213
pages

Voir plus
Voir moins

Vous aimerez aussi

Sciences est la nouvelle revue de la Faculté

de sciences économiques et de gestion de of Economics and Social Sciences

l'Université catholique de Bukavu (Province

du Sud-Kivu, RDCongo). L’objectif de la revue

est de collecter des données, d’organiser des 3 Revue de la faculté de sciences économiques et de gestion

connaissances sur différents aspects de la vie

UNIVERSITÉ CATHOLIQUE DE BUKAVUsocioéconomique de la région du Kivu.

AU SOMMAIRE ADOPTION

DE CE TROISI ME NUMÉRO

DES TECHNOLOGIES

technological progress and investment: déterminants des inégalités de revenu ET BIEN-ÊTREa non-technical survey à Burhale au sud-Kivu

Raouf Boucekkine et Bruno de Oliveira Cruz Godelive Batano Kusimwa,

Célestin Bucekuderhwa Bashige

et Jean-Baptiste Ntagoma Kushinganine

technology adoption in south Kivu

province subsistence farming

of democratic republic of the congo capital humain et croissance

Bucekuderhwa Bashige Célestin du bien-être à Kagera, en tanzanie

Christian Kamala Kaghoma

relation entre la satisfaction des

employés et leur intention de départ smuggling at the congolese-rwandan

des institutions de microinance au border: drivers, context and welfare

Kivu en rdc : effets modérateurs des impact

variables sociodémographiques Nene Morisho Mwana Biningo et Martin

Eddy Balemba Kanyurhi Doevenspeck

et Deogratias Bugandwa Munguakonkwa

22,50 €

ISBN : 978-2-343-11226-8

En collaboration avec la collection "Comptes rendus"

Fondée et dirigée par Eddie Tambwe

è

Maquette de couverture : Serge Lauret / Illustration : © mitay20 - Fotolia.com

Bukavu Journal

Adoption des technologies et Bien-être

of Economics and Social SciencesBukavu Journal

of Economics

and Social Sciences

Revue de la faculté de sciences économiques et de gestion

1

Revue d’économie et de sciences socialesRevue de la faculté

de sciences économiques

et de gestion

Publiée par l’Université catholique de Bukavu

Revue semestrielle

Comité de rédaction

Rédacteur en chef

Célestin BUCEKUDERHWA Bashige

Assistants de rédaction

Guillaume BIDUBULA Juwa

Christian KAMALA Kaghoma

éditeur Responsable

Université Catholique de Bukavu

recteurucb@ucbukavu.ac.cd, vraucb@ucbukavu.ac.cd

Comité Scientifque

Deogratias BUGAn DWA Munguakonkwa, Marc LABIE, Fréderic

KALALA Tshimpaka, Augustin MUTABAZI ngaboy’eka, Paul

KADUn DU Karhamikire, Paul GERADIn , Dieudonné MUHIn

DUKAdi-Kuruba, Paul-Robain nAMEGABE Rugarabura, Rita SUKADI,

Raouf BOUCEKKInE, Stephan MARYSSE, Jean-Baptiste nT AGOMA

Kushinganine, Eddy BALEMBA Kanyurhi, Ketty-Albert LUKUITSHI

Malaika, nene MORISHO Mwana Biningo, Paul Dontsop n ., Adamon

n DUn GU MUKASA, Janvier KILOSHO Buraye, Douglas AMULI Ibale,

John QUATTROCHI, Alice MUFnUGIZI n abintu.

Conception graphique : Serge Lauret.

© L’Harmattan, 2017

5-7 rue de l’école-Polytechnique

75005 Paris

ISBN : 978-2-343-11226-8

EAN : 97823431122682

Bukavu Journal of Economics and Social Sciences (BJESS) numéro 3, 2016UNIv ERSIt é cAt hol IqUE dE BUkAv U

Adoption

des technologies

et bien-être

En collaboration avec la collection « comptes rendus »

l’hARmAtt AN

3

Revue d’économie et de sciences socialesPrésentation

des auteurs

Raouf Boucekkine, PhD – est professeur des universités (classe exceptionnelle)

à l’Université d’Aix-Marseille, et membre senior de l’Institut Universitaire de

France. Il est actuellement directeur général de l’Institut d’Etudes Avancées

d’Aix-Marseille. Ses thèmes de recherche sont principalement la théorie de la

croissance et du développement et la macroéconomie dynamique. Il est

éditeur associé de nombreuses revues internationales de premier plan comme

Journal of Mathematical Economics, Journal of Economic Dynamics and

Control ou Macroeconomic Dynamics.

Bruno de Oliveira Cruz, PhD – holds a B. A. and a Master in economics from

the University of Brasilia, and a Master of Arts and PhD from Université

Catholique de Louvain. He has worked on. He has been researcher at IPEA

since 1996, working in themes on investment and technological adoption,

innovation and growth, regional and urban economics. In 2007, he was

awarded the best paper in the 4th Urban Research Seminar at the World Bank,

in the same year; he was nominated Deputy-director of Regional and Urban

Studies at IPEA. n ow, he is a Director for Economic and Social Research at

CODEPLAn (Brasília, Federal District, Brazil).

Célestin Bucekuderhwa Bashige, PhD – est professeur associé à la faculté des

Sciences Economiques et de Gestion et chercheur au Laboratoire d’Economie

4 Appliquée au Développement (LEAD) de l’Université Catholique de Bukavu.

Bukavu Journal of Economics and Social Sciences (BJESS) numéro 3, 2016Ses thèmes de recherche embrassent l’économie rurale, la sécurité

alimentaire, l’entrepreneuriat, la microéconomie du développement, l’accès aux

services publics, l’intégration régionale et le commerce transfrontalier. Il est

consultant dans plusieurs institutions et organismes internationaux comme la

Banque Mondiale, le Pn UD, l’EDF-RD, etc.

Eddy Balemba Kanyurhi, PhD – est professeur associé à la faculté des Sciences

Economiques et de gestion de l’Université Catholique de Bukavu (UCB) et

chercheur au Laboratoire d’Economie Appliquée au Développement (LEAD).

Il détient un doctorat en Sciences Economiques et Gestion de la Warocqué

School of Business and Economics de l’Université de Mons (Belgique). Ses

recherches portent sur la microfnance, le marketing bancaire et

l’entrepreneuriat. Il a récemment publié des articles dans les revues International Journal

of Bank Marketing, African Journal of Marketing Management et Bukavu

Journal of Economics and Social Sciences. Il mène aussi des consultances pour

le compte du Programme des n ations Unies pour le Développement (Pn UD),

Adam Smith International, EDF-RD, etc.

Deogratias Bugandwa Mungu Akonkwa, PhD – Docteur en Sciences

Economiques et de Gestion (Université Libre de Bruxelles) et détenteur

d’un Diplôme d’Etudes Approfondies en Gestion (Université Catholique de

5Louvain), Deogratias Bugandwa Mungu Akonkwa est professeur des cours

Revue d’économie et de sciences socialesde Méthodologies quantitatives en Economie et en Sciences Sociales à

l’Université Catholique de Bukavu, à l’Université de Goma et à l’Institut Supérieur

d’Informatique et de Gestion (Goma). Actuellement, il assure en même temps

les fonctions de Directeur Général Adjoint chargé des Opérations & Finances

de la Centrale des Mutuelles d’Epargne et de Crédit du Congo (MECRECO/

COOCEC). Ses principaux domaines de recherche sont la Gouvernance des

Institutions d’enseignement supérieur, la Gestion de la Qualité, le Marketing

Stratégique et la recherche marketing.

Godelive Batano Kusimwa – est assistante à la faculté des Sciences

économiques et de gestion de l’Université Catholique de Bukavu (UCB). Elle détient

une licence en Economie rurale. Elle participe aux études du Laboratoire

d’Economie Appliquée au Développement (LEAD) et du Centre d’Expertise

et de Gestion Minière (CEGEMI). Ses recherches portent sur la pauvreté et les

inégalités.

Jean-Baptiste Ntagoma Kushinganine, PhD – est professeur à la Faculté des

Sciences Economiques et de Gestion de l’Université Catholique de Bukavu

(UCB) et chercheur au Laboratoire d’Economie Appliquée au Développement

(LEAD). Il est titulaire d’un doctorat en Sciences économiques de

l’Université Catholique de Louvain. Ses travaux de recherche embrassent les thèmes

suivants : l’intégration régionale, la décentralisation, les fnances publiques,

la macroéconomie de l’environnement, la politique économique. II a conduit

plusieurs projets de recherche dans divers domaines, notamment en

économie de l’environnement, économie de développement et sur le commerce

transfrontalier. Il a été vice-recteur aux Afaires académiques de l’Université

Catholique de Bukavu et consultant dans plusieurs institutions des n

ationsUnies, notamment la Banque mondiale et la Commission Economique pour

l’Afrique. Il est actuellement membre du Comité n ational de la Recherche

Scientifque, du Comité national LMD en RDC et Conseiller Principal en 6

Bukavu Journal of Economics and Social Sciences (BJESS) numéro 3, 2016charge des Stratégies et Prospectives Economiques au Cabinet du Premier

Ministre de la RDC.

Christian Kamala Kaghoma, PhD – Kamala Kaghoma has a Ph.D in Economics

from Université de Yaoundé II-Soa with majors in Econometrics and

Environmental economics. He is currently a Professeur Associé and the Dean

of the faculty of Economics and Management of the Université Catholique de

Bukavu (UCB) where he has been teaching for more than a decade. Kamala’s

main areas of research include Public Economics, Environmental

Economics, Human resources Economics, Applied Macroeconomics,

Welfare Economics, Applied econometrics, Development microeconomics

and research methodology. His work focuses on child welfare related to w-a

ter and sanitation access, poverty dynamics, social mobility and

intergenerational (multidimensional) welfare transmission with emphasis on African

countries. He is currently working on youth employment, entrepreneurship

and internal migration in the Democra/c Republic of the Congo (DRC), his

country. Mr. Kamala Kaghoma has conducted several feld researches and

surveys in DRC and has a broad consultancy experience in Africa.

Nene Morisho Mwana Biningo, PhD – holds a PhD in Institutional Economics

from the University of Bayreuth in Germany. He also holds two masters

degrees in Macroeconomics and Management of Development Projects

(Development Studies). He is an Associate Professor at the Catholic University

of Bukavu (UCB) where he teaches Macroeconomics and Public Finances. His

felds of interest are Cross border trade, Regional Economic Integration and

Institutions and growth.

Martin Doevenspeck, PhD – Professor of Geographical Confict Research at the

University of Bayreuth, Germany. His research focuses on violent confict and

the political geography of climate change and risk in West and Central Africa. 7

Revue d’économie et de sciences socialestable

des matières

Technological progress and investment: a non-technical survey ......... 10

Raouf Boucekkine et Bruno de Oliveira Cruz

Technology adoption in South Kivu province subsistence

farming of Democratic Republic of the Cong .........................................o 32

Bucekuderhwa Bashige Célestin

Relation entre la satisfaction des employés et leur intention

de départ des institutions de microfnance au Kivu en RDC :

Efets modérateurs des variables sociodémographiques .........................77

Eddy Balemba Kanyurhi et Deogratias Bugandwa Munguakonkwa

Déterminants des inégalités de revenu à Burhale au Sud-Kivu ............124

Godelive Batano Kusimwa, Célestin Bucekuderhwa Bashige

et Jean-Baptiste Ntagoma Kushinganine

8

Bukavu Journal of Economics and Social Sciences (BJESS) numéro 3, 2016Capital humain et croissance du bien-être à Kagera,

en Tanzani e.................................................................................................. 162

Christian Kamala Kaghoma

Smuggling at the Congolese-Rwandan border: drivers, context

and welfare impact ......................................................................................190

Nene Morisho Mwana Biningo et Martin Doevenspeck

• • •

9

Revue d’économie et de sciences socialesTechnological progress

and investment

A non-technical survey

Raouf BOUCEKKINE et Bruno de Oliveira CRUZ*

RéSUMé/ABSTRACT

t his paper presents a non-technical overview of the recent investment literature with a special

emphasis on the connection between technological progress and the investment decision.

First of all, we acknowledge that some dramatic advances have been made in the 1990s in

understanding and modelling non-convex capital adjustment schemes and irreversibility.

Nonetheless, this new literature has not always satisfactorily accounted for the

investmentspecifc (or embodied) nature of technical progress. We argue that the recent technological

trends towards more embodiment have had a heavy impact on the way the investment

decision is taken and is to be taken. t his is turn should imply the reconsideration of many

empirical results, and a more careful modelling strategy taking into account the price variables and

scrupulously choosing the most appropriate level of (dis)aggregation.

Keywords: Investment, technological progress, Non-convex adjustment, irreversibility,

Embodiment.

Journal of Economic Literature: E22, E32, o40.

efective demand and employ-1. Introduction

ment. If only we knew more about

“Economic Teory can give rea- the determinants of investment!

sonably good account of how the But unfortunately, our knowledge

in this direction is still very mea-level of investment infuences

ger. One might well ask, what is

* Bruno Cruz acknowledges the fnancial wrong with the theory of

investsupport of CAPES Foudation (Brazil) for this 10 ment? Or perhaps, what is wrong research.

Bukavu Journal of Economics and Social Sciences (BJESS) numéro 3, 2016with the subject matter itself! For quite unsuccessful in empirically

one thing, this variable - the pivot forecasting this component of a-g

of modern macroeconomics- has gregate demand. One can observe

apparently lived a somewhat noma- that heuristic models, such as

accedic life among the various chap - lerator models, have shown to be

ters of economic theory. Perhaps empirically better adjusted than

it has not stayed long enough in models based on

micro-foundaany one place. Perhaps it has been tions. Te apparent superiority of

ill-treated” (Haavelmo, T. (1960) A heuristic models should not lead

Study in the Teory of Investment, us to forget about price variables

pp.3) and Tobin’s “q” when talking about

With this astonishing paragraph, investment. First of all, the superio -

Haavelmo (1960) begins his famous rity of estimated accelerator mo -

book A Study in the Teory est- dels, for example, does not mean

ment Teory in the early 1960s. that decision is made and how this

Many surveys on the subject have decision is oriented by economic

appeared since then (for instance, policy; there is not a theory behind

Jorgenson, 1963; Chirinko, 1993); the accelerator, it’s just a technique

theory has made an incredible step that apparently works in some

cirtowards a more comprehensible cumstances and that has its own

understanding of this important problems (see Oliner, Rudebusch

variable. Yet in 2000, Caballero, and Sichel, 1995, for example).

one of the most important resear- Second, the bad empirical

perforchers in the feld, opened his talk mances of optimal control-based

“Aggregate Investment: Lessons investment models can be also

parfrom the Previous Millennium” tially attributed to the way they are

with the following statement: tested in practice. For example, the

“But while we all may agree on the way the rate of capital depreciation

importance of investment for a n-a is traditionally treated in

econometion’s economic health, our unders- tric applications is highly

questiotanding of its determinants, both at nable, especially during the 1990s,

the microeconomic as well as the as it does not accurately capture the

macroeconomic level, has remained pace at which capital goods actually

limited. Te empirical investment become obsolete.

literature has been nearly merciless It seems therefore overwhelmin -

in evaluating investment theo- gly clear that the traditional user

ries.” (Caballero, R. 2000 American cost and Tobin’s q models are too

Economic Association Session. In much “stylized” to serve as

univerMemoriam: Robert Eisner, pp 1-2). sal and unquestionable models of

Tis pessimistic view about the both microeconomic and aggregate

evolution of our knowledge about investment. Tere is an urgent need

investment might be infuenced by to profoundly study and docu -

11the fact that researchers have been ment how the investment decision

Revue d’économie et de sciences socialesis efectively taken in real life over (summing up the investments done

a wide variety of microeconomic by all the establishments of their

cases. In particular, a much clo- sample) and an index of microeco -

ser inspection into how the capi- nomic investments and showed that

tal adjustment processes actually it is actually very high.

take place is required, especially in b. Recent empirical studies

connection with the pace of tech - confrm that the Information and

nological progress. Tis calls for Communication Technologies (ICT

two major focuses: hereafer) burst in the 1990s has

a. In the traditional optimal control considerably distorted the

investtheory developed (see for instance ment behavior at all levels: Te

assoChirinko (1993)), the capital adjust- ciated dramatic decrease in the

relament issue is settled by setting a tive price of capital during the 1990s

convex adjustment function, usually has rehabilitated price variables as

a quadratic function for tractability. a major determinant of investment

But is adjustment gradual in real decisions (see for instance Tevlin

life? Te answer is defnitely no. It is and Whelan, 2003). Tis has led

now known that investment at the some authors like Whelan (2002)

frm level is lumpy and infrequent to advocate another empirical

ap(Doms and Dunne, 1998), and that praisal of the investment decision,

these two characteristics are unli- based on a two-sector accounting

kely to completely disappear in the benchmark model in order to refect

aggregate (Cooper, Haltiwanger the pace of the relative price of cap -i

and Power, 1999). Doms and Dunne tal and the very fast depreciation of

worked on the investment patterns an increasingly large fraction of the

of 12 000 plants in US manufa-c capital stock. At the same time,

anoturing over the interval 1972-1989. ther complementary view of

investFor each frm, they constructed a ment emerged, resurgence of the

series of the proportion of the total vintage capital theory of the sixties:

equipment investment of the frm. It investment and technological

innoturns out that the largest investment vations are not “separated”,

investperiod accounts on average for more ment is the unique vehicle of inno -

than 25 percent of the 17 year of in- vations decisions, as Greenwood

vestment. Moreover, more than half and Jovanovic (2001) observe:

of the frms exhibit a capital growth ”In reality, advances in

technoof about 50 percent in a single year. logy tend to be embodied in the

laAlso, the second largest investment test vintages of capital. Tis means

spike ofen comes next to the largest that new capital is better than old

investment, which suggest that the capital, not just because machines

two biggest spikes correspond to a sufer wear and tear as they age, but

single investment episode. Finally, also because new capital is better

Doms and Dunne studied the corre- than the old capital was when the

12 lation between aggregate investment latter was new. It also means that

Bukavu Journal of Economics and Social Sciences (BJESS) numéro 3, 2016there can be no technological pro- 2. Characterization

gress without investment” (pp. 179- of microeconomic adjustment

180, Italics ours).

In this paper, we carefully review As we have mentioned in the

the state of art of the literature introduction, there is a compelling

investment regarding these two evidence that investment patterns

fundamental aspects. We start by at least at the plant level are far from

a non-technical discussion of the gradual, which goes at odds with

“new” investment theories which the optimal control investment

have recently introduced non- model with convex adjustment

convex adjustment costs and irre- costs. How could the economic

versibility in the heart of invest- theory deal with this clear

inconment theory (Section II). We frst sistency problem? We shall briefy

notice that these theories have review the recent stream of econo -

considerably improved our unders- mic literature devoted to this fun -

tanding of how the investment damental issue. We start with the

decision is taken both at the plant basic non-convex adjustment costs

and the aggregate levels. However, story, more sophisticated concepts

we also observe that, with some will be treated along the way.

very few exceptions, the modelling

of technological progress in such 2.1. Non-convex capital

adjusttheories is not markedly diferent ment costs

from traditional “exogenous” mo - Tis issue is very

comprehensideling in the neoclassical model; vely treated by Caballero (1999),

the whole action inherent in these among others. Following this

autheories come from non-convex thor, assume that a given frm has

*adjustment and other ingredients an optimal capital stock, K , which

like irreversibility. nonetheless as depends on the interest rate and

outlined just above, there are plenty on any proftability (exogenous)

of microeconomic and macroeco - variable (possibly technological

nomic studies pointing at a signi- innovations). Te crucial thing

fcant change in the composition is the specifcation of the

adjustof technological progress afer the ment costs. If one aims at

genera60s, and showing that embodiment ting infrequent and lumpy

investshould be seriously accounted for, ment patterns at the optimum, he

especially in the computer era. We should care about the functional

therefore end the survey with a de- form to assign to these costs. Te

tailed non-technical exposition on traditional simple convex form is

modelling embodiment both at the inadequate. In order to generate

micro and macro level, and on the infrequent investment, the

adjustmethodological consequences of ment costs function should be cho -

the information technologies boom sen such that it increases sharply

13in the 1990s (Section III). around the point of no adjustment.

Revue d’économie et de sciences socialesK1As mentioned by Caballero, a cost the following. Call K * , a measure K

proportional to the size of the nee- of capital imbalance. Assume that

ded adjustment is enough. now, the actual capital imbalance is near

this specifcation gives infrequen - the point which maximizes the

vacy, but not lumpiness. To get the lue of the frm (roughly speaking the

latter characteristic, there must be discounted sum of the present and

an advantage in bunching invest- (expected) future profts of the frm).

ment, and this can be achieved if Ten, the frm may not have any

for example adjustment requires incentive to pursue the adjustment

a fxed cost. For a positive invest- because of the incurred adjustment

ment I, the associated adjustment costs. Tis is specially true for small

cost is consequently: c +c I where adjustments because of the fxed f v

c and c are two positive constants. costs. Hence, the frm may perfectly f v

A similar adjustment costs func- choose to be inactive in such a case.

tion has to be set in case of disin - Indeed, it is possible to prove

rigovestment (when I < 0), including a rously that there exists a non-empty

positive fxed cost of disinvestment. range of inaction in the space of Z.

We are in a typical situation where More concretely, there exists a target

the adjustment technology exhibits point L such that there is no

investincreasing returns. For the relative ment for Z > L, and a target point U

importance of adjustment costs to such that there is no disinvestment

be constant over time, the latter for Z < U. Which ultimately means

*term is usually multiplied by K . that there exists a range of inaction

How does the optimization mo- (L, U). Infrequency and lumpiness

del perform with this modifed can therefore be generated within

adjustment costs function? And this alternative framework.

how does the obtained demand for 2. Is the q-theory robust to such

capital goods look like, in particu- a "realistic" specifcation of

adjustlar in relation with technical inno- ment costs? Recall that in the basic

vations? Te main properties of the optimization setting, we get the fo-l

optimization problems are stated in lowing relationship between

investCaballero (1999), section 3. Te fo-l ment and (marginal) q : q=1+C’(I),

lowing points can be highlighted: when the unit price of capital is equal

1. In such a model, there is room to 1 (s = 1). In the quadratic case, ie.

2for inaction. Tis a crucial departure when C(I)=bI , where b > 0 , we get

from the standard models with qua- a linear relationship between q and

dratic adjustment costs. In such mo- I. In the general convex case, we get

dels, capital accumulation and in- an implicit monotonic functional

vestment paths are smooth in time,

so there is no infrequent or lumpy 1. Te reader interested in the exact

mathematical solution can _nd all the technicalities in investment episode. In the

alternaCaballero (1999). Te technique makes use of

tive model, things are very diferent.

dynamic programming in continuous time 14 Te intuitive reasoning behind is with the associated optimality principle.

Bukavu Journal of Economics and Social Sciences (BJESS) numéro 3, 2016relation between q and I : I = φ(q). simply due to the use of linear

reTo each value of q corresponds a gressions, as the “true” relationship

single investment amount. Is this between q and capital is probably

functional relationship preserved highly nonlinear.

in the case of adjustment costs with 3. How do technological

innoincreasing returns? Te answer is vations enter this new investment

no as demonstrated by Caballero set-up? Actually, these new

microand Leahy (1996). In efect, the rela- economically founded adjustment

tionship between marginal q and Z, theories enrich in a considerable

the capital imbalance, is not func- way the discussion on the efects

tional over the Z space in that the of technological innovations (or

same value of q is associated with any other external shock) on the

diferent investment values. And in demand for equipment. In the stan -

the region of the Z space where this dard optimal control theory, there

relationship is functional, it is highly is no range for inaction. In the case

nonlinear. So in the very best case, of a technological improvement

the q-theory only holds "locally" but (raising the productivity of

capithe induced functional relationship tal goods for example), q increases

between q and I is by no way linear. since it is a measure of the value

Tis suggests two main points. to the frm of an additional unit

First of all, the q-theory is typically of (new) equipment. As q goes up,

a non-robust theory. Indeed, there investment is systematically

stimuexists a huge literature on this par- lated, because q is monotonically

ticular point. Many authors have related to I via the investment

equastudied the implications of diferent tion, q = 1 + C’(I), under the strict

specifcations of the adjustment convexity of the adjustment costs

costs departing from the initial function. When a range of inaction

quadratic formulation. Some have (optimally) arises, a technical

innoremoved the fxed costs; others vation does not necessarily trigger

1have added a strictly convex term to an investment boom .

the adjustment functions. In some Let us have a closer look at this

contributions, the monotonic rela- specifc point. A careful reading

tionship between q and investment of Caballero and Engel (1999)

is recovered but at the cost of less allows to notice that

technologirealistic generated investment pat- cal progress, purely disembodied

terns, notably in terms of lumpiness in this theory, exclusively

ope(as in the very well-known Abel rates through the optimal capital

*and Eberly’s 1994 paper). Because of stock, K , thus through the

imbathese robustness problems, the fai- lance ratio Z. Indeed, neither the

lures registered in its econometric threshold, L nor the threshold U

implementation are not surprising

at all. Second, a more favorable to

1. Tis may be true even in the absence of 15the q-theory, these failures may be fxed costs. See Abel and Eberly (1994).

Revue d’économie et de sciences socialesdepends on technological progress: 2. 2. Irreversibility and

In Caballero and Engel (1999), both uncertainty

are function of the fraction of p-ro Te literature relating

investfts foregone due to capital stock ment and uncertainty has

geneadjustment, which is ultimately rated two diferent conclusions.

assumed to be randomly distribu- In the one hand, the presence of

1ted . Tus, a technological impro- constant returns to scale and

symvement has essentially the virtue metric adjustment costs have led

of raising the optimal capital stock, to the conclusion that uncertainty

*K , which for given K, lowers the increases the value of investment

Kcapital imbalance Z since Z = . (see for instance Hartman, 1972, *K

Suppose that initially Z is in the and Abel, 1983 and 1985)). In this

range of inaction (L,U). A small set-up, the marginal value of capital

technological shock is unlikely to is a convex function of the stocha-s

lower Z at a value below the thres- tic process: Jensen’s inequality thus

hold L. For investment to occur, ie. implies a higher demand for

investfor Z to be shifed below L, the ma- ment. On the order hand, the

intrognitude of the technological impro - duction of irreversibility gives rise to

vement should be big enough. Tis another diferent mechanism (Dixit

is one the reasons why technologi- and Pindyck, 1994, and Pindyck,

cal difusion is not instantaneous. 1988). Irreversibility of investment

Investment involves some non- amounts to saying that undertaking

negligible adjustment costs which investment projects results in some

makes it optimal sometimes to not unrecoverable initial costs, the

soact, to not invest. In such a case, the called sunk costs. Uncertainty on

institutions have a role to play, and future benefts and costs of inves-t

economic policy crucially matters. ment projects makes the resulting

For example, when the technologi- investment problem trickier. If we

cal improvement is not enough big assume with Dixit and Pindyck

to encourage investment, a further that if the investment project is not

decrease in the interest rate (or any undertaken today, the frms retain

other component of the user cost of the option of undertaking the pro -

capital) may help the difusion (by ject tomorrow, there is a clear value

☐increasing K and then by lowering of waiting: the frms have always the

even more the capital imbalance Z). possibility to postpone investment

Uncertainty and irreversibility tomorrow in order to learn more

are other factors that call for ina-c about present and future project

tion and delayed adoption of inno - payofs. Tis value of waiting, this

vative tools, as it is explained in the option of waiting, is the main cha -

next sub-section. racteristic of the above mentioned

theories, and it has some crucial

implications in terms of investment

1. See page 8 of the article by Caballero and 16 patterns (lumpiness, infrequency, as Engel (1999).

Bukavu Journal of Economics and Social Sciences (BJESS) numéro 3, 2016observed in micro data) and econo- to the innovation by investing, that

mic policy. We shall summarize the is the threshold value of expected

related stream of literature in the MPK above which the frm invests,

few following points. is a priori signifcantly bigger than

1. As argued in Chirinko (1996), the typical user cost or discounting

\...In the traditional optimal control rate threshold values encountered

model àla Jorgenson, investment is in the traditional theory. To these

reversible because well-functioning typical terms, one has to add the

secondary markets exist or the rate value of waiting, which is in general

at which frms wish to reduce the a function of a measure of unce-r

capital stock is less than the rate tainty, typically the variance of a

of depreciation, ie. gross invest- stochastic price or technological

ment is always positive. Moreover, variables, and of the actual levels

in the traditional approach, there of the latter variables, including the

1is no room for fexible timing in capital depreciation rate .

investment: Either the frms invest It follows that investment

pat(I>0) or they don’t, and in the lat- terns are potentially less smooth

ter case, it is implicitly assumed when one has to account for

irrethat the investment opportunity versibility under uncertainty. As

lost will never be recovered. Te in the models with non-convex

optimal investment rule giving adjustment costs seen above, there

the optimal capital stock is basi- is room for inaction: provided the

cally: Investment is undertaken if fexibility of the investment

deciand only if the (expected) margi- sion timing, the frms can decide

nal proftability of capital, MPK, is to not act for a while since there is

bigger than a user cost of capital, a value of waiting, which generates

which includes the interest rate (the the infrequency characteristic.

discounting rate), the depreciation What are the theoretical

implicarate of capital, and the (expected) tions of this approach as to

investrate of change of the acquisition ment patterns? As we have just said,

price of capital. Tis is exactly irreversibility plus uncertainty and

what may happen if the frm, ini- time fexibility implies that

equiptially at equilibrium, is afected by ment purchases occur only in

a positive technical innovation; the spurts. To be precise, investment

curve is likely to be shifed upward, occurs if the (expected) proftability

so that investment takes place. In of marginal investment is enough

Dixit-Pyndick’s models, the rule high to compensate the cost of

capiis not that simple. Tere is an op- tal and the value of waiting. Tis is

tion of waiting. Even if a technical likely to happen in very good times

innovation unambiguously shifs (for example, when the demand for

upward MPK, uncertainty,

irrever1. Te reader interested in the exact formula

sibility plus time fexibility may

inof the value of waiting can fnd the necessary 17duce the frm to wait, and not react material in Pindyck (1988).

Revue d’économie et de sciences socialesthe goods produced by the frms (demand and/or technological)

is very high using historical stan- shock should be of a very large

dards, as in Pindyck, 1988). But this magnitude. For moderate

negais unlikely to happen for any good tive shocks, full utilization (and

draw in the distribution of the zero investment) is optimal. Te

stochastic environment. Typically rate of capacity utilization plays

the frms increase their productive therefore a central role in shaping

capacity only periodically. Tese investment patterns when

irrevermodels usually have marginal in- sibility and uncertainty matters,

crement, and are unable to generate and this should be of interest for

lumpiness. practitioners seeking for the best

Tere is a much more important investment decision in such a

new direction in the irreversibility context. Typical wisdom from this

literature with respect to the tra- approach is that frms must reuse

ditional non-convex adjustment all the units, that’s increase their

costs models: Te role of the rate rate of capacity utilization if po-s

of utilization of capital. Allow the sible, before investing. Certainly,

frms to choose this rate in pre- this behavior is not always

corrosence of uncertainty, irreversibi- borated by the empirical evidence.

lity and time fexibility, and forget In certain cases, frms do invest

about capital depreciation. Te even if they have not reached full

typical outcome is the following. capacity, which goes at odds with

In good times, there is little doubt one of the main implications of the

that the optimal decision should 1988 Pindyck’s seminal work. Tis

be to fully utilize the productive kind of behavior is highlighted in

capacity. In very good times, as several micro-econometric

stuwe have just mentioned above, the dies (see for example Licandro et

frms additionally increase their al. (2005) on Spanish data among

capital stock. What happens in many other frm level studies).

bad times? Tis crucial question Cruz and Pommeret (2010) show

is addressed in Dixit and Pindyck that accounting for

investment(1994), and before by Pindyck specifc technological progress

(1988) in a basic formal setting. (via vintage capital modelling) is

In bad times, since investment enough to explain this striking

is irreversible and capital depre- fnding. Indeed, if investment is

ciation is assumed to be zero, the the exclusive vehicle of

technolocapital stock held does not move. gical innovations and if a decisive

Tis inertia in the capital stock technological upgrading is taking

is coupled with a non-constant place, the decision to invest (ofen

optimal utilization rate. An ad- coupled with replacement of old

verse shock need not reduce this and less productive equipment) is

rate systematically. For the rate likely to be less intimately related

18 of utilization to fall, the adverse to the rate of utilization control.

Bukavu Journal of Economics and Social Sciences (BJESS) numéro 3, 2016We shall insist on the necessity to the empirical failures of this

theoaccount for this feature of techn-o ry. Tat is the usual tests do not

logical progress (that’s embodi- consider the possible existence of a

ment) in Section III, especially in multiple regime (q,I) relationship.

connection with the ICT burst. Teir failure may come from this

Te q-theory is generally non- omission. While this claim is so

robust to extensions of the basic far unproven, as correctly pointed

model incorporating non-convex out by Chirinko (1996), it gives an

adjustment costs. When the early idea about the complexity

benchmark model is extended to of the empirical debate surroun -

include uncertainty, irreversibi- ding the concepts of

irreversibility and fexibility in timing, then lity, uncertainty and non-convex

this non-robustness is much less adjustment costs. Te next

subclear, as it is brilliantly explai- section addresses very briefy this

ned by Dixit and Pindyck in their issue, among others.

1book . True, there is no monotonic

relationship between q and invest- 2.3. Some remarks on the “new”

ment over the whole space of capi- investment theories

tal imbalance Z. But, in contrast Some crucial issues remain to be

to the model with non-convex ad- addressed. First of all, if it is widely

justment costs where non-mono- admitted that non-convex

adjusttonicity may show up locally in the ment costs and/or irreversibility

-space, the picture is simpler here: are much more consistent with

Te Z space is divided in three re- observed investment behavior at

gions where investment is respec- the plant level that the Jorgenson

tively rising, falling or constant neoclassical model, it is crucial to

with increases in q. In other terms, study whether aggregation will

there exist three distinct regimes not “kill” them. In other terms, it

but none shows up a non-mo- is important to assert neatly that

notonic relationship between q non-convex adjustment costs and/

and investment. Tis leads many or irreversibility have a frst-order

theorists, following Dixit and impact in explaining aggregate

Pindyck, to argue that accounting investment patterns. Second, it is

for irreversibility and uncertainty important to examine if these new

does not only preserve the basic q- theories are not leaving in the dark,

theory but it also allows to remedy some crucial aspects of the

investment decision; we shall argue in

1. The reader can find a complete proof of this respect that embodied

techthese claims in the quoted book. These

pronical progress, and more generally perties are mathematically quite intuitive:

The presence of a fixed cost has a much the vintage composition of the

camore damaging effect on the concavity of pital stock, are important but still

the value functions of the firms, arising

insufciently dealt with in these

in the considered dynamic programming 19new theories.problems.

Revue d’économie et de sciences sociales1. With either non-convex ad- of frms exist in the sample (which

justment costs or irreversibility, is the case in the empirical

studthe investment patterns seem to ies conducted on US micro data),

be consistent with the empirical so that the law of large numbers

characteristics of micro adjust- prevails, one can frst take average

ment as documented by Doms and investment over the establishments

Dunne (1998), namely infrequency of the sample having

approximateand lumpiness. But as mentioned ly the same capital imbalance Z,

by Caballero (1999), the empiri- as an estimate of expected

investcal corroboration of these simple ment (conditional to the level of

adjustment rules is far from easy, imbalance). Ten, using the above

as”...frms respond diferently to mentioned simple expression of

exsimilar imbalances over time and pected investment in terms of the

across frms”. Tis leads Caballero hazard function, one can precisely

and Engel (1999) to construct a estimate the latter function. Te

stochastic version of the model second step simply requires

averawith non-convex adjustment costs, ging across all Z.

where the regions of inaction and 3. Te results obtained by

adjustment are not deterministi- Caballero, Engel and Haltiwanger

cally set, the basic idea being that (1995) (corroborated by further

large imbalances are more likely to empirical studies) seem to confrm

induce investment. A simple way to the importance of non-convexities

randomize the latter model is to as- and, incidentally of

irreversibisume that the fxed cost variable is lity, both at the micro and

aggrestochastic. Ten, the analysis relies gate levels. Te main result is that

on the so-called hazard functions, the hazard functions are clearly

H(Z), which describe the probabili- increasing consistently with the

ty to adjust when capital imbalance «new» investment theories

em(or any other convenient function phasizing the role of

non-conveof it) is equal to Z. xities and irreversibility. In

par2. In their celebrated 1999 pa- ticular, expected investment rises

per, Caballero and Engel provide more that proportionally with

the necessary material to estimate capital imbalance. Te linear

spethese hazard functions and more cifcations, consistent with the

traimportantly, to aggregate them. ditional formulations of the user

For a given imbalance Z, it is dem- cost and Tobin’s q-theories, rather

onstrated that the expected invest- predict constant hazard functions,

ment by any frm can be written as ie. the probability to adjust is the

basically the product of the adjust- same whatever is the level of

capiment to be done and the probability tal imbalance. Te same

conclufor the frm to undertake it (namely sion is made at the aggregate level.

H(Z)). Ten aggregation follows in Moreover, the estimated hazard

20 two steps. Provided a large number functions are shown to be very low

Bukavu Journal of Economics and Social Sciences (BJESS) numéro 3, 2016