Policy, Research, and External Affairs
WORKING PAPERS
International Economic Analysis
and Prospects
International Economics Department
The World Bank
June 1990
WPS 452
Modeling Investment
Behavior
in Developing Countries
An Application to Egypt
Nemat Shafik
This model of investment behavior takes into account certain
characteristics common to developing countries, such as the
oligopolistic structiure of markets, putty-clay technology, the
inelastic supply of nontraded capital goods, and financial re-
pression.
The Pobcy, Research, and 1xtemaal Affairs ('Compl x d;sinbutc- l'K I i' rkiig I'apers todissrrsunatc the fiund np of work in progesa and
to encourage the exchange of ideas amnng Ildnk sad ' r  rJ1 lthcr  n:rcsled in dcNclop,ment issues These papers carry the names of
Lhe aui2cirs, reflect onlv %heir siews, and should hc, .u '  Qi , ;d d-rrding   Ihc rindings, intcrpreLauons, and conclusions are the
authors' own The) sh;ud nt hw, a tn1s.t ,, L.- '  , J li l ll:K   If l) .rsD . s   artanagercnt, or an) of us member crounines



Policy, Research, and External Affairs
International Economic Analysis
and Prospects
WPS 452
This paper -  a product ol' the Inteniational Economic Analysis ane Prospects Division, Intemational
Economics Department --- is part of' a larger el'i'ort in PRE to undenstand the detenminants of private
investment in developing countries. Copies are available f'ree fronm lhe World Bank, 1818 If Street NW,
WashingtonDC2(0433. Pleasecontact Joseph Israel, roomii S7-2 18, exension 31285 (66pagcswithfigures
and tablcs).
Investment functions are notoriously difficult to       goods anti the quantity ol' capital available to the
estimate, particularly in developing countries.         private sector appears to be a more realistic
Shafik presents a model of the determinants of          proxy.
private investment that takes into accouilt
common characteristic.s of a developing                      Shalik tests thk, model economctrically for
economy.                                                Egypt, using the re:ent literature on
cointegration and error correction to avoid
Firms' decisions about investment are               spurious regression.; and to estimate the long-run
outcomes of the oligopolistic structure of' mar-        equilibrium  relationship between investment and
kets, putty-clay technology, the inelastic supply       its determinants.
of nontraded capital goods, and financial repres-
sion. These factors resulh in an importait role              She discisses thv limit of'testing
l'or markups, interTal finanicing, demand, andl tYie    econometricldly whe her thc govemment
cost of investment goods -- defined, not as the         "cro\wids in" or "cro"ws out" private investment
interest rate, but as the price outcomiie f'romii the   anid lIC imnposxsibilitv of conistructing the
interaction of supply and dernand in thie market        countcrfactual. It is not possible to conclude
fbor capital goods.                                     xsihthercro%k ding out or in occurred at the
nmacrocconomlic level ( to accept the alternative
Bly conistructinig an index of thc rclative piice   h\ potthesis) bu. it is po.;sibIe to draw conclu,ions
of' investmrn,t goods, it is possible to pros i(le a    a1bout \N hat did nlot happen (the null hypothesis).
more meaningful indicator of' the true cost of'
capital to the firm under a repressed finiancial             'I'Tle model also provides a framework for
svstem. In an econonmy with a well-functionlint          nalyzing tie cffects of,zovernment policy by
credit market, the Keyncsian equilibrium condii-        considering explicitly thlc role of a number of
tion equating the marginal efficiency ol'inxcst-        possible instrumalnts suchi as the exchange rate,
ment with the interest is likely to hoid. But           tItIe quantity of credit ava lable to the private
under financial repression or  ihere credit             sector, and the composition and financing of' the
markets are impcrfcct, ltlC interest rate is not a      governiminlt budget. Future research may choose
true reflectioni of the cost of capital to tie firm.    to test otiher cmpiricail pro ies, such as protec-
Instead, a Lombinlatioln of' the price of investenict   tioiI, w,%ilitin Ihie same fram  \work.
Thc PRE Worki]`z Pa)Cr SCriCs (IISeCII1TteS thI tidIJIngS ol s ork uIILdr ' a! \In the Bank's Policy. Research, and External
Affairs Complox vAn ohjectis k of tei sceios is to ct Lhece fin,linrs out quR kl,, C% Cn if presentation, are less than fullv
pT lished. The findinfgs, interp)reiatitns. WILd  oIx IL ;S10nI in theI OWS cr jld`  nOt neSes aril\ represent official Bank pToli y.
IProd -,  ai:h,- Pk<lIRI1l<!;'!i.''     sl



TABLE OF CONTENTS
TEX                                                                            PAGES
1. Introduction: The Issues ...................    2
1. 1. Introduction  ............            ......................   2
1.2. Background: T'he Literature on Investment in Developing Countries  .   3
2. Microfoundations for a Developing Economy  ....................   8
2.1. Market Structure, Pricing ..nd Optimization  .................   9
2.2. Investment Dynamics: An Error Correction Approach ....             .......  14
3. Empirical Evidence  ...........                                                    17
3.1. The Data  .............  17
3.2. Econometric Estimation: Methodology             ....................        29
3.4. Stationarity Testing .     .................................  31
3.5. Engle and Granger's Two-Step Estimator ...................  35
3.6. Unrestricted Dynamic Estimation .......              .................  43
3.7. An Evaluation of the Results  ........             ..................  46
4. Conclusions  ......... ..............................                              52
Appendix A: Data Sources and Derivations .          .......................    56
Bibliography ........,,,,,,,,,,,,,,,,,,,,,,,, 61
I am grateful to Paul Armington, Ajay Chhibber, William Easterly, and Mohamned A. El-Erian for
helpful comments and suggestions.



TAB}LES,                                                                     PAGES
Table 1: Testing for Unit Roots: Dickey-Fuller (DF) Augmented
Dickey-Fuller (ADF) and Cointegrating Regression
Durbin-Watson Tests (CRDW) .......................  33
Table 2: Cointegrating Vectors for Investment (Levels Regressions)  ...        .....  36
Table 3: Dynamic Equations for Investment (Difference Regressions) ...          .... 40
Table 4: Unrestricted Dynamic Equations .........................  44
FIGRE                                                                        PAGES
Figure 1: Private investment Ratio  .............................  18
Figure 2: Relative Price of Investment ...........................  24
Figure 3: Markup Behavior .           .................................  25
Figure 4: Composition of Government Investment  ......              .............  28
Figure 5: Estimated Private Investment -- From Engle-Granger Procedure . ... 42
Figure 6: Estimated Private Investment -- From Unrest.icted Dynamic Equation  47



-2-
1. Introduction: Thi, issues
1.1. Intrdugtion
The economic literatwire on investment has been characterized by considerable controversy,
even by the standards of economists. A number of different, often overlapping, models of
investment determination have been hypothesized and the empirical evidence has done little to
clarify which, if any, are accurate representations of the way in which capital formation occurs
in the economy. This is particularly true for developing countries where there has been less
empirical work, the data are less reliable, and the appropriateness of existing theoretical models
is debatable.
It is against this background that this paper suggests some methodological innovations in
modelling aggregate investment behavior. A theoretical framework for analyzing investment
decisions is presented in Section 2 that takes into account some of the structural features of a
developing economy. Starting from the firm's optimization problem, an aggregate investment
function is derived Lhat reflects the results of a survey of decision-making in fifty private sector
firms in Egypt. The mo&ol is then tested at the macroeconomic level in Section 3 using new
econometric techniques that have emerged in the recent literature on stationarity testing and
cointegration. The relationship between investment and an array of government policies is
highlighted in the economnetric analysis. The conclusions, both methodological and empirical, are
presented in Section 4.



- 3 .
1.2. Background: The Literature on Investment in Developing Countries
Since there are a number of good surveys of the investment literature available,' this
section focuses on empirical models that are relevant to developing countries. The relatively few
attempts to estimate investment functions for developing economies have tended to use fairly
eclectic models that combine features of the flexible accelerator, neoclassical, and structuralist
approaches. Few studies have attempted to apply "q" models, which use the ratio of the tnarket
valuation of the existing capital stock to its replacement cost, to developing countries since stock
market valuations of corporate fixed assets are often non-existent or else are not meaninigful.2
A number of these studies, particul iy the country-specific ones, have provided
considerable insights into the factors that influence capital formation in developing countries.
For example, Behrman's work on Chile explores the validity of putty-putty versus putty-clay
assumptions across a number of d.:fferent economic sectors.3  His results revealed that investment
functions tended to differ across sectors, both in terms of the variables that were relevant and in
'See Serven and Solimano, 1989; Precious, 1987; Bruaker, 1985; Nickell, 1978; Helliwell, 1976; Rowley and
Trivedi, 1975; and Meyer and Kuh, 1957.   The accelerator and flexible accelerator models are described in
Samuelson, 1939; Eisner, 1960; Meyer and Glauber, 1964; Brown, Solow, Ando and Karenken, 1963; and Eisner,
1967. The neoclassical model originates in the work of Jorgenson, 1963; Jorgenson, 1967; Jorgenson and Siebert,
1968; and Hall and Jorgenson, 1971. The "Q" theory of investment and the related adjustment costs literature are
presented in Tobin, 1967; ToDin, 1969; Tobin and Brainard, 1977; Hayashi, 1982. Modern versions of Keynes' model
based on the supply of capital goods can be found in Haavelmo, 1960; Witte, 1963; and Precious, 1987. The
Kaleckian profits model of investment determination is described in Kalecki, 1971. An example of a structuralist
model of investment behavior is provided in Taylor, 1987. Disequilibrium models of investment are described in
Malinvaud, 1980; Malinvaud, 1982; and Sneessens, 1987. Investment models that focus on financial constraints can
be found in Fazzari and Mott, 1984; Fazzari, Hubbard and Peterson, 1988a; and Fazzari, Hubbard and Peterson,
1988b.
2For example, in many developing country stock market shares are not truly traded so that quoted prices do
not reflect the market valuation and expectations of future profitability. Dailami uses "q" models to explain private
investment in Brazil and Korea. Dai!ami, 1987, 1990. Solimano applies a 'q' model to Chile, but also adds a
number of additional explanatory variables. Solimano, 1989.
3Behrman, 1972.



-4-
terms of their lag structure.  Pinell-Siles'  .Ady of private investment in India highlights the
dampening effect that the tax system has on capital formation because of the ,ailure to adjust
taxable income for inflation.4  Using panel data on Colombian firms, Bilsborrow found that the
availability of foreign exchange to implement planned capital formation and the internal flow of
funds were the most important determinants of investment.5 The importance of cash flow effects
reflect the uncertainty, informational constraints, and weak capital markets faced by Colombian
entrepreneurs.'
These earlier studies by Behrman, Bilsborrow, and Pinell-Siles were followed by a number
of more ambitious multi-country analyses that highlighted the role of government policy,
particularly public investment, on private capitdl formation in developing economies.7 The
theoretical framework adopted was sometimes ad hoc or some combination of neoclassical and
flexible accelerator models with additional variables to capture the effects of government policies.
Fry estimated investment functions for sixty-one developing countries using demand, relative
prices, the exchange rate, and the availability of domestic credit.8 His estimates find significant
effects on the investment ratio from the growth rate of GDP, the ratio of foreign exchange
receipts to GDP, the ratio of domestic credit to GDP, the purchasing power of exports, the ratio
of actual to expected prices, and the lagged investment ratio. Tun Wai and Wong's estimates of
4Pinell-Siles, 1979.
5Billsborrow- 1977.
'More recent work by Dailami on Colombia found that the high real marginal cost of capital, especially for
small- and medium-size firrs, served to constrain the expansion of capacity in the private sector. Dailarni, 1989.
7Fry, 1980; Sundararajan and Thakur, 1980; Tun Wai and Wong, 1982; Blejer and Khan, 1984.
aFry, 1980.



- 5 -
private investment functions for five courtries used government investment, the change in bank
credit to the private sector, and the inflow of foreign capital to the private sector as explanatory
variables.9 The econometric results indicated that government investment was the most important
explanatory variable for Greece, Korea, and Malaysia, whereas bank credit was critical in
Thailand and foreign capital inflow in Mexico. Retained earnings were included in the regressions
for Greece and Korea, the only countries for which data were available, but had insignificant
coefficients.
Private investment functions were estimated by Sundararajan and Thakur as part of a
growth model intended to measure the effects of public investment in India and Korea.'0 They
use a combined neoclassical and flexible accelerator model with additional terms for the public
sector capital stock and real savings available to the private sector." They found significant
coefficients on all the variables for India and Korea except for the public sector capital stock
When the long run multipliers for public investment were calculatc  the effects for India and
Korea were strikingly different. For India, the effect of public investment on private capital
formation was weak because of an initial strong crowding out effect that was not offset for many
periods. These negative effects were attributed to the high incremental capital-output ratio in the
public sector in India. In Korea, they found that the effect of public capital formatiou on
private investrent was unambiguously positive in the short and long run.12
'Tun Wai and Wong, 1982.
' Sundararajan and Thakur, 1980.
"Real savings available to the private sector was defined as total savings mninus public investment in real terms.
'2For an analysis of the impact of stock markets on private investment in Korea, see Dailarai, 1990.



- 6 -
Blejer and Khan's paper is one of the few that uses an optimizing framework for the firim
to derive an aggregate investment function to evaluate the effects of government policy."  The
resultiug model is essentially a flexible accelerator that allows for government policy to affect the
speed of adjustment to the desircd capital stock through a standard partial adjustment
mechanism. "  Blejer and Khan hypothesize that the factors that affect the speed of adjustment
to desired levels of capital are the stage of the business cycle, the availablity of financing, and
the level of public sector investment. They argue that it is important to distinguish between
public investment in infrastructulre, which is more likely to "crowd in" private investment, from
that in other areas. However, the empirical testing of this distinction is weakened by the
empirical proxies used for infrastructure investment.'5  Their empirical results from twenty-four
developing countries found an important positive effect on private investment from the degree of
capacity utilization and the availability of credit. They also claim evidence in support of their
1 Blejer and Khan, 1984.
14 Note that thlere are problems with such an approach that stem from the deficiencies of the standard
partial adjustment model. Under partial adjustment, agents incur costs for any kind of change, even a
desirable one. Consequently, in a growth situation, the model results in consistent undershooting of the
desired capital stock. It is possible to address this by using a generalized version of the partial adjustment
model, the error correction model, that will be described later. See Nickell, 1985.
i Blejer and Khan used two different proxies for infrastructure investment: (1) a proxy based on the
premise that infrastructure investments have a long gestation period and therefore the tre-id level of total
public inves,ment can represent infrastructure; and (2) a proxy that posits that because of its long run
nature, infrastructure investment is more likely to be anticipated. However, infrastructure investment is
usually very lumpy. Therefore, the measure based on the trend level of investment may be reflecting otber
types of investment spending that are fairly stable over time. Similiarly, expenditure on infrastructure is
often uncxpected since it can, by its nature, be postponed if neglect or deterioration is tolerable. Also,
because public investment in infrastructure in developing countries is often associated with borrowed
resources from banaks or donors, there is likely to be even greater uncertainty in formulating expectations
about future outlays.



-7-
positioz. that government infrastructure investment crowds in private investment whereas other
public investment crowds out private activitv.'6
A later study of the effect of public policy on private investment in Turkey by Chhibber
and van Wijnbergen used Blejer and Khan's framework, but used actual data on government
infrastructure spending." They also calculated real effective interest rates that took explicit
account of compensating balances.  Compensating balances are a means by which banks
circumvent low adrministered interest rates by requiring borrowers to place deposits in non-interest
bearing accounts as guarantees for loans,' Their econometric results show significant coefficients
for output, the real effective cost of borrowing, and private sector credit as a share of GNP. The other
two explanatory variables tried, an index of capacity utilization and the share of infrastructure in total
public investment, did not have sigaificant coefficients. They conclude that the effect of governrent policy
on private investment is complex and must be analyzed in light of a range of relevant policies including
exchange rates and institutional factors such as export promotion programs. The high rate of public
investment in Turkey resulted in some inflation and a raising of interest rates; however, it also insured
that the economy's adjustment effort was growth-oriented.
In summary, the empirical work on investme:; in developing countries has tended to draw from
the standard models in the literature and add elements that are relevant to the economy under
16In contrast, Balassa findb evidence that public investment and private investment are negatively
correlated using Blejer and Khan's data set. Balassa, 1988.
17 Chhibber and van Wijnbergen, 1988.
i8Chibber and van Wijbergen use a technique based on the relationship between commercial bank
deposits for transactions purposes and commercial bank loans for transactions uses to assess the
importance of compensating balances in Turkey. The excess of deposits over uses reflects the importance
of compensating balances. See Chhibber and van Wijnbergen, 1988 for a description of a technique
originally proposed by Ersel and Sak, 1989.



- 8 -
consideration. Many of the early studies, such as those of FTry and Tun Wai and Wong, simply produced
a list of variables that are correlated with investment in particular countries. Despite the problems
associated with fixed factor coefficients in the accelerator model and the limitations of the partial
adjustment model described above, Blejer and Khan's model represented an early attempt to develop the
theoretical underpinnings of an investment model tailored to a developing economy. The model developed
below is in the same spirit, but attempts to use realistic microfoundations as the starting p,oint for a
macroeconomic model of investment that incorporates the effects of government policy. In addition, the
econometric testing that follows will in.orporate the recent literature on stationarity testing and
cointegration to avoid the spurious correlations absociated with trended time series and to allow for an
analysis of the long run equilibrium relationship between investment and its determinants.
2. Microfoundations for a Developing Economy
The microfoundations described below have emerged from a survey of fifty private sector
firms in Egypt. The purpose of the survey was to identify those factors that influenced private
sector investment decisions in order to develop a realistic analytical model and to contribute to
the interpretation of the econometric results. The methodology, questionaire and detailed survey
results are available elsewhere.'9  The discussion here will be limited to deriving a theoretical
model of the firm's investment decision-making process that reflects the conditions that may
prevail in many developing, as well as in some developed, . onomies.
"9 Shafik, 1989.



2.1. Market Structure, Pricing and Optimization
Because most private firms in developing econormies are managed by their owner, the "black
box" assumption that the objectives of shareholders are the same as those of the firm's managers
is fairly plausible. Such an assunption would not necessarily be as credible in an economy where,
because of separation between management and shareholders, there may be multiple objectives
within the firm on the part of different decision-makers.20 Consequently, the objective function
hypothesized is a standard maximization of the expected utility (EU) of operating profits (ii):
(1)    Max EU(n ).
Output markets in developing countries are often characterized by oligopoly because of
market size, government policies, financial barriers, technological considerations, supply
eanstraints and a variety of structural features of developirg econornies. Consequently, it is
important to allow for the possibility of a divergence between price and marginal cost when
considering the firm's optimization problem. Similiarly, because investment decisions tend to be
irreversible since second hand markets for capital goods often do not function efficiently, and
there are large costs associated with asset liquidation, expectations about future demand and
profits are important.
X "Managerial" models consider the nature of the objective function in a corporate structure. See
Marris, 1964; Marris and Wood, 1971. The assumption of profit maximization may also be valid where
corporate managers have a shareholding stake in the ftrm or where the firm is faced with the threat of
bankruptcy. See Jensen and Meckling, 1976 and Grossman and Hart, 1982. The more recent literature
on "principal-agent" problems considers how principals (shareholders) can manipulate the incentive
structure so as to produce optimizing behavior on the part of agents (mnanagers).



- 10 -
Consider the following .inition of costs that distinguishes between direct production
costs and the indirect cost of capital as well as differences between imported and domestically
produced inputs:
(2)    C - WL + (l1-.)P DMD + leP5wM7 + [(_El)P kD + 9ePk-) [(d + r - z) (1-i)/(1-u)J
where
C    =    total costs
W    =    wages
L    =    labor
+    =    share of imported raw materials and intermediates
pmD   =    price of domestic materials and intermediates
Pw  =    price of imported materials Pnd intermediates
e    =    exchange rate
M    =    domestic raw materials and intermediates
Mw  =    imported raw materials and intermediates
I    =    investment
e    =    share of imported capital goods
Pk'  =    price of domestic capital goods
pW   =    price of imported capital goods
8    =    depreciation rate
r    =    interest rate
z    =    capital gains
i    =    present value of tax savings from investment incentives
u    =    rate of corporate taxation
The disaggregation of costs into a domestic and foreign component is useful because it
allows for the consideration of an explicit role for the exchange rate, an important feature of the
investment process in developing economies.2' This is because in many developing economies
21This is highlighted in Chhibber and Shafik, 1990.



foreign exchange is rationed and investment is usually highly import dependent given the small
size of the domestic capital goods industry.
Consider the simple case where there are N producers of a standardized product, a single
selling price, no new entry, cost functions of firms may be different, and inputs and outputs are
sold to price-takers. It is then possible to derive the markup over costs from a standard profit
maximizing framework that takes into account the interdependence of firms' decisions under
oligopoly.22
Each firm has a profit function:
(3)      i = p(Q)qj - C, (qi)
where p = f(Q) and Q = E qi.
The profit maximizing first and second order conditions are:
(4)     d i/dqi = p + qi (dp/dQ . dQ/dqi) - dC,/dqi = 0
(5)     d 2 t/dqi < 0 for all i.
22 For a detailed derivation, see Waterson, 1984. It is possible to derive the model with heterogeneous
products, differentiated prices, and potential entry of competitors, but the assumptions made here serve
to simplify the exposition.



- 12 -
Rearranging equation (4) and using the above definition of Q, it is possible to derive the deviation
of price from marginal cost as:
(6)              = si(l+ai)/4
where  -r   =  mark up over costs for the ith firm.
aj  =   dQ/dqi = effect of the ith firm's output on other firms' output
s;i =   ith firm's market share
I =   price elasticity of demand.
The outcome of the firm's optirnization problem depends on those features that characterize
an oligopolistic situation: the industry demand elasticity, market structure or concentration, and
beliefs about rival behavior. The price cost margin, or mark up, in equation (6) allows fox
situations in which firms may have different marginal costs of production, be of different sizes,
and hold different conjectures about how their rivals will react. Policies such as protection would
affect the mark up through the firm's market share and the price elasticity of demand.
The firms profits are determined by the mark up rate, costs, and the level of aggregate
demand in the economy:
(7)    n       = f(r, Y, C)
where  i       =  profits
Y      =  aggregate demand
C      =  costs.



- 13 -
In addition to determining aggregate investment in the economy, profits are crucial at the
firmn level because they generate the internal funds that facilit3te investment. This is particularly
important in developing countries where capital markets usually do not satisfy the conditions for
the Modigliani-Miller theorem to hold.23  Because stock markets, where they exist, are usually
underdeveloped and shareholders are rarely anonymous, there is little pressure on management to
distribute dividends. Where financial markets are characterized by artificially low adminsitered
interest rates, firms have a greater incentive to use debt financing since their borrowing costs are
being subsidized by depositors.
The firm's desired investment (I) depends on expected mark ups, demand and costs:
(8)   1 = f(.r, Ye, CO)
These determinants of the capital stock in equation (8) reflect an array of variables outlined above
including market shares, price elasticities, rival behavior, exchange rates, wages, interest rates,
taxes, investment incentives, and the price of capital goods.
23 The Modigliani-Miller theorem posits that firms are indifferent between internal and external sources
of financing under specific assumptions about the way in which capital mnarkets operate. Modigliani and
Miller, 1958. Even in indLstrialized economnies with well-funclioning capital markets there is considerable
empirical evidence that rettsntions are an important determinant of investment. For example, in the United
Kingdom, retained profits accounted for approximately 75% of the finance for new investment. See King,
1977, p. 209.



- 14-
2.2. Investment Dynamics: An Error Correction Approach
The process by which firms move from actual to desired levels of capital stock is
hypothesized to follow an error correction process. There are a number of advantages to such an
approach. Firstly, error correction models have proven to be usefvl for explaining a variety of
long run macroeconomic relationships.24 Secondly, unlike the more common partial adjustment
model, the error correction approach implies that firms incur no costs for changes that are
planned.25 Thirdly, the recent literature on cointegration provides a theoretical rationale for the
empirical success of error correction models.26 Specifically, the error correction levels term
captures the long run equilibrium relationship between variables while the differenced terms
capture the dynamics. Granger establishes that error correction models produce cointegrating sets
of variables and that cointegrating series can be represented by an error correction process.27
Series are said to cointegrate if some linear combination produces a stationary, or "white noise,"
error. Fourthly, the error correction framework is intuitively appealing because it provides a
realistic representation of how rational, but fallible, agents make decisions. In the context of a
developing country in which information may be far from perfect, decisions about the long run
capital stock desired are likely to be characterized by gradualism and revision.
24 Davidson, Hendry, Srba, and Yeo, 1978; Bean, '981; Currie, 1981; Salmon, 1982; Henry and
Minford, 1988.
25 Nickell, 198.5.
26 Engle and Granger, 1987; Jenkinson, 1986; Dolado and Jenkinson, 1987; Giovanetti, 1987; Henry
and Minford, 1988.
27 Granger, 1983 and 1986.



- 1 5 -
The firm's intertemporal optimization problem is to minimize the costs associated with
adjusting to the desired capital stock over an infinite horizon. Given the following quadratic cost
or loss function:28
(9)    Min  EVc(It - It*)2 + (It - it.1)2
Differentiating the above, an equation of the error correction form is obtained:
(10)   &It = aOA It + al(I., - It ,*) + a2A I,..
This formulation implies that investment responds both to changes last period as in the
partial adjustment model as well as to changes in the target, I*. The levels term, (It, - It., *),
captures the divergence from the long run equilibrium relationship caused by the costs of
adjustment.
The relationship between the desired level of investment and the desired capital stock is defined
conventionally as:
(11)   it = K* - (I -
where o = rate of depreciation.
28 For other derivations of error correction models, see Nickell, 1985; and Davidson, Hendry, Srba
and Yeo, 1978. For an earlier derivation of an error correction-type model intended to explain farmers'
supply response to prices, see Nowshirvani, 1971.



- 16 -
Thus the firm chooses a desired level of investment given a lagged capital stock so as to achieve
its desired capital stock. With irreversible or putty-clay technology, it is necessary to assume that
0.
1t 2 O.
Using equation (8) that defines the stochastic process generating the optimal target investment to
substitute into equation (10) and assuming that expectations are realized, one obtains the dynamic
reduced form for investment:
(12)   Ait = DOAr, + PIAYt + 2ACt + a, (,.r  tl - 4Yt. -5Ct.,)+a2AIt., + et
where et = error term.
This is the equation that will be estimated econometrically below.
The above model provides a framework for analyzing the consequences of an array of
government policies for private in-estment. Policies that affeet aggregate demand enter directly
through the accelerator term. Protective tariffs and quotas as well as a domestic licensing system
that restricts new entrants alter firm mark ups and result in higher investrment in those sectors.
Government policies that affect the costs associated with investment can enter through a number
of channels defined in equation (2) including the exchange rate, interest rate, tax incentives, and
rate of corporate taxation.



- 17 -
3. Empirical Evidence
3.1. The Data
The period chosen for econometric analysis, 1960-1986, was, to some extent, determined
by data availability. The sample encompasses two different periods in Egypt's economic history,
especially in terms of the relationship between the public and private sectors. Prior to 1974,
Egypt followed an essentially statist import substitution policy whereas after 1974 the government
sought to encourage the private sector under the banner of the "open door policy," or "infitah.''
The evolution of private investment over the sample period is depicted in Figure 1. Whether the
underlying determinants of investment were stable under these two subperiods will be explored
econometrically.
The data in Egypt, like that in most developing countries, are limited in terrs of both
quantity and quality. Quarterly observations are not available for the vast majority of economic
variables and methods for their estimation are sometimes shrouded in mystery.29 Consequently,
a number of different sources were used and cross referenced whenever possible.  Even when
reasonably reliable series are available, there is the perpetual problem of matching theoretical
concepts in economics to the statistics compiled by governments. Nevertheless, while the precise
magnitudes of the variables are debatable, the general trends are confirmed by the historical
experience as well as by the survey results.
"For a useful description of data sources and causes of discrepancies, see Mabro and Radwan, 1976,
pp. 242-65. For a more recent discussion of the idiosyncrasies of Egyptian economic data, see Hansen,
1988.



REAL PRIVATE IWESrMENT/REAL GDP
0.08
0.07
0.06
0.05
0.04'I
0.03 -
0.02-                                                                                    0
1964   1966   1968   1970  1972  1974   1976  1978  1980  1982  1984  1986
YEAR
°   PVI/GDP



- 19 -
The defiaitions, derivations, and sources of the variables used in the equations that follow
are available in the data appendix to this paper.  A few comrnents concerning their definitions
are made below, but the detailed derivations are provided in the appendix. All data are annual,
expressed in real 1980 prices and are in logarithms.30  In the discussion that follows, a "D"
before a variable name represents a differenced variable where the lag operator, (I-L) is used.
The first lag of a variable is represented by (-1) and the second lag by (-2).  The following
acronyms are used:
PRIVI         =      private investment
GDP           =      GDP at factor costs (non-oil)
R             =      real interest rate
RIW           =      real interest rate/average wage
MARKUP   =   markup
ICOSTS        =      relative price of investment goods
PRVCRD        =      private credit
PVCRDY        =      private credit/GDP
GVIINF        =      government investment in infrastructure
An investment deflator was anstructed using a weighted average of the investment
components of the wholesale price index (domestic machinery, imported machinery, construction,
and transport equipment) with variable weights based on actual shares of these inputs in total
investment costs.    Because the machinery component of the official WPI only includes
30 It is conventional in the econometric literature on investment to express all variables in real terms
since the investrment rwccess is perceived as a "real" phenomenon. This view was challenged by Anderson
who argued for the Lse of nominal prices since signals are transmitted in nominal terms and it is not
possible to accurately represent the process by which agents translate these signals into a real expenditure
framework.  However, Bean has pointed out that using a nominal framework implies that all price
movements arc; unanticipated, whereas only divergences of actual from expected prices should matter.
In a developing economy accustomed to a fairly steady rate of inflation as in Egypt, economic agents are
likely to anticipate inflationary trends. Consequently, Anderson's nominal framework seems inappropriate
and all variables have been expressed in real terms. For a discussion, see Anderson, 1981, p. 89 and Bean,
1981, p. 104. In a hyperinflation economy, nominal prices mnay become important signals for investors.
See Oks, 1987 for an analysis of the Argentine experience.



- 20
domestically produced capital goods (which only constitute about 200/o of total machinery inputs),
a separate weight was given to the price of imported capital goods. This was constructed by using
a price index for machinery exports of the major industrial countries, Egypt's major trading
partners, multiplied by the parallel market exchange rate in Egypt.  Thus, the exchange rate
enters as an explicit determinant of investment costs in order to reflect its importance in
determining the price of imports. The resulting investment deflator was used to put private and
public capital formation in real terms and to analyze the evolution of investment costs.
In order to test the microfoundaticns, empirical proxies were needed to represent profits
which are a function of demand, costs and the markup.  The demand proxy used is the non-oil
gross domestic product. Revenues from petroleum were removed to avoid double counting since
they accrued to the government and did not act directly as a source of private demand. Instead,
the effect of oil rents operated through the government budget rather than directly through the
accelerator. The effect of remittances of migrant labour on demand is ambiguous. Remittances
are repatriated to Egypt either in the form of financial assets or, possibly more importantly, in
kind as imports of goods. Financial remittances held in Egyptian pounds (LE), foreign exchange
accounts in Egypt are invested by the banks in the Eurocurrency market and imply no net inflow
of foreign exchange to Egypt, although interest incorme from abroad does accrue to the migrant
investor and a commission is earned by the bank. Financial remittances held in LE, however,
have the effect of increasing domnestic credit and the country's net foreign exchange reserves.
In contrast, remittances repatriated in the form of goods have a dampening effect on domestic



- 21 -
demand since they substitute for domestic production.  Consequently, GDP, rather than GNP
which includes some estimate for rermittances, is used here as the preferred proxy for demand.3
Two different empirical representations of the cost of capital goods have been considered.
Some elements of the theoretical representation of the cost of capital defined in equation (2) will
not be considered in the empirical work for Egyp. As with most empirical analyses of investrnent,
the effect of the rate of appreciation of capital goods (z) is neglected because of the absence of
data and the fact that without an active second hand market for machinery, this capital gain
cannot be readily realized. The effect of taxation (u) and investment incentives (i) will not be
included, again for lack of data and because they are relatively unimportant because of widespread
tax evasion and the introduction of tax holidays under the "open door" policy Data on the rate
of depreciation of the capital stock is not available and the practice of using a constant rate as
a proxy will have no effect on the econometric results.  Although the view that depreciation is
an economic variable that depends on the firm's scrapping and maintenance decisions is more
attractive, it is empirically intractable in most countries.3'
The elements of the cost of capital that will be evaluated directly will be the cost of credit,
(r-p), the relative cost of factors, (RAW) and the relative price of capital goods,
[(I-O)PkD+flePkw]. The cost of credit is proxied by the WPI-deflated discount rate (R)." The
3' Although the governrnent tries to estimate the value of remittances, including those in kind, it is
generally believed that the official statistics are underestimates.
32 See Nickel], 1978 for a discussion.
33The discount rate is an adequate proxy for borrowing costs since the difference between the two
has been fairly constant as a result of central bank regulation of fees and comrnissions.



- 22 -
relative cost of factors is represented by the ratio of the real interest rate to the average wage in
the economy (R/W). The cost of capital term does not take into account the implications of
compensating balanc-,s for the effective cost of borrowing.  Since the practice of requiring
compensating balances is not legal, there are no data available to evaluate the impact on real
borrowing costs. In the absence of data in the Egyptian case, it was necessary to assume that the
degree to which compensating balances respond to higher inflation is constant and therefore will
have no effect on the coefficient estimates.
The relative price of capital goods is represented by the variable ICOSTS which is based
on the ratio of the investment deflator to the GDP deflator. Treating the cost of investment
goods separately from the cost of borrowing is desirable because it isolates the effect of
neoclassical price factors from Keynesian considerations about the interaction of demand and
supply in the capital goods market. In effect, the ICOSTS variable operationalizes Keynes'
marginal efficiency of capital for an economy where, because of credit market imperfections, it
is distinct from  the interest rate.34   For traded capital goods, supply is highly elastic and
therefore the price to firms depends solely on the world price and the exchange rate.  However,
for nontraded capital goods, supply is more inelastic and one would expect considerable increases
in the price of construction and land, for example, in the case of an investment boom. The use
of variable weights in the investment deflator also captures the effect of relative prices on
changing shares of tradable and nontradable capital goods. It is hypothesized that ICOSTS is a
34 Recall from the standard presentation of Keynes' model, investment occurs until the marginal
efficiency of investment is equal to the rate of interest. However, this equilibrium relationship emerges
out of the interaction between demand and supply in the capital goods producing industry. Because the
supply function for capital goods is upward sloping, the response of investment to a change in the rate
of interest is gradual. For a discussion, see Precious, 1987.



- 23 -
more realistic representation of the cost of capital to the firm in a financially repressed developing
economy than the neoclassical interest rate variable.
The movement of the relative cost of investment goods is depicted in Figure 2. The
downward trend in the price of investment goods after the 1967 war reflects some of the early
attempts of the government to encourage private investment. Investmnent incentives, such as
subsidies to buildings materials and machinery, and the real appreciation of the exchange rate
during the oil windfall had the effect of reducing the relative cost of investment goods. This is
consistent with the survey findings where firms, especially those established during the "open door
policy," were found to be characterized by greater capital intensity.  After 1980, the price of
investment goods rose sharply reflecting the increasing price of both imported investment goods
subject to a depreciating exchange rate and non-traded investment goods responding to growth
in demand. This pattern in the movemnent of investment costs had important implications for
firms' decisions about factor shares, an issue that will be discussed at a later stage.35 The proxy
used for markups is the ratio of the wholesa'e price index to an index of wages in the economy.
Although this proxy does not capture the complexity of mark up determination described above,
it does provide a crude indicator of the evo;ution of the profit rate at the aggregate level.
Alternatively, this mark-up variable can be interpreted as the irverse of real wages. Figure 3
depicts the movement of narkups over the period. Markups were relatively high during the 1960s,
a pattern that coincides with the survey findings that, for firrns that survived the nationalizations,
the 1960s were a highly p-ofitable period. Specifically, the absence of competition, subsidies to
inputs, relatively low wages and considerable unsatisfied demand meant that firms were able to
35 Evidence about firms' technological choices in favour of greater capital intensity emerged from the
survey. See chapter 4 in Shafik, 1989 for a discussion.



RELATIVE PRICE OF INVESTMENT GOODS
(INVESTMENT DEFLATOR/GOP DEFLATOR)
0.22 -
0.2-
0.18
0.16
0.14
0.12
0.1
0.08
0.06                                                                                              t
0.04 04,-
0.02-
0 -a
-0.02
-0.04
-0.06
-0.08
-0. 1
-0.12
-0.14-                  *                                                     
1960   62    64    66    68    70    72    74    76    78    80    82    84    86
YEAR
0    ICOSTS



MARKUP BEHAVIOR
(LOG(VPI/WAGE INDEX))
0.7   - s--
0.6 4
0.5 -
0.4 -
0.3                                                                                       1
0.2
0.1
-0.1
-0.2
-0.3                                                                                       0
-0.4
-0.5
-0.6
1960   62   64    66    68    70   72    74   76    78   80    82   84   86
Y'EAR
0   MARKUP



- 26 -
charge a high markup over costs. Part of this markup) can alsc be considered a risk premium
given the highly uncertain environment in which firms were operating. The trend of markups is
generally downward with small upturn during the oil boom of the late 1970s. After 1980, there
is a squeeze on markups as a result of rising wages. This does not imply that markups were
negative after 1980, but only that wages were increasing at a faster pace than were output prices.
An important factor that it has not been possible to capture econometrically is the effect
of protection. The importance of securing protection, often before an investment is made, was
an important theme among the firms surveyed. However, because rates of effective protection are
industry-specific, and often firm-specific, it is virtually impossible to construct a meaningful
indicator of the overall protective regime and its effect on markups. Instead, the implications of
protection for firms' markups were considered at the sectoral level and are available elsewhere.
The justification for introducing government policy variables in the econometrics is that
they may affect costs, markups or demand independently of the empirical proxies used here. For
example, government investment in infrastructure may reduce the costs faced by firms, although
it may not be reflected in the ICOSTS variable. Similarly, government borrowing on the domestic
credit market may reduce credit availability to the private sector in a rationed market directly
although it may have no effect on administered interest rates. The effect of public policy will be
evaluated through both government expenditure variables, particularly public investment in
infrastructure (GVIINF), and through the implications of the financing of the government deficit
on the availability of private credit.
34 Shafik, 1989.



- 27 -
The variable for governrnent investmnent in infrastructure is the sum of public investment
in agriculture, irrigation, electricity, transport, construction and utilities.3   The evolution of
government investment expenditure in infrastructure (GVIINF) and in other areas is depicted in
Figure 4. There is a general decline in public investment in the wake of the 1967 war with a
recovery during tbe windfall period of 1975-80. Aggregate investment and that in infrastructure
grew as a share of GDP during the oil boom of the 1970s. The squeeze on public investment did
not occur until the early 1980s and seems to have fallen disproportionately on infrastructure and
industry
In order to evaluate the potential rationing effect of government administered interest
rates, quantity variables for credit will be considered in addition to the more conventional real
interest rate term. The quantity of credit to the private sector, both the level (PVCRD) and as
a share of GDP (PVCRDY), will bea considered. The quantity of credit is likely to be important
in a credit market where interest rates are subsidized, balance sheets are unreliable,  and
reputation is an important determinant of access to bank credit. In addition, the quantity of
credit captures the effect of financial remittances held in Egyptian pounds, which may be an
important factor in investment determination.  The more commonly discussed channel for
crowding out, the government deficit, will also be considered. The conventional view is that
deficit financing bids up interest rates which reduces private capital formation and results in
" Total government investment and non-infrastructure public investment were also tried as explanatory
variables but were found to be insignificant. Non-infrastructure investment was defined as the residual
from total investment which consisted of government investmnent in industry, petroleum, trade and finan .e,
housing, and services. This in part reflects the very long lags associated with public investment in areas
such as health and education services.



COMPOSITION OF GOVERNMENT INVESTMENT
(IN REAL TERMS AS A SHARE OF GDP)
0.28-
0.25, 
0.24                                                                                                .sX
X~~~~~~~
0.22                                                        X0
0.2 
0.18                                                                                                 5  
0.16                                                                             X                   0
0.14                                                                                                 8 N00
0.12 
0.1                           X
0.06                                                                                                 r
0.04
0.02       X
0-                  4
1965     67      69       71      73      75      77       79      81      83      85
YEAR
Z JIN FPA           EZS   OTHER



- 29 -
indirect crowding out.38   However, in a rationed credit market with administered prices, the
effect of deficit financing will be on the quantity of credit rather than on the price and therefore
is more likely to be reflected in the PVCRD variable.
3.2. Econometric Estimation: Methodology
Investment functions are notoriously difficult to estimate, often for two very different
reasons. Many of the macroeconomic time series that are relevant to investment decisions, such
as output, are trended and tend to generate spurious regressions. In addition, other v.Ariables that
are expected to matter, such as the interest rate, are often not significant because they are
stationary whereas investment is usually characterized by a trend.
Granger and Newbold recommend differencing data when spurious correlations are
suspected."  By differencing, stationarity of the time series is more likely and more mneaningful
parameter estimates can be obtained. However, simply differencing all of the timie series is both
ad hoc ar,J results in the loss of information about the equilibrium relationship betwecn the levels.
In add:tion, it is possible to introduce a trend into already stationary timne series by indiscriminate
diffe:encing.
3'There is also an indirect channel for crowding in when bonds are closer substitutes for money than
they are for capital.  The resulting portfolio effects as agents switch into capital reinforces the
expansionary effect of a fiscal stimulus. This result is obtained by B. Freidman in a three asset model
with bonds, money and capital. B. Freidman, 1978, 1985. Given the absence of an active bond market
in most developing economies, this channel for indirect crowding is not very relevant.
39 Granger and Newbold, 1974.



- 30 -
As an interesting aside, some ol the standard models of investrment in the literature
(accelerator, flexible accelerator, neoclassical, putty-clay, partial adjustment, and profits models)
were estimated in the levels on Egyptian data.40 The results in the levels indicated that some
combination of putty-clay, profits and partial adjustment wt,.ld produce a well-fitting investment
function. However, once the data were differenced, virtually all of the models collapsed. This
implies that what has been interpreted as causality in a number of empirical studies of investment
may have been spurious correlations between trended variables.
The recent literature on cointegiation and stationarity testing provides a more rigorous
framework for avoiding spurious regression while retaining long run information about the
equilibrium relationship in the levels. Essentially, the intutition behind cointegration is that
econometric results are legitimate only when time series are stationary.4' Therefore it is
necessary to test the time series properties in order to deternisne what degree of differencing, if
any, is necessary to de-trend the data. Once stationarity is achieved, if somne linear combination
of the variables results in a "white noise" error term, the series are said to be cointegrated. This
implies that it is possible to explain the evolution of the time series through the interaction of a
set of non-trended data that results in an error term that is random, thereby leaving nothing left
to explain econometrically.
40 See Shafik, 1989 for these results.
4' For a survey of the literature, see a special issue of the Oxford Bulletin of Economics and Statistics
with articles by Hendry, 1986; Granger, 1986; Hall, 1986; Jenkinson, 1986; as well as work by Dolado and
Jenkinson, 1987; and Engle and Granger, 1987.



- 31
3.4. Stationarity Testing
Dickey-Fuller (DF), Augmented Dickey-Fuller (ADF), and the Cointegrating Regression
Durbin-Watson (CRDW) test proposed by Sargan and Bhagrava were used to test whether
variables were stationary (1(0)) or needed to be first differenced (1(1)) or second differenced (1(2))
to induce stationarity.42  The Dickey-Fuller test where the null hypothesis is a simple unit root
(1(1)) takes the form:
n
a Xt =  Y    + ,Eaj AX.j + et  where H: 1(1).
'Where the null hypothesis is 1(2), the test statistics is:
P,  n
AAXt =           +AX, +Ea j AAXIj + et where H: 1(2).
J=l
The test statistics is the standard "t" test on the lagged dependent variable (p). Because
the test is sensitive to whether a drift (C) andlor a time trend (T) are included, it was repeated
in different forms for each variable. The Augmented Dickey-Fuller test includes second and third
lags of the left hand side variable to capture any additional dynamics. The critical values for the
ADF test are the same as those for the DF test.43
42 Dickey and Fuller, 1979; Dickey and Fuller, 1981; and Sargan and Bhargava, 1983.
4 Engle and Granger, 1987, p. 269.



- 32 -
The Cointegrating Regression Durbin-Watson test is the standard Durbin-Watson statistic
that results from regressing the difference of the variable on a constant when the null is l(l) and
the second difference on a constant when the null is 1(2). The need to try an array of test
statistics reflects the low power of the alternative tests of stationarity and the evolving nature of
this literature.
The results of the DF, ADF and CRDW tests are presented in Table 1. The DF and ADF
tests are reported separately for regressions with only the lagged dependent variable and with the
addition of a constant term (C) and with a time trend (T). The results indicate that the majority
of the time series have simple unit roots, or are l(l). The -only exception is the real interest rate,
R which is I(O). However, the relative price of factors, R/W, is l(l).
The fact that the time series have simple unit roots is analytically convenient since
stationarity is achieved by first differencing. The levels of the series can be used to express the
long run equilibrium relationship by which agents are adjusting their actual to desired capital
stock.  Because their target capital stock is changing over time, agents correct for their
expectational errors in the levels terms.
Two different methods for estimating an error correction model with cointegrating series
will be used below.  The first will be a two step procedure advocated by Engle and Granger
which tests for cointegration at the levels stage before considering the dynamic properties."
The validity of the second stage dynamnic results depends on having an appropriate specification
"Engle and Granger, 1987.



- 33 -
Table 1: TESTING FOR UNIT ROOTS: DICKEY-FULLER (DF), AUGMENTED
DICKEY-FULLER (ADF) AND COINTEGRATING REGRESSION
DURBIN-WATSON TESTS (CRDW)
VARIABLE                OF         DF         AOF        ADF   DF W/ C    DF W/ C  ADF WI C
H10:I(1)   HO0:I(2)   HO0:I1()   HO:I(2)   110:I(1)   H10:I(2)   H10:I(1)
PRIVr                 2.85      -3.48       1.88       -2.81      -0.75       -4.30      -0.46
GDP                   3.39      -3.54        2.32      -3.13       0.13       -4.81        0.82
RI                   -3.83      -6.99      -3.51       -6.49      -3.75       -6.85       -3.51
R/W                   0.31      -5.87       0.31       -5.30      -1.61       -6.18       -1.47
MARKUP               -0.69      -2.53      -0.80       -1.99       0.53       -2.95        0.77
ICOSTS               -0.79      -4.67      -0.78       -3.77      -0.96       -4.58       -0.85
PVCRD                 2.74      -4.20       1.51       -3.54      .1.22       -5.21        0.88
BYCRDY               -1.01      -5.34      -1.02       -4.79      -0.02       -5.48        0.16
GVIINF                0.19      -3.36       0.53       -2.92      -1.04        -3.26      -2.41
CRITICAL VALUES   t-2.61                                         t.3.20
ADF W/ C DF W/ CST DF W/ C&T ADF w/ C&TADF W/ C&T             CRDW       CROW
H0:1(2)   H0:I(1)   H0:I(2)   H0:I(1)   H0:I(2)          HO:I(1)   HO:I(2)
PRIVI                -3.30      -1.75      -4.21       -1.74      -3.21         0.04       1.76
GDP                  -4.44      -1.55      -4.79       -1.86      -4.89         0.05        2.00
R                    -6.30      -4.00      -6.69       -3.46      -6.15         1.46        2.68
R/W                  -S.29      -3.58      -5.91       -3.08      -5.39         0.49        2.59
MARKUP               -2.23      -1.32      -2.64       -1.34      -1.98         0.06        1.10
ICOSTS               -3.65      -0.60      -5.03       -0.44      -5.15         0.32        1.70
PVCRD                -4.47      -1.00      -6.24       -1.59      -5.46         0.06        2.16
PVCRDY               -5.01      -1.54      -6.11       -1.76      -5.42         0.22        2.25
GVIINF               -2.93      -0.80      -3.22       -0.71      -3.95         0.17        1.57
DBFGD                -S.84      -2.82      -8.04       -4.11      -5.67         0.33        2.87
GOVEX                -2.37      -1.58      -2.73       -2.54      -1.95         1.33        2.47
GOVI                 -2.02      -0.88      -2.83       -1.20      -1.94         1.09        2.29
NrGVEX               -3.92      -1.44      -4.61       -0.74      -4.35         1.98        2.71
CRITICAL VALUES                t-2.85                                      CRDW-1.07
NOTE: Wl C IS WITH A DRIFT TERM; W/ C.T IS WITII OOTII A DRIFT TERM   AND A TIME TREND.



- 34 -
at the levels stage. Because of the limitations of the Engle-Granger procedure and the weak power
of cointegration tests, the model will also be estimated using a full dynamic version.  Starting
from the most general unrestricted dynamic equation possible, the model will be reparameterized
until the most parsirnonious version i3 obtained. This data-based approach to modelling stems
from the view that although econormic theory should guide the selection of variables that are
included, the actual model should emerge from the data.4  Some authors have argued that
general dynamic modelling is suierior to the Engle-Granger two stage procedure.46 Rather than
a desire to dive into the methodological debates between econometricians, the purpose of using
two different estimation techniques here is to provide confirmation of the results. Hopefully, by
arriving at a similiar model via two different routes, the validity of the argument will be
strengthened.
Cointegration testing is still at an early stage, so the results must be treated as tentative,
especially given the relatively small sample size. The small number of observations limits the
degree to which alternative lag structures can be explored without causing problems with degrees
of freedom. It will be several decades before most developing economies have sufficient reliable
data to be able to estimate these types of models with confidence.  In the interimr  however,
45 See Hendry and Richard, 1983 for a description of this methodology and Bean, 1981 for an
application to investrnent in the United Kingdom.
46 Jenkinson, 1987; Banerjee et al, 1987. The major problem with the Engic-Granger procedure is that
the validity of the dynamic differenced results hinges crucially on the appropriateness of the first stage
levels results, i.e. the equilibrium long run relationship hypothesized.   With unrestricted dynamic
modelling, the choice of explanatory variables is based on empirical significance.  The limitation of
unrestricted dynamic modelling is on the number of explanatory variables that can be included without
losing degrees of freedom.



- 35 -
economic policy must be made and it seems unwise to do it without the benefit of better
econometric techniques.
3.5. Engle and Granger's Two-Step Estimator
The first stage of the Engle and Granger procedure involves exploring the levels or
equilibrium part of the error correction model to establish whether the variables cointegr. te.
Evidence of cointegration includes an R2 that is close to unity at the levels stage, significant
coefficients47, a significantly non-zero Cointegrating Regression Durbin-Watson statistic, and
significant Dickey-Fuller and Augmented Dickey-Fuller tests on the residuals from the levels
regression. With cointegrating variables, the coefficient estimates from this levels regression can
be interpreted as the long run multipliers. The second stage involves running regressions using
stationary time series (in this case, first differences) and including the lagged residuals from the
levels regressions as an explanatory variable. This lagged residual term, RES(-I), is intended to
capture the error correction process as agents adjust for expectational errors about the equilibrium
relationship in the previous period.
The first stage cointegrating levels regressions for investment are presented in Table 2.
Equation I represents the simplest version of the modei presented above with no explicit
government policy variables.  All of the variables are significant and appropriately signed and
" Note that because of autocorrelation of the residuals, the "t" statistics from the levels regression
are biased upwards and therefore it is not possible to assess the true significance of the coefficient
estimates.  Howevc., it is possible to accept the insignificance of coefficients at the levels stage since if
a variable is insignificant when "t" statistics are upwardly bias, it will certainly be insignificant for the
true value of the "t" statistics.



- 36 -
Tablc2: COINTEGRATING VECTORS FOR INVESTIENT (LEVELS REGRESSIONS)
VARIABLE    (1)        (2)      (3)      (4)      (5)      (6)       (7)      (8)
C  -32.54  -23.62  -29.88  -19.42   -9.05  -40.28   -32.66  -27.79
(6.90)  (3.36)  (7.11)  (3.59)  (2.35)  (5.76)   (4.08)  (3.22)
GDP    4.16      2.79     4.02    2.70      1.26      5.02     4.29      2 q7
(8.18)  (3.03)  (9.05)  (3.88)  (2.97)  (6.56)   (5.07)  (2.62)
ZCOSTS   -1.89   -1.14   -2.62   -1.94   -1.58   -2.19    -2.37   -1.79
(3.54)  (1.57)  (4.96)  (3.29)  (2.43)  (3.03)   (3.41)  (2.40)
KARZUP    1.23      0.51     1.72    1.33                2.20      1.96     1.20
(2.53)  (0.78)  (3.77)  (2.45)               (2.19)   (2.03)  (1.17)
R/W                                                  1.14    -0.01      3.29
(0.75)  (0.003)  (1.99)
GVINP               0 .54             0.36     0.71                         1.06
(1.79)           (1.53)  (3.39)                        (2.50)
PVCRDY                       0.93    1.35      1.00               0.76
(2.90)  (3.73)  (2.63)               (1.70)
DBFGD
GO VEX
R2    0.95      0.94     0.91    0.97      0.95     0.94      0.95     0.94
R2(ADJ)    0.95      0.92     0.96    (.96    0.94       0.92      0.93      0.92
CRDW     1.53     1.40    1.47    1.80       1.24     1.27      1.23     1.25
F  151.90   64.96  152.71   94.17   89.75    65.52    58.78   35.98
DF   -3.95   -3.73   -4.06   -5-09   -3.85    -3.00    -3.02   -3.15
ADF(2)   -2.86    -2.08    -3.54   -4.17   -3.08    -2.33    -2.24    -2.06



37 -
the cointegration statistics are promising.   Equation 2 considers the effect of government
infrastructure investment and results in a significant coefficient as well as positive indications of
cointegration.  Similarly, the quantity of private credit which is included in equation 3 is
significant and generates favorable cointegration statistics. This wouid seem to imply that there
was some rationing in credit markets that served to crowd out private investment.
On the surface, this result would appear to be inconsistent with the survey findings and
interviews with banks that credit markets were very liquid throughout much of the - .iod, except
after the imposition of credit ceilings by the Central Bank as part of a reform package negotiated
with the International Monetary Fund in 1987. Established firms never complained about a
shortage of credit prior to the imposition of ceilings, implying that government borrowing did not
crowd out some of the private sector through the financial system. In fact, some firms complained
that the banks put pressure on them to borrow more. Because real interest rates were negative
throughout the period, established flrms did tend to borrow heavily since credit was, in some
sense, a "free good." However, credit was rationed at the margin, especially for new, poorly
connected firms. Government borrowing may have crowded out these new borrowers but, given
the conservatism of the financal system, it is not clear that the banks would have extended them
loans anyway.
Other government spending variables, such as total government investment and non-
infrastructure government investment, were also tried in all possible combinations and always
appeared with an insignificant sign and generated no improvemnent in the cointegration



- 38 -
statistics.48 Given that "t" statistics are biased upwards when positive autocorrelation exists,
insignificant coefficients at the levels stage imply that these variables should be omitted.
The combined effects of government investment in infrastructure and rationing in credit
markets are considered in equation 4. The resualing levels equation has the best cointegration
statistics as evidenced by the highly significant DF, ADF, and CRDW  tests.   Equation 5
illustrates the importance of the markup variable by indicating the poor performance of an
equation that does not include markups.  The preferability of expressing the costs of capital as
d function of the cost of investment goods (ICOSTS) and quantities of credit (PVCRDY) over the
more conventional factor price variable, RJW, is evidenced by equations 6 and 7. Factor costs
are insignificant when they are included with ICOSTS in equation 6 and with PVCRDY in
equation 7. The only case where the more conventional neoclassical representation of the cost of
capital is significant is in equation 8 where it appears with a positive sign. This is not surprising
in a rationed credit market where a rise in the real interest rate generates a larger quantity of
credit available to investors through the banking system. Consequently, the significance of R/W
in equation 8 seems to be merely oerving as a proxy for the quantity of credit variable, PVCRDY.
The preferred specifications from these levels results are equations 3 and 4 which include
the quantity of private credit and government investmei,. in infrastructure.  The coefficients are
significant and appropriately signed. The equations indicate that in the long run, the accelerator
has the greatest effect on investment, followed by the cost variable (ICOSTS). The markup is also
48 The insignificance of these other government spending variables also occurred when they were
exoress d in differences. The results are not reported here because, strictly speaking, if there is no long
run relationship at the levels stage, one would not expect to find significant effects in the dynamics. This
was consistently so for these variables.



- 39 -
important with a long run coefficient that is over one. The magnitude of the effect of the
quantity of credit is greater than that for government infrastructure investment, but both are
significant. The cointegration tests are favourable implying that the error is "white noise."  As
in the only other known application of the Engle and Granger procedure to investment by Henry
and Minford for the United Kingdomn, a fairly complex specification was necessary to find
evidence of cointegration for investment.49   This suggests that previous studies that relied on
fairly simple models of the determinants of investment may have been misspecified.  These
preferred specifications will be used for the second staige of the Engle and Granger procedure to
explore the dynarmics of the investmnent process.
The differenced dynamic equations for investment are reported in Table 3. The residuals
fro i the levels regressions are included in lagged formn, (RES(-I)), to capture the process by which
agents adjust to prediction errors in the last period.  Ideally, it would be possible to include
several lags of each differenced variable and to repararnaterize according to significance.
However, in order to preserve degrees of freedom, only one lagged difference was included.
lhe results of running an unrestricted version of equation 3 from table 2 are reported in
equation I of Table 3. Reparameterizations based on significance are reported in equations 2, 3,
and 4. A similar exercise that includes governrnent investment in infrastructure is repeated in
equation 5 with reparamaterizations reported in equations 6, 7, and 8 of Table 3.  The
reparameterizations allow the alternative lag structures to be defined by the data. In general, the
variables have significant and appropriate signs and the diagnostic statistics are good. In
49 Henry and Minford, 1988.



- 40 -
Table 3: DYNAMIC EQUATIONS FOR INVESTMENT (DIFFERENCE REGRESSIONS)
(Dependent variable is the difference of the
logarithm of real private investment)
VARIABLE        (1)        (2)       (3)       (4)        (5)       (6)        (7)        (8)
C      0.08      0.09      0.08       0.04      0.05       0.06       0.01     -0.05
(1.41)  (1.89)*   (1.42)    (0.61)    (0.49)    (1.16)    (0.08)    (0.86)
DGDP      1.97       1.39      1.34      1.55       1.90      1.61       1.88       2.65
(2.44)*" (2.19)** (2.03)**  (1.77)*   (1.37)  (2.14)** (2.64)** (3.26)**
DGDP(-1)      -0.72                                    -0.84
(0.88)                                   (0.96)
DXCOSTS      -2.00    -1.45    -'.44    -1.88         -1.75      -2.11      -2.20     -2.68
(2.44)**  (1.96)*  (1.92)* (2.08)**   (1.29) (2.65)** (2.97)** (3.07)**
DICOSTS(-1)       0.79                                      1.32       1.17      0.87
(1.10)                                   (1.58)  (1.79)*    (1.39)
DMARXUP       0.64                1.30        0.9      0.25                             1.01
(1.12)            (2.69)**   (1.64)    (0.32)                            (0.61)
DMARKUP(-1)       0.86       1.36                           0.91       1.11       1.05
(1.59) (2.93)**                          (1.44) (2.37)** (2.41)**
DGVIINF                                                0.21                            0.11
(0.89)                          (0.61)
r1GVIINF(-1)                                                0.18      0.43       0. 20
(0.47)  (1.82)*    (0.77)
DPVCRDY       1.15      1.16      1.11       0.80      1.35       1.10       1.27       1.27
(3.78)** (3.95)** (3.84)** (2.78)** (4.05)** (3.62)** (4.26)** (4.81)**
DPVCRDY(-1)       0.44       0.62      0.56                 0.27
(1.42) (2.54)** (2.00)**                 (0.73)
DPRIVI(-l)       0.19                 0.11      0.37       0.22                 0.33       0.50
(0.86)              (0.53)  (1.76)*   (0.66)                 (1.72)* (3.11)**
RES(-1)     -0.61      -0.39    -0.43    -0.63        -0.72      -0.52      -0.85     -1.33
(2.58)** (2.23)** (2.19)** (2.60)**    (1.09)   (1.81)* (2.60)*"  (4.82)**
R2      0.64      0.56       0.56      0.42       0.86      0.73       0.78      0.76
R2(ADJ)       0.39      0.41      0.38       0.22      0.62       0.57       0.63      0.62
SE      0.17      0.17       0.17      0.19      0.14       0.15       0.14      0.15
DW      1.73      1.74       1.86      1.89      2.10       1.56       1.91       1.66
F      2.53      3.77      3.14       2.16      3.61       4.54       5.00      5.43
DF    -5.14    -4.65    -5.03    -4.65          -4.78      -3.46      -4.22     -3.63
ADF(2)     -4.21      -3.88    -4.19    -4.30        -4.06      -2.43     -3.39      -3.16
CHOW                 0.31                1.21                 5.10      10.78
NOTES: -NUMBERS IN PARENTHESES ARE T STATISTICS.  ONE ASTERISK IMPLIES SIGNIFICANCE AT
THE 10% LEVEL.   TWO ASTERISKS IMPLY SIGNIFICANCE AT THE 5% LEVEL.



- 41 -
differenced form, the coefficient on the ICOSTS variable has the greatest magnitude, followed by
the accelerator (DGDP) and the markup. The coefficient on government infrastructure investment
is smaller and significarntly positive in equation 6. C ven the long lead time on investment in
infrastructure, it is not surprising that the effects on private investment are initially small in
magnitude.
These results of the second stage of the Engle-Granger procedure provide strong evidence
of the appropriateness of an error correction framework. The lagged residuals from the levels
regressions, (RES(-I)), which represent the equilibrium error term, are always significant, implying
that an error correction mechanism exists whereby agents adjust their expectations to
unanticipated changes. This implies that the equations have a long run or equilibrium solution
in the level of investment. The lagged private investment term is significant in equations 4, 7, and
8. Given the limited scope for exploring further lagged effects, it is likely that the coefficient on
lagged private investment is capturing the effects of further lags of the right hand side variables
that cannot be included separately in the regressions.
The overall fit of the equation is evidenced by the plot of the actual evolution of private
investment and that predicted by equation 2 of Table 3 that appears in Figure 5. -  The fitted
values for private investment are very close to the actuals over the 1960-86 period.  This is
remarkable given the array of shocks during this period- two wars, two oil shocks, and a
fundamental change in economic policy orientation.  Chow tests for paramneter stability were
constructed for the best regressions in Table 3, equations 2, 4, 6 and 7, to test for a structural
break with the introduction of the "open door policy" reforms in 1974. The results of the Chow
5Note that Figure 5 depicts private investrent in first differences.



ESTIMATED PRIVATE INVESTMENT
FROM ENGLE AND GRANGER PROCEDURE
0.6
0.5
0.4
0.3
0.2
-0.2
62     64     66     68    70    72       74    76    78    80    82    84    86
YEAR
O    ACTUAL              +    FlTED



- 43 -
tests, also reported in Table 3, show considerable parameter stability for equations 2 and 4 during
the period. This implies that the coefficient estimates for equations 2 and 4 in table 3 were not
significantly different in the 1962-74 period from those in the 1975-86 period. These Chow test
results are encouraging since they indicate that the underlying determinants of investment in the
economy, or "deep parameters," have been adequately captured.    These results from  the
Engle-Granger procedure will be compared with those from general dynamic modelling below.
3.6. Unrestricted Dynamic Estimation
The results of the unrestricted dynamic equations are reported in Table 4. Again, because
of the limited number of observations, it is not possible to include several lags of all of the
variables.51  Instead, one lagged difference of each variable was included along with a lagged
error correction term that reflects the relationship in the levels:
ecm = (YK - Ex,)
" This may be an advantage of the Engle-Granger procedure when dealing with small samples. Engle
and Granger suggest estirmating the simplest error correction model initially and then considering the
effects of lags. Such an approach does preserve degrees of freedom, although the validity of the results
hinge on having the correct specification at the first stage. Engle and Granger, 1987.



- 44 -
Table : UNRESTRICTED DYNAMIC EQUATIONS
(Dependent variable is the difference of the logarithm of
real private investment)
VARIABLE        (1)       (2)        (3)       (4)
C    -1.32    -2.10    -0.12         -0.13
(0.73) (2.26)**   (0.67)    (0.75)
DGDP      1.02       1.06      1.41      0.83
(1.12) (2.15)**   (1.63)    (1.43)
DGDP(-1)      -0.71                0.47
(0.65)              (0.53)
DICOSTS     -0.29                -0.68
(0.32)              (0.74)
DICOSTS(-1)       1.44       1.00      0.43
(1.69)   (1.92)*   (0.53)
DMARKUP    -0.06                  0.44
(0.08)              (0.68)
DKARKUP(-1)       0.96      0.88       0.87      0.93
(1.38)   (1.95)*   (1.36)   (1.86)*
DGVIXNF       0.36      0.34
(1.57) (2.16)**
DGVIINF(-1)       0.41       0.33
(1.52)   (1.84)*
DPVCRDY       1.12      1.05      1.07       1.10
(3.64)** (4.30)** (3.08)** (3.67)**
DPRVCRDY(-1)      0.66       0.74      0.87      0.81
(1.87)* (3.50)** (2.16)** (2.97)**
DGO VEX
DGOVEX(-l)
ONIGVEX
DNIGVEX(-1)
DGOVI
OGOVI(-1 )
DPRIVI(-1)      -0.07                -0.10
(0.41)              (0.44)
ECM(-1)     -0.16    -0.25    -0.08         -0.08
(0.80) (2.35)**   (1.38)   (1.77)*
R2      0.85      0.83      0.54       O.S0
R2(ADJ)      0.60       0.71      0.21       0.34
SE      0.15      0.13       0.19      0.18
DW      1.64      1.66       1.59      1.87
F      3.33      6.75      1.62       3.51
DF      -4.1     -4.21     _4.73      -5.29
ADF(2)     -2.16      -2.31     -3.55      -4.11
CHow                 1.13                2.07
NOTE: THE ECM TERM IS DEFINED AS THE DIFFERENCE INI TIHE LEVELS OF TIIE LEFT IIAND SIDE AND RIGHIT
HAND SIDE VARIABLES.



- 45 -
where Yt is the left hand side variable and XY are the right hand side variables.52 Although not
ideal, the lagged difference includes information from two years in the past, which coincides with
the survey findings that the average lag between the conception and implementation of an
investment project was two years. Consequently, although being able to include more lags would
be desirable, given the constraint, the present approach is certainly adequate.
The results in equation I of Table 4 and the resulting reparameterization in equation 2
confirm the importance of government infrastructure investment and the quantity of credit.
Equations were also estimated that consider the effects of non-investment government expenditure
and total government investnent which, like those from the Engle-Granger estimates, always
generated insignificant coefficients. Equation 3 reports the results of including only the PVCRDY
variable along with the core model. The repararneterization in equation 4 implies that only
demand and internal funds variables, i.e. output, markups and credit, matter.
An interesting feature of the results in Table 4 is that the ICOSTS variable takes a positive
sign in equation 2 that is significant at the 10% level. This may seem paradoxical since a rise in
investment costs would usually be associated with a fall in investnm-nt. However, in a highly
oligopolistic market, changes in costs may not matter in the short run since they can be passed
on to consumers. In addition, a rise in investmnent costs may be reflecting, in part, increases in
aggregate demand in the short run. However, the long run effect of a rise in investment costs on
52 This formulation of the error correction term restricts the sum of the coefficients to unity. Ideally,
the coefficients of the levels terms would be estimated directly and then could be included in a composite
error correction term. However, given the restrictions on degrees of freedom, it was not possible to use
this approach.  The validity of the assumption that the coefficients sum to unity implied by this
formulation was confirmed by the significance of the error correction term in the regressions reported in
Table 4.



- 46 -
private capital formation is significantly negative as evidenced by the cointegrating vectors in
Table 2.
As with the previous results, the unrestricted equations indicate the importance of
government infrastructure investment and quantities of private credit, as well as a positive role
for government expenditure.   The error correction  term  is significant in  all of the
reparameterizations, lending support to the model hypothesized as well as to this formulation of
the error correction model. The lagged private investment term is never significant although its
inclusion tends to improve the diagnostic statistics. As before, this may be because of lagged
variables that had to be omitted to preserve degrees of freedom. Equations 2 and 4 perform
particularly well with significant coefficients and cointegration tests that indicate white noise
errors. Figure 6 plots the actual and fitted values for private investment using equation 2 from
Table 4.53 Once again, the model fits very well over the sample period. The Chow tests for a
structural break in 1974 also indicate considerable paramneter stability. Like the previous results
from the Engle-Granger procedure, the parameter stability implied by the Chow tests is an
indication that the consequences of the policy changes between the pre and post infitah periods
have been captured through the economic deterrninants on the right hand side of the regressions.
3.7. An Evaluation of the Results
The best results from the Engle and Granger procedure are equation 2 in Table 3 and for
the generalized dynamic modelling equation 2 in Table 4 perform particularly well. The diagnostic
tests from both procedures are not markedly different, nor are the coefficient estimates. The plot
"Note that Figure 6 depicts private investment in first differences.



0.8           15ApC\+  f
~O.A
3~~~~'       4-O I



- 48 -
of actual and fitted values for the unrestricted version is a slightly better representation of the
data than that from the Engle-Granger procedure.  Both estimates confirm the core model
hypothesized above which identified the role of derand, markups, and costs in investment
determination. The long run multipliers in Table 2 reveal that the accelerator has the largest
impact on private capital formation, followed by investment costs and the markup.  The
cointegration tests indicate that these best equations resulted in approximately stationary error
processes at the first stage and white noise errors in the second.
The two stage and the unrestricted dynamic modelling provide strong evidence on the
crucial role of the quantity of credit. The interpretation of this PVCRDY is as an indication of
rationing in credit markets where banks face administered interest rates and have imperfect
information about borrowers. In Egypt this was particularly problematic because of widespread
tax evasion. As a consequence, banks often had to make decisions about loans without access to
the firm's true balance sheets. Although government borrowing did not crowd out established
private sector firms who had access to abu;dant amounts of credit, there appears to have been
some crowding out at the margin of less established firms and perhaps of potential firms.
Ultimately, this reflected weaknesses in the financial system which operated with highly imperfect
information and in a climate of considerable uncertainty.
The more Keynesian ICOSTS variable consistently outperformed the interest rate variables,
R and R/W. This is because the ICOSTS variable takes into explicit account the interaction of
demand and supply in the capital goods market and includes the important distinction between
tradable and non-tradable capital goods, thereby providing a more realistic measure of the user
cost of capital. The combination of the ICOSTS variable and PVCRDY provides a much better



- 49 -
measure of the cost of capital to the firm in a repressed financial system than the neoclassical
interest rate.
The insignificant effect of interest rates on investment has been a common and often
problematic finding in much empirical work in both developed and developing economies. A
number of explanations have been proposed in the literature to explain why it is not possible to
obtain a significant coefficient for the cost of funds when in theory the interest rate should be a
crucial variable.  These include uncertainty about internal rates of return, unsophisticated
investment decision procedures, the long time frarne of investment decisions compared to short
run fluctuations in interest rates, and the possibility that changes in borrowing costs are
overshadowed by variations in demand.54
In this model, the insignificance of the interest rate is justified by the existence of markup
pricing, the preference for using internal funds for investment financing," and the effects of
"financial repression." In a rationed financial market, the allocation of credit has little to do with
the price of borrowing and much more to do with the quantity.   In the Egyptian case, the
existence of rationing did not imply that credit was in short supply since the banks were highly
liquid during much of the period.  Rat'er, it meant that what credit there was available was
allocated according to non-price criteria such as reputation.
54 See Hay and Morris, 1979, pp. 393-394 for a discussion.
" In addition to firms preferring internal financing of investment, the banks avoid providing funds
for longer term activities because they are not permitted under Central Bank regulations to charge higher
interest rates on longer term loans.  Instcad, the banks prefer to roll over short term loans, which
obviously adds grea:er uncertainty for the firm.



- 50 -
This is not to imply that the interest rate was not important, since its low level was a
crucial factor determining the tendency of firmns to become overindebted, as evidenced by the
frequent overleveraging among the firms surveyed. Rather, it is argued that in a repressed
financial system, the nterest ra e serves a different role than that of clearing the market for
credit.  Specifically, the rcte of interest becomes a mechanism whereby the government taxes
savers on behalf of itself and private sector borrowers.
Neither modelling procedure provided any evidence of crowding out as a result of
government policy. None of the government spending variables had a significantly negative
coefficient. Equation 2 in Table 4 provides strong evidence of the positive effects of government
investment in infrastructure on private capital formation. The long run coefficient on GVIINF
in equation 2 of Table 2 is also significantly positive. This implies that infrastructure investment
is complementary to private activity by reducing costs to firms and therefore induces higher levels
of private investment.
The error correction mechanism is consistently significant, whether as the lagged residual
from the levels regression in the Engle and Granger procedure or as the lagged levels term in the
unrestricted estimates, thereby confirming its appropriateness as a representation of the data. In
general, the error correction model seems to be an appropriate means of capturing the process by
which agents adapt to the highly complex environment in which investment decisions are made.
The plots of the actual and fitted values from the two procedures indicate that the model has
strong predictive content. Without having to rely on ad hoc dunmny variables, it has been possible
to identify the underlying economic processes that determined private investment.



- 51 -
Although the above econometric analysis lends considerable support to the mode' of
investment determination hypothesized, it also highlights the limits of this type of analysis and
the difficulties in assessing the impact of government policy. In theory, by estimating the signs
and magnitudes of public policy multipliers it should be possible to respond to the long-running
debate about crowding out versus crowding in.56 However, at the aggregate level, it is virtually
impossible to prove definitively whether crowding out or crowding in has occurred because of the
problem of constructing the counterfactual.
In order to prove that crowding out occurred, for example, it would be necessary to show
not only that the public sector expanded while the private sector contracted, but also that the
private sector sought to expand and was displaced by the public sector. It would be easy to
conclude that a negative multiplier on a governrment policy variable implied crowding out when
in fact it may be reflecting a possibly wise countercyclical policy on the part of the state. What
it is possible to conclude at the macroeconomic level, however, is what did not happen. For
example, a positive public policy multiplier is a fair indication that crowding out did not occur
as is a negative multiplier a reasonable basis for concluding that crowding in did not occur.
The results above for Egypt provide mixed evidence of crowding out at the macroeconomic
level. The effects of government policy only operated through direct, non-price channels, such
as rationing in credit markets and government investmnent. Indirect channels such as interest rates
5 Buiter concludes from his theoretical survey about crowding out versus crowding in that ultimately
the question is an empirical one that must be addressed through the estimation of public policy multipliers.
Buiter, 1985.



- 52 -
and bond substitutability were not relevant. This is a particular feature of the crowding out
versus crowding in debate in many developing countries.
The government did not crowd out established firms in the credit market and, by
maintaining artificially low interest rates, implicitly subsidized the borrowing of these firms. At
the margin, however, the government may have crowded out newer borrowers and thus contributed
to a lower level of private investment. This may not have been a major economic loss since there
is evidence that the efficiency of much of the private investment that was made was very low. On
the other hand, greater _ccess to credit for newer and smalier firms and the resulting competitive
pressures might have contributed to a more efficient use of investmnent resources.
The economic results also lend some support to crowding in as a result of aggregate
government expenditure and, in particular, to government investment in infrastructure. However,
the positive coefficient on GVIINF may bc reflecting a procyclical stance on the part of the state.
Whether the results are capturing causality or simply the coincidence of government spending
with private investment during a boom period is not clear.
4. Conclusions
This paper has attempted to address some of the weaknesses in the theoretical and
empirical literature on investment. By integrating the microfoundations and macroeconomics of
investment at the empirical level, it has been possible to provide a more realistic model of the
causes of capital formation in a developing economy.  Rather than simply hypothesizing a
theoretical model of the determinants of aggregate investment, the macroeconomic specification



- 53 -
is based on detailed case studies of firm decision-making.   Its application to the Egyptian
experience generated a number of different determinants of investment that might be considered
in other country case studies.
The nicrofoundations revealed the complex channels through which government policy can
affect private activity. While it was not possible to test all of the transmission mechanisms
empirically, the model provides a general framework for analyzing aspects that are relevant to
particular economies. The econometric approach taken to test the model addressed many of the
methodological weaknesses in the existing literature on investment.  Thus, for example, the
common problem of "spurious correlations" in investment functions was addressed by testing the
time series for stationarity and insuring that the residuals from the econometric estimation were
"white noise."
By using two different estimation techniques along with the microfoundations, the validity
of the model hypothesized was conf.rmed further. The use of an error correction approach and
the application of unrestricted dynamic modelling allowed for some of the effects of expectations
and uncertainty on aggregate investment. A comparison of the actual and estimated values for
private investment over the 1960-86 period revealed how well the model fit the data despite the
considerable shocks to the economy over the period.
The limits of econometric testing of whether the government "crowds in" or "crowds out"
private investment and the impossibility of constructing the counterfactual were also discussed.
While it is not possible to conclude whether crowding out or in occurred at the macroeconomic
level, i.e. to accept the alternative hypothesis, it is possible to draw conclusions about what did



- 54-
not happen, i.e. to reject the null hypothesis. In addition, the model provided a framework for
analyzing the effects of government policy by considering explicitly the role of a number of
possible instruments such as 'he exchange rate, the quantity of credit available to the private
sector, and the composition and financing of the government budget. Future research may choose
to test other empirical proxies, such as protection, within the same framework.
The model of investment behavior that emerged had features in common with a number
of different approaches in the literature as well as aspects that are particularly relevant to a
developing economy. Firms' decisions about investment were outcomes of the oligopolistic
structure of markets, putty-clay technology, the inelastic supply of non-traded capital goods and
rinancial repression. The consequences of these factors for investment determination were an
important role for mark ups, internal financing, demand, and the cost of investment goods
defined, not as the interest rate, but as the price outcome from the interaction of supply and
demand in the market for capital goods.
By constructing an index of the relative price of investment goods, it has been possible to
provide a more meaningful indicator of the true cost of capital to the firm under a repressed
financial system. In an economy with a well functioning credit market, the Keynesian equilibrium
condition equating the marginal efficiency of investment with the interest rate is likely to hold.
However, under financial repression or where credit markets are imperfect, the interest rate is not
a true reflection of the cost of capital to the firm. Instead, a combination of the price of
investment goods and the quantity of credit available to the private sector appears to be a more
realistic proxy.



- 55 -
Further exteglsions of the model have interesting implications for a range of stabilization
policies.   For example, policies such as devaluation, credit ceilings and reductions in public
investment, have negative effects on private capital formation.  in the context of protection,
putty-clay tochnology and inelastic demand for imports, the likely outcome of devaluation is a
reduction in investment, at least in the short run until efficiency gains are achieved.5" Similarly,
credit ceilings imposed on a repressed financial system are likely to result in a worsening
allocation of resources since those who possess borrowing power are often different from those
whose projects have the highest returns. The model also provided evidence of the potential
negative impact of fiscal austerity and cuts in public infrastructure investment on capital
formation in the private sector.
Some of the fall in investment that results from such stabilization policies may be desirable
if, for example, more efficient production results. Howevet, the costs of this transition in factor
efficiency may be high because firms are locked into their technologica! choices for long periods.
An evaluation of the efficacy of stabilization policies cannot occur in an institutional vacuum and
must take into account these differential effects of government policy over the short and long run.
The question of how government stabilization policy can facilitate this transition to more efficient
production could be an important extension of this research.
57 See Chhibber and Shafik, 1990 for an analysis of the impact of devaluation for private investment
in Indonesia that cont asts the short versus long run effects.



- 56 -
Appendix A: Data Sources and Derivations
Sources of Data
The data used in the econometrics came from a variety of sources. The sources are listed
below with the variable(s) obtained from them. Note that when a variable is listed with more
than one source, this means that different parts of the time series came from different sources or
that series from a number of sources were cross referenced and reconciled. Note also that a "N"
at the end of a variable's acronym indicates its nominal value. All indices use 1980 as the base
year (1980=100).
For a general discussion of the sources of economic data in Egypt, see the statistical
appendix in Mabro and Radwan.5g  For a discussion of the evolution of the measurement of
capital formation in Egypt, see Radwan.59 A more recent discussion of economic data in Egypt
is available in Hansen.
The sources used were the following:
Central Authority for Public Mobilization and Statistics, Statistical Yearbook, various issues.
- consumer price index, disaggregated (CPI)
- wholesale price index, disaggregated (WPI)
- WPI for domestic machinery (WPIDMACH)
- WPI for construction (WPIBLDG)
- WPI for transport equipment (WPITRANS)
- gross domestic product at factor costs (GDPN)
Central Authority for Public Mobilization and Statistics, Emplovment, Wages, and Hours of
Work, various issues (1968-78).
- number of workers in the private sector
58Mabro, R. and S. Radwan, 1976, pp. 242-265.
5 The commodity flow approach used to measure capital formation in Egypt is based on data for
domestic production, imports, and exports of capital goods. See Radwan, 1974, pp. 74-81.
0 Hansen, 1988.



- 57 -
Commander, S., The State and Agricultural Development in Egypt since 1973, London: Ithaca
Press, 1987.
- average wages in agriculture (AVGWAG)
Ikramn, K. Egypt: Economic NManagement in a Period of Transition, Baltimore: Johns Hopkins
University Press, 1980.
- domestic bank financing of the government deficit (DBFGDN)
- government expenditure (GOVEX)
International Monetary Fund, International Financial Statistics, *arious issues.
- domestic credit claims on the private sector (PRVCRDN)
- discount rate, end of period (DISCR)
- consumer price index (CPI)
- GDP deflator (GDPDEF)
- London interbank interest rate (LIBOR)
- world wholesale price index (WWPI)
International Monetary Fund, "Arab Republic of Egypt - Recent Economic Developments," various
issues.
- private investment (PRIVIN)
- government investment (GOVIN)
- government investment in infrastructure (GVIINFN)
- gross domestic product at factor prices (GDPN)
Mabro, R. and S. Radwan, The Industrialization of Egypt. 1939-73, London, 1976.
- gross domestic product at factor prices (GDPN)
Ministry of Planning, "Follow-up Report to the Five Year Plan," (in Arabic), various issues.
- private investment (PRIVIN)
- government investment (GOVIND
Pick's, World Currency Yearbook, various issues.
- nominal parallel mnarket price of foreign exchange (EBM)
Shura Council, "Investrnent Policy during the Period 1959/60-1982/83," Arab Republic of Egypt,
unpublished report, 1985 (in Arabic).



- 58 -
- government investment by economic activity (GOVIN)
- private investment by econornic activity (PRIVIN)61
- number of workers by economic activity (TOTLAB)
- wages by economic activity (TOTWAGN)
- final consumption at market prices by the public sector (GOVCONN)
- government investment in agriculture and irrigation (GVIAGN)
- governrent investment in electricity (GVIELECN)
- government investment in construction (GVICONN)
- government investment in transportation (GVITRNN)
- government investment in communications (GVICOMN)
- government investment in utilities (GVIUTILN)
- investment costs in machinery (ICMACHN)
- investment costs in buildings (ICBLDGN)
- investment costs in transport equipment (ICTRANSN)
- investment costs in other (ICOTHERN)
- gross domestic product at factor costs disaggregated by economic activity
World Bank, "Arao Republic of Egypt: Issues of Trade Strategy and Investment Planning,"
January 1983.
- gross domestic product at factor prices (GDPN)
- GDP from the petroleum sector (GDPPETRN)
- government deficit (GOVDEFN)
- domestic bank financing of the government deficit (DBFGDN)
- total government tax revenue (GVTXRVN)
World Bank, "Arab Republic of Egypt. Current Economic Situation and Economic Reform
Program," July 1986.
- gross domestic product at factor prices (GDPN)
- GDP from the petroleum sector (GDPPETRN)
- government investment in infrastructure (GVIINFN)
World Bank, "Arab Republic of Egypt: Country Econormic NMemorandum," September 1988.
- gross domestic product at factor prices (GDPN)
- GDP from the petroleum sector (GDPPETRN)
- government investment in infrastructure (GVIINFN)
- government expenditure (GOVEX)
61 The Shura Council data on private and government investment by economic activity were compared
with the figures given in various issues of the Ministry of Planmng's Follow-up Report to the Five Year
Plan. The figures were generally consistent and discrepancies could usually be explained by differences
in definLitions and categories used.



- 59 -
United Nations, Yearbook of National Accounts Statistics, various issues.
- indirect taxes (INDTAXN)
United Nations, Monthly Bulletin of Statistics, various issues.
- export price index for machinery and equipment from the major industrialized  countries
(XUVKIMP)
Derivations
The derivations used for the variables included in the econometrics are as follows in alphabetical
order:
AVGWAG            =    TOTWAGN/TOTLAB
GDP               =     log(GDPNO)
GDPNO             =    (GDPN-GDPPETRN)/GDPDEF
GOVI              =     log[GOVIN/IDEF]
GVEX              =     log[GVEXN/GDPDEF]
DBFGD             =     log[DBFGD/DBFGD]
GVINFN            =     GVIAGN+GVIELECN+GVICONN+GVITRNN+GVIUTILN
GVIINF            =     log[GVINFN/IDEF]
ICOSTS            =     log[IDEF/GDPDEFI
ICTOTN            =     ICMACHN+ICBLDGN+ICTRANSN
IDEF              =     (0.194*WMACH*WP[DMACH)+(0.0806*WMACH*IMACfII)
+(WBLDG*WPIBLDG)+(WTRANS*WPITRANS)
+(WOTHER*WPI)
IMACHI            =     XUVKIMP*EBM
INDTAX            -     INDTAXN/GDPDEF
KP                =     (1-0.043)KP(-I)+PRIVI
MARKUP            =     log[WPUWAGEINDX]
NIGVEX            =     GOVEX-GOVI
PRIVI             =     log[PRIVIN/IDEF]
PROFIT            =     log(GDPNO-TOTWAG-INDTAX)
PVCRD             =     log[PRVCREDN/GDPDEF]
PVCRDY            =     log[PRVCREDN/GDPDEF/GDPNO]
R                 =     log( I+DISCRATE)/(1 +(WPI-WPI(- 1 ))/WPI(- 1))
R(W W                   log(1+R)/(1+AVGWAG)
TOTWAG            =    TOTWAGN/WPI
WBLDG                   ICBLDGN/ICTOTN
WMACH             =     ICMACHN/ICTOTN
WOTHER            =     ICOTHERN/ICTOTN
WTRANS            =     ICTRANSN/ICTOTN



- 60 -
The investment deflator used above was constructed by using a weighted average of price indices
that constitute investment costs. Since the weights based on the Shura Council data were only available
for 1961-82, the shares of investment costs were assuned fixed before 1961 and after 1982."  The price
index for domestic capital goods was the WPI for machinery and equipment which only includes indigenous
capital goods. The indices for construction and transport were also obtained from the disaggregated ' PI.
The WPI was not available in disaggregated form before 1977, therefore the aggregate WPI had to be
used.  The only exception was the price index for construction for which a series was available from
1970.63 A price index for machinery exports of the major industrial countnes was used as an index of
imported capital goods prices.  This was multiplied by the nominal parallel market exchange rate to
obtain an index of imported capital goods prices in Egypt. The parallel market exchange rate was used
since it was the operative rate for much of the period and captures the opportunity cost of foreign
exchange for the period of rationing pnor to 1974.  The relative shares of domestic versus imported
machinery is assumed to be fixed.  The weights are based on estimates from the World Bank that
imported machinery constitutes 80.6% of total machinery inputs.64 This is a fairly plausible figure given
the small size of the domestic capital goods industry and the fact that indigenous and imported machinery
are far from perfect substitutes.
The index of wages used to calculate the markup proxy is based on two sources. For 1960-1981,
the series for average wages in the economy calculated from the Shura Council was used. However, after
1981, data on aggregate wages are scarce since the Central Authority for Public Mobilization and Statistics
is several years behind in compiling the labour force surveys. Consequently, data on agricultural wages
was used for 1981-1986 to construct the wage index.61 This is not an implausible supposition since wages
in the agricultural sector are fairly market-determined and would represent a close approximation to wages
in the private sector.
62The weights used before 1961 were: machinery-0.35; building-0.41; transport-0.15; other-0.09. The
weights used after 1982 were: machinery-0.40; building-0.38; transport-0.16; and other-0.05.
h3 Assaad constructs a Paaschc index for labour, capital and mnaterials in the construction sector.
Assaad, 1989.
"World Bank, 1980.
6eCommander, 1987, p. 92.



- 6 1 -
Bibliography
Anderson, C., 1981, "A New Approach to the Empirical Investigation of Investment
Expenditures," Economic Journal.
Assaad, R., 1989, "A Construction Index for Egypt," rnimeograph, Cairo.
Balassa, B., 1988, "Public Finance and Economic Development," PPR Working Paper Number 31,
The World Bank.
Banerjee, A., J. Doaldo, D. Hendry and G. Smith, 1986, "Exploring Equilibrium Relationships
in Econometrics Through Static Models: Some Monte Carlo Evidence," Oxford Bulletin of
Economics and Statistics.
Bean, C., 1981, "An Econometric Miodel of Manufacturing Investment in the United Kingdom,"
Econoriic Journal.
Behrman, J., 1972, "Sectoral Investment Determination in Developing Economy," American
Econornic Review.
Bilsborrow, R., 1977, "The Determinants of Fixed Investment by Manufacturing Firms in a
Developing Country," International Economic Review.
Blejer, M. and M. Khan, 1984, "Government Policy and Private Investment in Developing
Countries," IMF Staff Papers.
Brown, E., R. Solow, A. Ando, and J. Kareken, 1963, "Lags in Fiscal and Monetary Policy,"
Comrnission on Money and Credit.
Brunker, D., 1985, "Empirical Studies of Investment Behavior: Abstracts," Aus:ralia Bureau of
Industry Economics. Working Paper 34.
Buiter, W., 1985, "A Cuide to Public Sector Debt and Deficits," Econormic Policy.
Chhibber, A. and N. Shafik, 1990, "Does Devaluation Hurt Private Investment?
The Indonesian Case," PRE Working Paper, The World Bank.
Chhibber, A. and S. van Wijnbergen, 19S8, "Public Policy and Private Investment in Turkey,"
PPR Working Paper, World Bank.
Commander, S., 1987, The State and Agricultural Development in Egypt since 1973, London:
Ithaca Press.
Currie, D., 1981,"Some Long Run Features of Dynamic Time Series Mlodels," Econonic
Journal.



- 62 -
Dailami, M., 1987, "Expectation, Stock Market Volatility and Private Investment Behavior:
Theory and Empirical Evidence for Brazil," mimeograph, The World Bank.
Dailami, M., 1989, "Policy Changes that Encourage Private Business Investment in Colombia,"
PPR Working Paper Number 266, The World Bank.
Dailami, M., 1990, "Stock Markets, Cost of Credit and Private Investment. The Korean Case,"
niimeograph, The World Bank.
Davidson, J., D. Hendry, F. Srba, and S. Yeo, 1978, "Econometric Modelling of the Aggregate
Time Series Relationship between Consurer's Expenditure and Income in the United
Kingdom," Econornic Journal.
Dickey, D. and W. Fuller, 1979, "Distribution of the Estimators for Autoregressive Time Series
with a Unit Root," Journal of American Statistical Association.
Dickey, D. and W. Fuller, 1981, "The Liklihood Ratio Statistics for Autoregressive Tiime Series
with a Unit Root," Econometrica.
Dolado, J. and T. Jenkinson, 1987, "Cointegration: A Survey of Recent Developments," Oxford
Applied Economics Discussion Paper number 39, Institute of Econornics and Statistics,
Oxford University.
Eisner, R., 1960, "A Distributed Lag Investment Function," Econometrica.
Eisner, R., 1967, "A Permanent Income Theory for Investment: Some Empirical Explorations,"
American Economic Review, volume 57.
Engle, R. and C. Granger, 1987, "Co-integration and Error Correction: Representation,
Estimation and Testing," Econometrica.
Ersel, H. and G. Sak, "Public Borrowing and the Development of Financial Markets in Turkey,"
mimeograph, Capital Markets Board of Turkey, 1987.
Fazzari, S. and T. Mott, 1984, "The Investment Theories of Kalecki and Keynes: An Empirical
Study of Firm Data, 1970-1982," Department of Economics Working Paper Ser:es 75,
Washington University.
Fazzari, S., R. Hubbard and B. Peterson, 1988a, "Financing Constraints and Corporate
Investment," Brookings Papers on Economic Activity.
Fazzari, S., R. Hubbard and B. Peterson, 1988b, "Investment, Financing Decisions, and Tax
Policy," American Economic Review.
Friedman, B., 1985, "Crowding Out or Crowding In? Evidence on Debt-Equity Substitutability,"
National Bureau of Economic Research Working Paper Series, Number 1965.



63 -
Friedman, B., 1978, "Crowding Out or Crowding in? Economic Consequences of Financing
Government Deficits," Brookings Paper on Economic Acticity. Number 3.
Fry, M., 1980, "Saving, Investment, Growth and the Cost of Financial Repression," WQrld
Development.
Giovannetti, G., 1987, "The Purchasing Power Hypothesis in the Long Run: An Application of
Cointegration," mimeograph, Faculty of Statistics, University of Rome.
Granger, C., 1983, "Co-integrating Variables and Error Correction Models," University of
California, San Diego Discussion Paper number 83-13.
Granger, C., 1986, "Developments in the Study of Cointegrated Economic Variables,"
Oxford Bulletin of Economics and Statistics.
Granger, C. and P. Newbold, 1974, "Spurious Regressions in Econometrics," Jounal of
Econo=tric5
Grossman, S.J. and O.D. Hart, 1982, "Corporate Financial Structure and Managerial Incentive,"
in J. McCall, editor, Thr Economics of Information and Uncertainty. Chicago: University
of Chicago Press.
Haavelmo, T., 1960, A Study in the Theory of Investment, Chicago: Chicago University Press.
Hall, R. and D. Jorgenson, 1971, "Application of the Theory of Optimum Capital
Accumulation," in G. Fromm, editor, Tax Incentives and Capital Spending, Washington,
D.C.: The Brookings Institution.
Hall, S., 1986, "An Application of the Granger and Engle Two-Step Estimation Procedure to
United Kingdom Aggregate Wage Data," Oxford Bulletin of Economics and Statistics.
Hansen, B., 1988, "The Political Economy of Poverty, Equity and Growth: Egypt," unpublished
manuscript.
Hay, D. and D. Morris, 1979, Industrial Economics, Oxford: Oxford University Press.
Hayashi, F., 1982, "Tobin's Marginal q and average q: A Neoclassical Interpretation,"
Econonetrica.
Helliwell, J., 1976, editor, Aggregate Investment, London: Penguin.
Henry, S. and M Minford, 1988, "United Kingdom Manufacturing Investment in the 1980s: Some
Estimates using Error Correction Mod,ls," Discussion Paper number 137, National Institute
of Economic Research.
Ikram, K., 1980, Egypt: Economic Management in a Period of Transition, Baltimore: Johns
Hopkins Press.



- 64 -
Jenkinson, T., 1986, "Testing Neo-classical Theories of ULbour Demand: An Application of
Cointegration Techniques," Oxford Bulletin of Economics and Statistics.
Jensen, M. and W.V Meckling, 1976, "Theory of the Firm: Managerial Behavior, Agency Costs
and Ownership Structure," Journal of Financial Economics.
Jorgenson, D., 1963, "Capital Theory and Investment Behavior," American Economic Review,
volume 53.
Jorgenson, D., 1967, "The Theory of Investment Behavior," in R. Ferber, editor, Determinants
of Investment Behavior, New York: Columbia University Press.
Jorgenson, D. and C. Siebert, 1968, "Theories of Corporate Behavior," American Econonmic
Review.
Kalecki, M., 1971, Selected Essays on the Dynairncs of the Capit,list Economy. 1933-1970,
Cambridge: Cambridge University Press.
King, M., 1977, Public Policy and the Corporations, London: Chapman and Hall.
Mabro, R. and S. Radwan, 1976, The Industrialization of Egypt. 1939-73, Oxford: Clarendon
Press.
Malinvaud, E., 1980, Profitability and Unemployment, Cambridge: Cambridge University Press.
Malinvaud, E., 1982, "'Wages and Unemployment,' Economic Journal.
Niarris, R., 1964, The Econornic Theorv of Mlanagerial Capitalism. London: Macrnillan.
Marris, R. and A. Wood, 1971, The Corporate Economy, London: Macmillan, 1971.
Meyer, J. and R. Glauber, 1964, Investment Decisions. Economic Forecasting and Public l'olicv,
Boston: Hanrard University Press.
Meyer, J. and E. Kuh, 1957, The Investment l)ecision, C'ambridge: Harvard University Press.
Nlodigliani, F. and NI. Mliller, 195S, "The Cost of Capital, Corporation Finance and the Theory
of Investment," American Econornic Review.
Nickell, S., 1978, The Investment Decision of Firms, Welwyn: J. Nisbet and Company and
Cambridge University Press.
Nickell, S., 1985, "Error Correction, Partial Adjustmnent and All That: An Expository Note,"
Oxford Bulletin of Econorics-and Statistics.
Nowshirvani, V., 1971, "A Modified Adaptive Expectations Model," American Journal of
Agricuitural Economics.



- 65 -
Oks, D, 1987, "Public Sector Deficits in Argentina: Measurement and Macroeconomic Effects,"
D.Phii. thesis, St. Antony's College, Oxford University.
Pick's, World Currency Yearbook, New York: International Currency Analysis, various issues.
Pinell-Siles, A., 1979, "Determinants of Private Industrial Investment in India," World Bank
Staff Working Paper number 333.
Precious, NM., 1987, Rational  xpectations. Non-Market Clearing and Investment Theory,
Oxford: Clarendon Press
Radwan, S., 1974, Capital Formation in Eg2vptian lndustr and Agriculture. 1882-1967, London:
Ithacn Press.
RoAley, J. and P. Trivedi, 1975, Econometrics of Investment, New York: John Wiley and Sons.
Salmon, M., 1982, "Error Correction Mechanisms," Econormic Journal.
Samuelson, P., 1939, "Interaction between the Multiplier Analysis and   * Principle of
Acceleration," Review of Economics and Statistics.
Sargari, J. and A. Bhargava, 1983, "Testing Residuals from Least Squares Regression being
C :nerated by the Gaussian Random Walk," Econometrica.
Serven, L. and A. Solirnano, 1989, "Private Investment aid Macroeconomic Adjustment: An
aOerliew," PPR Working Paper 339, World Bank.
Siiatik, N., 1989, 'Private Investment and Public Policy in Egypt, 1960-1986," DPhI-'I thesis,
Oxford University.
C ness Xns. H., 19S7, 'Investment and the Inflation-Unemployment Trade-Off in a '-a 'roeconornic
R3aioning Model with Monopolistic Competition," European Economic RexN-x.
Solinuio, A., 'IA9, 'Hlow Private Investment Reacts to Changing Macroecononmc  
The Case o)f Chile in the 1980s,' PPR Working Paper Number  1, lTe WNor2J  Bank.
Sundarar.ijan, V. and S. Thakur, 1980, "Public Investment, Crowding Out. anu4 CGrocth: A
D.na mriC Model Applied to India and Korea," INIF Staff Papers.
ra.  ..; . . icroi policy :n the Tropics: How Sensible Peopie Stand," World Vve 2r.cnn.
I:r::; j.  6, (. mrnent on Crockett-Friend and Jorgenson," in R. Ferber, editor,
At .tn1,..."  ,^.In%estm   ehavior, New York: National Bureau oI lconornuc Researcil.
Tobin.  , i69, '. General Equilibrium A-proach to Monetarv Theorv.' 1'nc    -niev 
LrI- tq MJn kT.g.



* 66 -
Tobin, J. and W. Brainard, 1977, "Asset Markets and the Cost of Capital," in B. Balassa and
R. Nelson, editors, Economic Progess, Private Values and Public Policy: Essays in Honor
of Wlliam Eeliner, Amsterdam: North Holland.
Tun Wai, U. and C. Wong, 1982, "Determinants of Private Investment in Developing Countries,"
Joural of Develogment Studies.
Waterson, M., 1984, Econoraic Theory of the Industry, Cambridge: Cambridge University Press.
Witte, J., 1963, "The Microfoundations of the Social Investment Function," Journal of Political
EconoQLy.



PRE Working Paper Series
Contact
Title                           Author Db                                for paper
WPS429  Ghana's Cocoa Pricing Policy      Merrill J. Bateman      June 1990        C. Spooner
Alexander Meeraus                        30464
David M. Newbery
William Asenso Okyere
Gerald T. O'Mara
WPS430  Rural-Urban Growth Linkages in    Peter B. Hazell         May 1990         C. Spooner
India                           Steven Haggblade                         30464
WPS431  Recent Developments in Marketing  Panos Varangis          May 1990         D. Gustafson
and Pricing Systems for Agricultural Takamasa Akiyama                     33714
Export Commodities in Sub-Saharan Ehon Thigpen
Africa
WPS432  Policy Choices in the Newly       Bela Balassa            May 1990         N. Campbell
Industrializing Ccuntries                                                33769
WPS433  India: Protection Structure and   Francois Ettori
Competitiveness of IndL .try
WPS434  Tax Sensitivity of Foreign Direct    Anwar Shah           June 1990        A. Bhalla
Investment: An Empirical        Joel Slemrod                             37699
Assessment
WPS435  Rational Expectations and         Boum-Jong Choe          June 1990        S. Lipscomb
Commodity Price Forecasts                                                33718
WPS436  Commodity Price Forecasts and     Boum-Jong Choe          June 1990        S. Lipsc:omb
Futures Prices                                                            33718
WPS437  Institutional Development Work in   Cheryl W. Gray        June 1990        L. Lockyer
the Bank: A Review of 84 Bank    Lynn S. Khadiagala                       36969
Projects                        Richard J. Moore
WPS438  How Redistribution Hurts          Milan Vodopivec         June 990         J. Lutz
Productivity in a Socialist Economy                                      36970
(Yugoslavia)
WPS439  Indicative Planning in Developing    Bola Balassa         May 1990         N. Campbell
Countries                                                                 33769
WPS440  Financial Sector Policy in Thailand:  William Easterly    June 1990        R. Luz
A Macroeconomic Perspective      Patrick Honohan                          34303
WPS441  Inefficient Private Renegotiation  Kenneth Kletzer
of Sovereign Debt
WPS442  Indian Womer., Health, and        Meera Chatterjee
Productivity



PRE Working Pager Series
Contact
Ttle                             Author                    Date 
WPS443  The Inflation-Stabilization Cycles    Miguel A. Kiguel
in Argentina and Brazil          Nissan Liviatan
WPS444  The Political Economy of Inflation    Stephan Haggard      June 1990        A. Oropesa
and Stabilization in Middle-Income   Robert Kaufman                        39176
Countries
WPS445  Pricing, Cost Recovery, and        Rachel E. Kranton       June 1990        W. Wright
Production Efficiency in Transport:                                        33744
A Critique
WPS446  A Model of Crop and Livestock     Gerald T. O'Mara
Markets in Mexico                Merlinda Ingco
WPS447  Analyzing the Effects of U.S.     Gerald T. O'Mara
Agricultural Policy on Mexican
Agricultural Markets using the
MEXAGMKTS Model
WPS448  A Model of U.S. Corn, Sorghum,    Richard E. Just
and Soybean Markets and the
Role of Government Programs
(USAGMKTS)
WPS449  Analysis of the Effects of U.S.    Richard E. Just
Macroeconomic Policy or. U S.
Agriculturre Using the USAGMKTS
Model
WPS450  Portfolio Effects of Debt-Equity   Daniel Oks
Swaps and Debt Exchanges
with Some Applications to
Latin America
WPS451  Productivity, Impe,rect Competiton  Ann E. Harrison
and Trade Liberalization in
the C6te d'lvoire
WPS452  Modeling Investment Behavior in    Nemat Shafik            June 1990        J Israel
Developing Countries: An                                                  31285
Application to Egypt
WPS453  Do Steel Prices Move Together"    Ying Oian
A Co-integration Test
WPS454  Asset and Liability Management    Toshiya Masuoka          June 1990        S. Bertelsmeier
in the Developing Countries  Modern                                       33767
Financial Techniques -- A Primer