.62 43.26

37.63 3?.20 36.98

Table 5 Sr tinder alternative combinations of #EL” and nxIMj _.$







56.93 53.84 52.71

39.66 42.42 43.62

30.43 35.00 37.20

-1.03 - 1.50 -

G.A. Provopoulos, income tax incidence in a developing country

Table 6 S,Munder alternative combinations of ELM and n,(,,





0.10 0.50 0.90 1.30

18.23 52.71 66.74 74.34

17.00 43.62 52.80 57.45

15.93 37.20 43.68 46.82

Table 7 SF under alternative combinations of EF and e&r).

0.00 0.50





100.00 100.00 100.00

68.93 68.97 69.00


52.64 52.68

Table 8 Sf under alternative combinations of .L$and E&,.

s-4 L(A)




0.00 9.50 1.00

100.00 fQO.00 100.00

68.93 68.27 69.00

52.59 52.64 52.68

Table 9 SE under alternative combinations of E: and CF.






100.00 100.00 100.00

68.30 68.69 68.97

51.86 52.31 52.64

0.50 1.00



GA. I’n~vopoulos,Income tax incidence in a developing country Table 10 SF under alternative combinations



0.50 1.00

100.00 100.00 100.00

68.97 68.93 68.9 1




and E$

52.64 52.59 52.57

Table li Sz under alternative combinations

of E: and nxo,




1.00 __--

- 0.50 - l..Oo - 150

100.00 lOO.GO 100.00

61.89 65.66 613.97

44.8: 5O.!M 52.64

Table 12 SF under altcraative combinations

of a: and nX,Ab




100.00 100.00 1011.00 100.00

67.65 68.97 69.37 69.57

51.12 52.64 53.11 53.33



0.50 0.90 1.30

Before concluding the section a few remarks are in order. Attention was focusecl here on a closed-economy setting therefore the implications of factor mobility in or out. of Greece have not been explored. It would certainly be useful to know, even very tentatively, how the conclusions reached here might ble modified after partial allowance is made for international factor mobility. So far as labour emigration is concerned it was earlier noted that it has been reduced! to a negligible scale according to more recent experience. Since ta:< incidence is also (dependent upon factor mobility at an mtercountry level the above remark: tends to impy that the burden on industrial labour ot’ a tax would not be reduced to levels below those indicated by our earlier firidings.

GA. Provopoulos, Income tax incidence in a developing country


Slightly S,ifferent conclusions might however be reached with respect to capital. Regarding capital mobility in or out of Greece one should particularly focus on capital inflow rather than outflow for re:sons already cited. The introduction of a capital tax might thus be marSf sted in a reduced capital inflow. The further consequence could be that the relative burden of capital decreases relative to what our previous findings would suggest. 7. Conclusions To sum up, this study despite its limitations has provided sufficient insight into the problem of tax incidence in a developing economy. A fairly simple general equilibrium model has provided the analytical framework within which incidence aspects nave been investigated. The model’s main novelty lies in the fact that input supplies were allowed to vary in the context of a dual-economy structure. One should however be reminded that results thus obtained reflect linear first-order approximations. We are consequently bound to study the introduction of small tax-rates. Appropriate shifting formulas were subsequently developed for both a labour and a capital income tax in the indlustrial sector. These formulas were expressed in terms of the factor-supply and product-demand elasticities as well as the technological properties of the sectoral production functions. A basic finding emerging from this anslysis is that for the taxed factor to Ps: the futl bzlrden some very restrztive conditions must hold, even unaer perfectly conrpetitive assumptions. In the majority of cases examined the percentage burden is below a hundred percent. Until, however, empirical evidence establishes the magnitudes of the parameters involved, no precise idea about the extent of shifting can be gained.


Dracatos,C., 1964,

Production functions for Greek manufacturing, Special Studies Series (Bank of Greece, Athens). Feldsteiri, M.S.. 1974a, Tax incidence in a growing economy with variable factor supply, Quarterly Journal of Economics 88, 351-573. Feldstein, M.S., 1974b, Incidence of a capital income tau in a growing economy with variable savings rates, The Review of Economic Studies 41, 5Of--5 13. Harberger, A.C., 1062, The incidence of the corporation income fax, Journal of Polilical Economy 70,215-240. Harris, J.R. and M.P. Todaro, 1970, Migration, unemployment and development: A two-sector analysis, American Economic Review ti0, 126-142. [email protected] A.A., 1973, The demand for labour in Greek manufacturing, Research Monograph Series 20 (Center of Planning and Economic Research, Athens). Koutsoyianni-Kokkova, A., 1964, Production functions for Greek manufacturing (Center of Planning and Economic Research, Athens). Lianos, T.P., 1974, Capital-labour substitution in a developing country: The cw of Greece, European Economic Review 6, 129-14,.


G.A. Provopoulos, hwme tax incicienccin a developing country

McLure, C.E., Jr., 1970. Tax incidence, macroecc.romic policy and absolute prices. Quaftcrly Journal of Economics 84,254-267. McLure, CR., Jr., 1975, General equilkbrium incidence analysis: The Kuberger model after ten years, Journal of Public Economics 4, 125-161. Miesxkowski, P.M., 1967, On the theory of tax incidence, Journal of Political Exmomy 75,250262.

Miesxkowski, P.M., 1969, Tax incidence theory: The effects of taxes on the distribution of income, Journal of Economic Literature 3,1103-1124. Musgrave, R.A. and P.B. Musgrave, 1973, Public finance in theory and practice (McGraw-Hill, New York). Pepelasis, A.A. and P.A. Yotopoulos, 1962, Surplus labour in Greek agricultc-e: 1953-60, Research Monograph Series 2 {Center of Planning and Economic Research, At+ 1s). Provopoulos, G.A., 1977, Incidence effects of the Greek fiscal structure: Sornti theoretical and empirical aspects, Unpublished Ph.D. thesis (University of Pssex, Colchcster). Report on the long-run perspectives of the Greek economy, 1967 (National Research Institute, Atb:as). Yotopoulos, P.A., 1967, Allocative efficiency in economic development: A cross-section andyds of tiipirus farming, Research Monograph Series 18 (Center of Planning and Economic Research, Athens).