Toward the crisis: a Kaleckian-Keynesian interpretation of the instability of growth and capital accumulation in Brazil

Abstract

This article examines theoretically and empirically the instability of Brazilian investment and growth for the past couple of decades, highlighting the evolution that led to the current crisis. A theoretical discussion highlights the importance of Kaleckian and Keynesian approaches in understanding the semi-stagnation of the Brazilian economy since the 1990s. Empirical evidence shows that investment has increased until 2013, but not to the point of getting the economy back on the track of high growth rates and higher investment-GDP ratios. The econometric findings are compatible with the theoretical underpinnings of investment activity based on Keynes and Kalecki and suggest the existence of room for activist policies in Brazil in order to stimulate economic activity.

Full text

Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=cira20
Download by: [187.36.0.208] Date: 06 March 2017, At: 04:22
International Review of Applied Economics
ISSN: 0269-2171 (Print) 1465-3486 (Online) Journal homepage: http://www.tandfonline.com/loi/cira20
Toward the crisis: a Kaleckian-Keynesian
interpretation of the instability of growth and
capital accumulation in Brazil
Eduardo Maldonado Filho, Fernando Ferrari Filho & Marcelo Milan
To cite this article: Eduardo Maldonado Filho, Fernando Ferrari Filho & Marcelo Milan
(2017): Toward the crisis: a Kaleckian-Keynesian interpretation of the instability of growth
and capital accumulation in Brazil, International Review of Applied Economics, DOI:
10.1080/02692171.2017.1297387
To link to this article: http://dx.doi.org/10.1080/02692171.2017.1297387
Published online: 06 Mar 2017.
Submit your article to this journal
View related articles
View Crossmark data

INTERNATIONAL REVIEW OF APPLIED ECONOMICS, 2017
http://dx.doi.org/10.1080/02692171.2017.1297387
Toward the crisis: a Kaleckian-Keynesian interpretation of the
instability of growth and capital accumulation in Brazil
Eduardo Maldonado Filho, Fernando Ferrari Filho and Marcelo Milan
Department of Economics and International Relations, UFRGS, Porto Alegre, Brazil
ABSTRACT
This article examines theoretically and empirically the instability
of Brazilian investment and growth for the past couple of decades,
highlighting the evolution that led to the current crisis. A theoretical
discussion highlights the importance of Kaleckian and Keynesian
approaches in understanding the semi-stagnation of the Brazilian
economy since the 1990s. Empirical evidence shows that investment
has increased until 2013, but not to the point of getting the economy
back on the track of high growth rates and higher investment-GDP
ratios. The econometric ndings are compatible with the theoretical
underpinnings of investment activity based on Keynes and Kalecki
and suggest the existence of room for activist policies in Brazil in order
to stimulate economic activity.
1. Introduction
Aer the introduction of the Real Plan (RP) in 1994, when, aer a chronic inationary pro-
cess that took place in Brazil in the 1980s and early 1990s, monetary stability was achieved,
1
the average growth rate of real GDP, between 1995 and 2014, remained around 3.0% a year,
and its performance was characterized by a stop-and-go process. It should be noted that
GDP performance has been even weaker and more volatile since the subprime mortgage
crisis and the great recession that followed, despite the counter-cyclical macroeconomic
policies implemented by the economic authorities to avoid a contagion of the Brazilian
economy: between 2009 and 2014, the average real GDP growth was around 2.6% a year.
is leads us to the following question: Why, in the period 1994–2013, when average
annual ination had been around 7.2%, the performance of the Brazilian economy had been
weak, especially when compared to other emerging countries, and fragile, turning into a
full collapse in 2015–2016? is question becomes even more relevant if we analyze the
growth rate of Brazilian economy from a long-term perspective: between 1950 and 2014,
real GDP grew at an average rate of 4.9% per year. In the 1950s, 1960s, and 1970s, when
the economic growth process was driven by the State and by Keynesian macroeconomic
policies, real annual growth rates were higher than the historical average. In the 1980s and,
especially, 1990s, when neoliberal reforms started to dictate the rules for economic policies,
KEYWORDS
Investment; Brazilian
economy; Kalecki; Keynes;
VEC Models
JEL CLASSIFICATION
E12; E22; O54
ARTICLE HISTORY
Received 8 May 2016
Accepted16 February 2017
© 2017 Informa UK Limited, trading as Taylor & Francis Group
CONTACT Fernando Ferrari Filho ferrari@ufrgs.br

2 E. MALDONADO FILHO ET AL.
and also in the 2000s, when economic policy was slightly changed, annual average real
growth rates were far below the average.2
In our view, the low and unstable GDP growth rates in the last two decades may be partly
explained by the poor performance of gross xed capital formation: between 1990 and 2014,
the investment/GDP ratio hovered around 17.4%,3 within a range between 15.3% and 19.5%.
With this overall view in mind, the goal of this article is to show that the instability of invest-
ment, both public and private, has been a restrictive factor preventing a robust and consistent
economic growth in Brazil. In order to do this, departing from Kaleckian and Keynesian
approaches to investment, we analyze theoretically and empirically the relationship between
capital accumulation and GDP and, aer showing the key variables determining investment
for the Brazilian economy, present some conclusions and reections.
Besides this brief introduction, this article has three sections: the next section summarizes
the investment theories of Kalecki and Keynes to identify the main important variables
that aect investment and, as a result, economic growth in a capitalist economy, in order
to inform the statistical exercises. e third section analyzes the question empirically, using
statistics and a time series econometric model. It investigates the determinants of invest-
ment in Brazil for the 1994–2013 period. Finally, the last section presents the conclusions.
2. Kaleckian and Keynesian theories of investment
is section aims to present, in the light of Kaleckian and Keynesian theories, how invest-
ment is aected by other variables, that is, what are the determinants of investment,
4
so that
we can sort out the variables for the empirical analysis in the next section.
2.1. Investment cycles and trends in the Kaleckian view
Kalecki (1969) developed an investment theory in two steps. In the rst, he presented
the investment theory for a static economy, assuming that it does not have a tendency for
long-term growth. In Kalecki’s conception, this is a short-term analysis.5 e investment
theory has the goal of simply to contribute to an explanation of cyclical uctuations in
output, that is, how instability is pervasive in capitalism. In the second step, he introduces
‘development factors’, thus including the tendency for economic growth. us, his invest-
ment theory seeks to explain cyclical uctuations along a trajectory of economic growth.
is subsection explores only the original Kalecki’s investment theory, but there is a large
literature that examines the theoretical and empirical considerations of his theory, usually
providing support for its internal consistency.6
Initially, it is important to remember that, for Kalecki, the amount of entrepreneur-
ial capital (i.e. equity capital) is the main factor determining the size of a rm. Kalecki
acknowledges that rms also use capital provided by outside investors, but he shows that
the borrowing power of a rm is limited by the amount of entrepreneurial capital. erefore,
the expansion of the rm depends, ultimately, on the internal accumulation of capital; that
is, the accumulation nanced from the rm’s gross prots.7 Another important aspect to
be pointed out regarding the size of entrepreneurial capital, or why rms prefer retained
prots, is related to the ‘increasing risk’ associated with investment expansion.
According to Kalecki (1969, 92), for a given amount of capital, the risk increases with
additional investment, since it boosts the gearing ratios. us, the size of entrepreneurial

INTERNATIONAL REVIEW OF APPLIED ECONOMICS 3
capital also imposes a limit to the investments through the principle of ‘increasing risk’
(Kalecki 1937). According to this principle that shows that greater indebtedness increases
the possibility of the rm’s bankruptcy, Kalecki assumed that investment is a function of
the rm’s retained prots. Moreover, the principle of ‘increasing risk’ means that when the
uncertainty is high, the investment is low.
For Kalecki, it is important to make a clear distinction between investments in xed
capital and in inventories, since the factors that determine them are dierent. Moreover,
considering the investment in xed capital, Kalecki emphasizes that there is an essential
distinction between the decision to invest and the investment itself. Given that and other
considerations, according to Kalecki (1969), the nal formula for total investment of the
economy is the following:
where It+θ is the total investment in period t+θ, St is the amount of gross current savings,
∆Pt/∆t is the rate of change in aggregate prots, ∆Ot/∆t is the growth rate of output, d′t is
a constant, even though it is subject to long-term variation, especially as a consequence of
technological progress, and a, c, and e represent the coecients of St, ∆Pt/∆t, and ∆Ot/∆t,
respectively.8
In other words, the result obtained by Kalecki is that the total investment depends both
on the level of economic activity and on the rate of variation in the level of economic activity,
and it is clear that the incorporation of the determinants of investment in inventories does
not fundamentally change the determinants of total investment.
Moreover, according to Kalecki, the explanation for the tendency for long-term growth
in the level of investment and, consequently, of the output, is fundamentally associated
with the ‘development factors’ (the constant d′ above). If the specic parameter that meas-
ures the intensity of those factors is positive, then the economy will present a long-term
growth trend. Absent those factors, the system converges to a stationary state. erefore, the
existence of a stable long-term growth trend depends on a constant expansion in the rate
of innovations. If the intensity of ‘development factors’ decrease in the long term, this will
determine, as a consequence, a tendency for deceleration of investment rate and, therefore,
of economic growth over time.9 In fact, Kalecki considers the following three factors to be
important for determining the long-term investment level: innovations, rentier’s savings,
and population growth.
Regarding innovations, which is the most important factor for the long-term dynamics,
he argues that their occurrence aects positively the decision to invest, and has a role that
is similar to prot variations. at is, a positive rate of new inventions has a direct impact
on investment decisions. It is worth noting that the conception of innovation in Kalecki is
very broad, including not only technological development, but also the introduction of new
products and access to new sources of raw materials, provided that these factors demand
investments in new xed capital goods. e two other factors pointed out above are seen as
aecting the growth rate of a capitalist economy: negatively in the case of rentiers’ savings,
and positively in the case of population growth.
To sum up, the main variable aecting investment is ‘expected protability that induces
capital accumulation [and] […] which makes investment possible, partly through internally
generated funds’ (Arestis 1996, 25).
(1)
It+𝜃
=a∕(1+c)S
t
+b
�
(ΔP
t
∕Δt)+e(ΔO
t
∕Δt)+d
�
t,

4 E. MALDONADO FILHO ET AL.
2.2. Institutionality of investment in the Keynesian view
In e General eory of Employment, Interest and Money (GT), Keynes argues that uctu-
ations in eective demand and employment levels occur because, in a world in which the
future is uncertain and unknown, individuals prefer to hold on money and, consequently,
their decisions on expenditures, consumption and, especially, investment, are postponed.
us, economic crises appear because money is an alternative form of wealth.10
If, according to the principle of eective demand (PED), the investment is the main
variable that determines income and employment, a question arises: Which factors inu-
ence investment? In Chapter 11 of GT, Keynes formalizes his investment theory in terms
of the ‘marginal eciency of capital’ (MEC), which refers to the discount rate that equals
the expected income stream (demand price) to the cost of investment (oer price). In other
words, the MEC shows the volume of investment that can be expected for each interest rate.
erefore, for Keynes, investment is determined by the equality between MEC and the cost
of capital. Furthermore, Keynes’s investment theory was developed under the hypothesis
that money matters for the performance of the economic activity.
is general idea is explored by post-Keynesian authors (Asimakopulos 1971; Carvalho
1988; Crotty 1992; Davidson 1972; Lavoie 2014; Minsky 1975) that show how nancial and
monetary conditions aect rms’ investments. Keynes’ arguments about uncertainty based
on unknown and immeasurable probability is discussed by Carvalho (1988). Crotty (1992)
argues that Keynes investment theory is superior to mainstream approaches. Davidson
(1972) shows the relationship between money, nancial institutions, and economic growth.
Lavoie (2014) provides theoretical rigor to the model by showing its coherence, while
Minsky (1975) develops, based on the idea of lender’s risk and borrowers’ risk, his nan-
cial fragility hypothesis.
Returning to Keynes, in a context of fundamental/radical uncertainty, in which is not
possible to foresee the future, investment is unstable due to the weak/fragile expectations
that form the MEC (since these are based on conventions). Furthermore, since liquidity
preference by individuals is a form of hedging against uncertainty regarding their trans-
action and production plans, thus conditioning the dynamics of the productive process,
investment instability (income and employment) is recurrent.
Chapter 12 presents the reasons to the volatility of investment. It shows that due to the
fact that investment decision-making is based on weak/fragile long-term expectations, the
volatility of investment occurs. In this context, the degree of condence and conventions,
or, more broadly, institutions, are crucial to induce the entrepreneurs’ investment decisions.
In Keynes’ words (2007, 161), a substantial part of decision-making ‘can only be taken as
a result of animal spirits’. Eichner and Kregel (1975, 1301) reinforce that the rate of discre-
tionary expenditures, that is, investment, depends ‘on the “animal spirits” of entrepreneur
and the volatility of their expectations’.
Keynes’s approach is peculiar, because his investment theory has a dynamics of its own,
since the process of investment nancing, that is, the role of credit, and not only of the
animal spirits, is essential for its execution. at investment-credit relationship became
known as the nancing-investment-savings-funding circuit.11 On the other hand, Keynes’s
main argument was based on the limitations of monetary policy and credit to inuence
the demand for investment. Such limitations were due to the low elasticity of MEC, the
impossibility of reducing the interest rate (faced with the ‘liquidity trap’) and to the liquidity

INTERNATIONAL REVIEW OF APPLIED ECONOMICS 5
preference by the nancial system, thus reducing the ow of credit. Finally, the lack of
investment would generate an insuciency of eective demand and recurrent economic
uctuations or crises. In short, from a Keynesian perspective, credit, expectations, and
interest rate are the main variables that aect investment.
3. The dynamics of investment in Brazil: an empirical analysis
is section deals empirically with both the evolution of investment and the possible deter-
minants suggested in the previous section. Despite the fact that the above theoretical analysis
shows many variables to explain investment in xed capital – such as entrepreneur’ stock
of capital, internal funds, debt, prots, aggregate economic activity or output, inventories,
expectations, innovations, rentiers’ savings, population, MEC, animal spirits, price of capital
goods, interest rates, number of rms, money holdings, institutions, economic policies, and
credit –, we select some of them that will be included in the statistical and econometric
analysis. ose selected were GDP growth (as a proxy of realized investment in the past),
prots, credit, interest rate, and expectations. e selection was based on two aspects: rst,
not all variables included in the model are available for the Brazilian economy. Second, even
when they are, they may not be adequate for a statistical treatment.
e empirical analysis is rst carried out by descriptive illustrations, without exhausting
the analysis of all possible determinants, and then by means of an econometric vector error
correction model (VECM) in order to deepen the understanding of the determinants of
investment in Brazil. As a caveat, one should be aware that working with data about the
Brazilian economy is another challenge, because of the frequent changes in methodology,
dierent annual bases of comparison, incompleteness, and discontinuity of the series.
12
is
limitation should be considered when evaluating the signicance and robustness of the tests
performed below. is is particularly important regarding variables used for applying the
theoretical approaches, which are complex and not easily replicable even for databases that
are more reliable. For example, the time lags discussed by Kalecki in his investment function
are not xed or determined, although in his statistical applications they vary between one
year and one quarter, and this imposes a diculty in the statistical treatment of his model
for Brazilian data. Most data required for estimating the macroeconomic models are updated
up to 2013, not capturing the descent into the 2014 stagnation and the 2015–2016 full-blown
crisis. Moreover, even for the series with information for the entire period, a lapse of 20years
may be relatively short, given changes in data collection methodology and possible noises
in the series, for estimating long-run relationships suggested by the VECM methodology.
3.1. Descriptive empirical analysis
Figure 1 shows the real growth rate for Growth Fixed Capital Formation (GFCF). e
contribution of GFCF to GDP growth, in percentage points, is also presented, showing a
behavior similar to the real growth of GFCF (correlation coecient of 0.99). As it would be
expected from Kalecki, investment growth, excluding changes in inventories, has short and
long-term uctuations. ere is a downward trend for both series until 2003, a resumption
of real growth from then to 2008 and another decline until 2013, with a sharp drop during
the global nancial crisis, and a strong cyclical, short-term expansion in 2010.

6 E. MALDONADO FILHO ET AL.
e evolution of the real index (at 1980 prices) of the two main components of investment
(machinery and equipment and structures) for the period is shown in Figure 2. Stagnation
in the neoliberal period (1990–2002) was followed by a strong expansion in investment in
machinery and equipment starting in 2003, and investment in structures shows a similar
pattern, with a less intense rhythm in the post-neoliberal period. e small drop in invest-
ment in structures aer the GFC suggests a positive eect of the counter-cyclical policies,
mainly the Program for Accelerating Growth.
Figure 3 illustrates the evolution of the GFCF/GDP ratio and the strong positive correlation
(coecient equal to 0.96) between real GDP growth and the growth of real GFCF, as expected
by the Kaleckian view, especially aer the GFC and the GR. is relationship is also suggested
by Figure 1, given the important contribution of investment to GDP growth. e GFCF/GDP
ratio uctuates between 15.3% and 20.7%, with a trend of gradual decline until 2003, and
a slight increase thereaer, stagnating aer the global crisis, and a falling trend aer 2011.
In the Kaleckian investment model, there is a clear relationship between investment and
GDP, albeit with lags. In the Keynesian theory, GDP depends on eective demand, and
investment is the most important component of the latter in terms of the multiplier eects.
e tight relationship between the variables in levels also suggests a strong relationship in
terms of growth rates.
e relationship between real GFCF (at prices of 2013), credit outstanding to the private
sector as a percentage of GDP, and the average annual real interest rate (SELIC) can be seen
in Figure 4. ere is a strong relationship (coecient equal to 0.84) between the credit/GDP
ratio and real investment as expected by Keynesian theory. Regarding the real interest rate,
up to 2003 the sharp decline observed had not aected the actual level of investment. From
then on, there is a clear inverse relationship between the two variables (correlation of −0.61).
-10%
-5%
0%
5%
10%
15%
20%
25%
1994 1995 1996 1997 1998 1999 2000200120022003 20042005 2006 20072008 20092010 20112012 2013
-2%
-1%
0%
1%
2%
3%
4%
5%
Real Growth of GFCF Contribution of GFCF to GDP Growth
Figure 1.Real growth rate of GFCF (left axis) and GFCF’s contribution to the GDP growth (right axis) in Brazil
(% annual): 1994–2013. Source: Elaborated by the authors using data from IPEADATA (2016). Information
on the base year for deflating nominal growth rates was not disclosed by the institute).
Note: The dotted line represents the best-fit polynomial (third degree) to the growth trend of GFCF.

INTERNATIONAL REVIEW OF APPLIED ECONOMICS 7
Figure 5 shows the relationship between the Index of Industrial Entrepreneur Condence
(IIEC) – a proxy for entrepreneur ‘expectations’ in which gures above and below 50 rep-
resent, respectively, optimism and pessimism, and real GFCF (at prices of 2013). e index
0
100
200
300
400
500
600
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 20052006 2007 2008 2009 2010 20112012 2013
GFCF - Machinery and Equipments GFCF - Structures
Figure 2.Index of real GFCF’s main components in Brazil (1994–2013) (1980=100). Source: Elaborated by
the authors using data from IPEADATA (2016).
-10%
-5%
0%
5%
10%
15%
20%
25%
1994 1995 19961997 19981999200020012002 20032004 2005 200620072008 20092010 201120122013
-1%
0%
1%
2%
3%
4%
5%
6%
7%
8%
GFCF Real Growth GFCF/PIB ratio Real GDP Growth
Figure 3.Real GDP growth rate (right axis), Real growth of GFCF (left axis), and GFCF/GDP ratio (left axis)
in Brazil (1994–2013). Source: Elaborated by the authors using data from IPEADATA (2016).

8 E. MALDONADO FILHO ET AL.
0
100,000
200,000
300,000
400,000
500,000
600,000
700,000
800,000
900,000
1,000,000
1994 1995 1996 1997 1998 19992000 2001 2002 2003 2004 2005 200620072008200920102011 20122013
0%
10%
20%
30%
40%
50%
60%
Real GFCFl Credit to the Private Sector/GDP Real Interest Ratel
Figure 4.Real GFCF (left axis, R$ million), credit to the private sector as % of GDP (right axis) and real
average annual interest rate (right axis) in Brazil (1994–2013). Source: Elaborated by the authors using
data from IPEADATA (2016) and BCB (2016).
0
100,000
200,000
300,000
400,000
500,000
600,000
700,000
800,000
900,000
1,000,000
1994 1995199619971998199920002001 2002 2003 2004 2005 20062007 2008 2009 20102011 2012 2013
0
10
20
30
40
50
60
70
80
Real GFCF Industrial Business Owners Confidence Index
Figure 5.Real GFCF (left axis, R$ million) and IIEC/CNI (1994–2013). Source: Elaborated by the authors
using data from IPEADATA (2016).

INTERNATIONAL REVIEW OF APPLIED ECONOMICS 9
series is relatively constant, with optimistic entrepreneurs throughout the period, and a
negligible fall in the index at the end of the period, but without any signicant impact on
real investment. e variable IIEC is not likely to capture the complexity of Keynesian
degree of condence on expectations. Surveys can measure a given state of condence,
but real investment decisions are carried out in a context of true uncertainty, that cannot
be measured in any meaningful way. e gure shows a growing investment for 2003 on,
but the index is more or less stable, and from 2011 to 2013 investment grows when the
condence index falls. So, this is not a good measure of Keynesian expectations and will
not be used in the econometric analysis.
Finally, Figure 6 displays the trends for real GFCF, real GDP (correlation between them
equal to 0.97) and real Gross Operating Surplus (correlation with real GFCF equal to 0.94).
ese descriptive ndings suggest that Kalecki’s view has something to say about investment
behavior in Brazil. We further investigate these issues by means of an econometric model.
3.2. Econometric analysis
e econometric literature regarding models based on the theories of Kalecki and Keynes is
meager, given the problems associated with collecting data faithful to the theories outlined
above. A book that seeks to estimate these relationships was put together by Baddeley (2003).
Among many dierent applied approaches to investment, she compares the conventional
model of Tobin’s Q with a Keynesian-Kaleckian model (or Post-Keynesian), emphasizing
the role of uncertainty and cyclical factors. For the estimated model of Tobin, investment
depends on current and lagged values of the stock of capital, Tobin’s Q, the price of capital
goods, capacity utilization, the volatility of the price of capital goods and the turnover of
the stock exchange. For the Keynesian-Kaleckian model, investment depends on current
and lagged values of prots’ growth, capital stock growth, corporate savings, a time trend,
capacity utilization, the volatility of the price of capital goods, and the turnover of stock
exchange. Annual data from the US economy for the period 1970–1998 were used, a short
horizon considering the high number of estimated parameters. She found that the combined
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
0
100,000
200,000
300,000
400,000
500,000
600,000
700,000
800,000
900,000
1,000,000
RealGFCF RealGDPRealGOS
Figure 6.Real GFCF, real GOS (right axis) and real GDP (1994–2013) (R$ million). Source: Elaborated by
the authors using data from IPEADATA (2016).

10 E. MALDONADO FILHO ET AL.
Keynesian-Kaleckian model is superior to the Tobin’s Q model, lending empirical support
to the former theoretical view.
Other econometric studies include Fazzari and Mott (1986–87), Holt and Pressman
(eds.) (2007) e Alexiou (2010). Fazzari and Mott (1986–87) use a panel data of 835 US
manufacturing rms to identify the determinants of investment for the 1970–1982 period.
ey found that internal liquidity matters for investment. ey also found that capacity
utilization and the level of interest commitments matter for capital accumulation. e
second part of the book edited by Holt and Pressman (2007) is dedicated to the empirical
analysis of business investment, containing four chapters that discuss R&D investment in
Australia, using Kalecki’s link between innovation and investment, a susceptibility model
of investment, related to the work of Courvisanos, the eects of speculation and bubbles on
investment based on Keynes and Kalecki (and conrming the validity of their approach),
and the role of equity supply for rms in the US. e studies reject the neoclassical per-
spective on investment and show the relevance of Kalecki and Keynes for understanding
capital accumulation patterns. Alexiou (2010) uses a panel data comprising G-7 rms for the
period 1972–2005. e author found out that prots, capital stocks (with a positive signal,
what is contrary to Kalecki’s view), GDP, and capacity utilization are statistically signicant
variables for investment, validating the Keynesian and Kaleckian theories.
e approach adopted in this study is similar to Baddeley’s, estimating a modied and
combined linear Kaleckian-Keynesian model, but introducing only a small number of
explanatory variables in logs of levels and parameters to be estimated.13 e purpose of
this specication is to provide a theoretical convergence, testing the combined power of
the most important variables in each macro theory, but with adjustments imposed by data
availability. From the Kaleckian side, the model includes, besides the dependent variable
real GFCF
14
(deated by the price index of capital goods) and real GDP (including GFCF on
the expenditure side) for the period and real prots (Gross Operational Surplus – GOS).15
e problem with specications inspired by the work of Kalecki is that the theory com-
bines variables in levels and dierentiated (rates of change). e latter tend to behave as
stationary processes, potentially changing the order of integration of the variables in the
model. And they can have negative values, making it impossible to perform logarithmic
transformation of the data. Important variables traditionally included in neo-Kaleckian
models of investment such as capacity utilization, were le out to simplify the discussion,
the same happening to scal policy indicators (decits and public debt), exchange rates,
capital ows, depreciation, expectations, and autoregressive terms for investment.
From the Keynesian approach, we le out variables suggested by Davidson (1994),
16
such
as the deator for prices of capital goods, the industrial entrepreneurs condence index
(IIEC/CNI) as a proxy for expectations regarding quasi-rents,17 and the number of rms.
e model includes, besides real GFCF, the short-term real interest rate and the credit to
the private sector/GDP ratio. e latter is a way to incorporate the initial stage proposed
by the theory of the circuit nance–investment–savings–funding. Just like the Kaleckian
approach, despite the dierences between the two theories in terms of dynamics, the insti-
tutional determinations of investment in Keynes are dicult to measure and were not
treated in this work. So, the model is given by:
(2)
It=a+bOYt+b1Pt+b2Rt+b3Ct+
𝜀
t,

INTERNATIONAL REVIEW OF APPLIED ECONOMICS 11
where I is real GFCF, Y is the real GDP, P is the real GOS, R is real interest rate (SELIC) and
C is the credit to private sector to GDP ratio.
e econometric analysis of time series involves the identication of the stationary of the
data to avoid spurious OLS regressions. e visual inspection of the gures above suggests
that the selected variables, except growth of investment and growth of GDP, are non-sta-
tionary, that is, they have a unit root, or at least they are stationary around a deterministic
time trend. Dickey-Fuller DF-GLS unit root tests with data adjustment for trends were
implemented, including variables in levels and with and without a lag. e number of lags
used in the tests was suggested by the Minimum AIC (MAIC) developed by Ng and Perron.
Table A2, in annex, summarizes the results. e variables used in the Kaleckian-Keynesian
model in levels are not stationary in all specications considering the time trend and with
one lag, suggesting estimation using VECM.
en the number of lags was determined based on FPE, AIC, SBIC, HQIC, and LR
statistics. e statistical tests suggest the use of four lags for determining the existence of a
cointegrated vector, with or without a constant, for some statistics, and two lags for others.
e next step involves the cointegration analysis. e specication of the equation was cho-
sen based on the existence of at least one cointegration relationship and the adherence to the
theoretical combined model. Using the multiple-trace test method proposed by Johansen
(1995), the models with an unrestricted constant, restricted constant, linear trend, restricted
trend and no trend, lagged four or three years, yielded no statistical result. Reducing the
lags to two years, there were 4 cointegrating equations, except for the specication with no
constant and no trend, which yielded 3 cointegrating equations.18
Finally, we proceeded to the estimation of the VECM.19 e analysis below considers only
the cointegrating vector for the long-term relationship, with two lags for the independent
variables, not the coecients that measure short-term adjustments to restore ‘equilibrium’
when the relationship between the variables is disturbed by random shocks. e ve dierent
specications (respectively with an unrestricted constraint, with a restricted constrains, with
a linear trend, with a restricted trend, and with no constant nor trend) are presented below:
(i) It – 1.14Yt – 0.78Pt+0.03Rt – 0.03Ct+15.04=εt;
(ii) It – 0.67Yt – 0.81Pt+0.02Rt – 0.22Ct+8.06=εt;
(iii) It – 0.09Yt – 0.61Pt+0.03Rt – 0.55Ct+0.00T – 3.84=εt;
(iv) It – 0.20Yt – 0.62Pt+0.03Rt – 0.49Ct+0.00T – 1.92=εt;
(v) It – 0.07Yt – 0.89Pt – 0.00Rt – 0.46Ct=εt.
e coecients for real GDP and the credit-to-GDP ratio show a wide variance between
the specications. In the rst specication for the cointegrating equation, all signs are as
expected, but the credit-to-GDP ratio coecient was not statistically signicant at 10%.
In the second specication, all the signs were the expected and all coecients were signif-
icant. In the third specication, the signs were as expected, but the real GDP coecient
was not signicant. e fourth specication had all the coecients with the expected sign
and statistically signicant. Since the coecient for the time trend is zero, it seems that the
second specication is the best one. Finally, the h specication had an unexpected sign
for the coecient interest rate, but it was not signicant, like the coecient for real GDP.
Since the variables are cointegrated, there must be a joint causality relationship between
them, possibly according to the theories of Kalecki and Keynes. But since the cointegration
treatment creates statistical problems for testing for Granger non-causality, a VAR with

12 E. MALDONADO FILHO ET AL.
dierentiated variables was estimated. e Phillips–Perron test was implemented, and all
dierentiated variables were stationary. e Augmented Dickey–Fuller test equally rejected
the null of unit root for the dierentiated variables. Yet, the modied Dickey–Fuller test
failed to reject the null for the dierentiated variables, with or without a trend. e visual
inspection also suggests that the dierentiated variables are stationary. e VAR results are
not reported here, since the goal is to perform the Granger non-causality test. e results
were as follows. Using two lags in the VAR with all ve variables in the order of the Equation
(2) above, at 10% we fail to reject all Granger non-causality tests but the ones for the real
interest rate, meaning that the coecients on the lags of all variables are not jointly zero
in the equation for the real interest rate. is suggests that all variables Granger-cause the
real interest rate (investment and output on the demand side, and prots and credit on the
supply side). e fact that real interest rates do not aect investment is a claim made by
Kalecki, although he discussed the long-run interest rate. On the other hand, Arestis (1996)
argues that the interest rate matters for the Kaleckian view.
Using one lag in the VAR with all ve variables in the order of the Equation (2), at 10%
we reject the null of Granger-non causality for the coecients on the lag of the credit to
GDP ratio to predict real GFCF, that is, credit Granger-causes investment, as predicted by
the circuit nance–investment–saving–funding. e same was found for the coecients
on the lag of real GFCF and real GDP to predict real GOS, conrming the Kaleckian view.
is was also conrmed by bivariate VARs with two lags. ere is no Granger-causality rela-
tionship between real GFCF and real GDP, but we reject that real GFCF does not Granger-
cause real GOS. Real GFCF also Granger-causes the real interest rate, and credit to GDP
ratio Granger-causes real GFCF. It is worth mentioning that those results were obtained by
dierentiated variables, so the relationships are between the changes in the variables, unlike
the VECM results, that are reported for the levels of the variables.
ere are not many econometric studies dealing with the relationship between invest-
ment instability and growth instability. Ramey and Ramey (1995), from a conventional
perspective, nd that output volatility causes lower output growth, but investment is not
relevant for output instability. To the best of our knowledge, there are no studies about this
relationship for Brazil. In this study, since there is a long-run relationship between capital
accumulation and real GDP in levels, the same must be true for the variables in dierences
(as approximations for rates of change). e Granger-non causality tests are not conclusive,
but Figure 1 suggests that investment has been unstable in Brazil, and that its contribution
to GDP ts this volatility. Furthermore, Figure 3 shows that real GDP and real investment
growth move together over their cycles and their trends. erefore, instability in capital
accumulation is likely to feed output instability.
Overall, in terms of the VECM results, prots and the real interest rate had signicant
coecients and the expected sign in all specications, vindicating both Kalecki and Keynes
for the long-run equilibrium of capital accumulation in Brazil. is means that when a
positive shock occurs, there is an increase in prots, a decrease in the real interest rate, and
a simultaneous increase in investment to restore the long-run equilibrium. Real GDP is
not always signicant for the long-run equilibrium, like the credit-to-GDP ratio, but they
behave as expected respectively by the Kaleckian and the Keynesian theories. So, overall
the two approaches suggest good analytical elements to understand the empirical behav-
ior of investment in Brazil. e Granger non-causality tests suggest dierent results, with
interest rates being preceded by decisions of investment, production, credit, and realization

INTERNATIONAL REVIEW OF APPLIED ECONOMICS 13
of prots. Credit relative to GDP can forecast investment spending, and spending and
production help predicting prots.
e results suggest, and visual inspection of Figures 1 and 3 is clear, that the instability
of output growth in Brazil cannot be separated from instability of investment, since both
change in the same direction in the long run. e same is true for capital accumulation
and prots. Prots grow when investment grows, creating the internal funds for the next
round of accumulation. So, investment is self-nanced in the long run. However, credit
seems to play an important role, closing the gap between the time of investment and the
time of earning prots. And real interest rates represent a drag on the investment in the long
run, perhaps less in terms of nancing costs than in terms of providing another source of
income in nancialized economies. is means that the current crisis in Brazil is a result
of the substantial reduction of investment that took place in 2014 and 2015. Output could
not grow in such a situation. Despite the rapid growth of credit to the private sector relative
to the growth of the output in the previous years, it is possibly still insucient to foster a
sustained level of investment. At the same time, falling but still high real interest rates are
another source of dismal investment levels.
e scarcity of data for a longer period and the lack of variables that capture the key
aspects of theories of investment in the tradition of Kalecki and Keynes, such as expecta-
tions, prevent a dynamic treatment appropriate to completely understand the determinants
of investment in Brazil. Additional research is therefore needed in order to better measure
and develop the variables consistent with the theories of investment. Similarly, even with
these limitations, other statistical tests could illuminate dierent dimensions of research
and show the key constraints to dynamic and sustainable growth of the Brazilian economy.
4. Concluding remarks
e paper presented theoretical and empirical evidence to interpret the weak performance of
the Brazilian economy from 1994 to 2013. In short, the result can be explained by the aggre-
gate investment behavior and the macroeconomic environment. e weak performance
can be explained by a structural lack of investment, with a cyclical pattern consistent with
high levels of instability. us, it seems to be in accordance with the investment theories of
Kalecki and Keynes and their emphasis on the role of output, prots, credit, and interest
rates. Of course, their theories are complex and cannot be fully captured by existing data
and econometric techniques.
e econometric results provide additional support to the conclusion that the trend of
a volatile and sluggish growth in Brazil over the past 20years is related to a volatile invest-
ment that cannot be sustained at high levels. e expansion of credit and the reduction
in real interest rates were insucient to sustain the expansion of capital accumulation in
Brazil. e ndings point to the fact that high real short-term interest rates in Brazil, albeit
showing a downward trend, appear as a possible constraint to sustained expansion of the
accumulation, especially when high short-term rates index the longer term rates. Also, there
are limits to the expansion of credit as a driver of GFCF if prots (internal funds or savings)
do not generate the funding necessary for closing the circuit. And prots depend on capital
accumulation according to VAR results, and both move together in the long run according
to the VECM estimated. us, nance and funding problems may also help explain the
problems of capital accumulation and growth in Brazil. ese trends ultimately points to

14 E. MALDONADO FILHO ET AL.
low levels of economic activity caused by a lack of adequate levels of eective demand,
since both capital accumulation and production move in the same direction in the long
run. If investment falters, the economy must stagnate or collapse, as has happened during
the 2015–2016 crisis. e Brazilian economy in the past decades in general and during the
current crisis in particular is just the proof of that.
Notes
1. Between July 1994, when the Real was created as a legal tender currency, and December
2014, the average ination rate was 7.9% per year. Authors’ calculations based on statistical
information from IPEADATA (2016).
2. In the 1950s, 1960s, 1970s, 1980s, 1990s and 2000s, the average annual growth rates of GDP
were, respectively, 7.1, 6.1, 8.8, 3.0, 1.9 and 3.4%. Authors’ calculations based on statistical
information from IBGE (2016).
3. Authors’ calculations based on Table A1, in the annex.
4. Based on both theoretical models, this section shows that the main variables that aect
investment and should be dealt with in an empirical analysis are capitalist savings or retained
prots, changes in prots, output growth, technical progress, expectations, credit (or debt)
and interest rates.
5. We must emphasize that the concepts of ‘short’ and ‘long’ term used by Kalecki are dierent
from the Marshallian concepts. e separation by Kalecki of short and long term seems to
be related to the dierence between an analysis that ignores the factors that inuence long-
term behavior (therefore, an analysis of short term considers only cyclical uctuations in a
static economy), and a long-term analysis that takes into account the factors that inuence
economic growth.
6. For additional details, see, for instance, Arestis (1996), Sawyer (1985), Duménil and Lévy
(2012), Lavoie (2014), and Skott (2012). Arestis (1996) highlights the importance of prots
and interest rates for investment in Kalecki’s view. Sawyer (1985) presents the main features
of the dierent Kaleckian theories of investment, business cycles, and growth. Duménil and
Lévy (2012) and Lavoie (2014) develop formal models for Kaleckian investment theory, adding
mathematical rigor to the analysis, and conrming its theoretical relevance. Finally, Skott
(2012) presents a few shortcomings in the Kaleckian theory. He challenges a few behavioral
aspects of Post-Kaleckian investment theories (more than Kalecki’s proper), mainly the ones
related to the role of capacity utilization. He also criticizes the lack of empirical support for
investiment functions with specic results regarding capacity utilization.
7. us, for Kalecki (1969, 91) ‘the variety in the size of enterprises in the same industry at a given
time can be easily explained in terms of dierences in entrepreneurial capital’. Obviously, this
does not answer the question of why the amount of entrepreneurial capital diers between
rms in the same sector. In other words, why do some rms have a larger entrepreneurial
capital than others?
8. He assumes that a/(1+c) < 1.
9. It should be noted that, in Kalecki’s theory, the trend for decreasing intensity of ‘development
factors’ implies the decrease of prot rate and an increase of idle capacity level. However, it is
important to highlight that the direction of causality is dierent from that proposed by Marx.
For Marx, it is the decreasing trend of prot rate that results from technical progress (or, in
Kalecki’s terminology, from the increase of ‘development factors intensity’), which, in turn,
determines the deceleration of the capital accumulation process (and therefore of economic
growth) and ultimately causes the periodical crisis of the capitalist system.
10. Keynes (2007, Chapter 17) shows that there is an insuciency of eective demand – and,
therefore, of investment – due to the fact that individuals allocate income in the form of non-
reproducible wealth instead of allocating it for the acquisition of goods produced by work.
11. For additional details, see Studart (1993).
12. For a specic discussion about the data on investment in Brazil, see Santos et al. (2015).

INTERNATIONAL REVIEW OF APPLIED ECONOMICS 15
13. Another study on the determinants of investment in Brazil, but focusing on private investment
for the period 1970–2005, with a dierent econometric methodology, is provided by Luporini
and Alves (2010). e use of variables in levels in our model shows the limits of statistics
and econometrics to accommodate more complex macroeconomic models that incorporate
changes in the variables considered by capitalists in their decisions to invest.
14. is study does not split the GFCF into residential xed investment and business xed
investment, which would allow us to isolate and focus on the capitalists’ accumulation
decisions. For a detailed study segregating investment by sectors, see Bielschowsky, Sque,
and Vasconcellos (2015).
15. e institute responsible for supplying the series, IBGE, uses two dierent methodologies for
calculating gross operational surplus. ere is therefore one series with data from 1990 to 2009
and another one with data ranging from 2000 to 2013. We used the rst one for the period
1994–2009 and the second one for the period 2010–2013, since the dierence between the
series is not large. Prots were adjusted based on the GDP deator for year 2013 prices, the
GFCF was corrected by the deator for prices of capital goods, and the real value of capital
stock at 2013 prices is supplied directly by IPEADATA (2016).
16. According to Davidson (1994), investment (or the maximum amount of capital goods
desired by the rms) depends on the market price of capital goods, quasi-rent expectations,
interest rates, and the number of rms in the economy. Analytically, DI=f(pI, i, α, β), where
pI= market price, i=quasi-rent expectations, α= interest rate and β= number of rms,
where fpI<0, fi>0, fα<0 and fβ>0.
17. IPEADATA (2016) provides data about the condence index since 1999, which increases
the diculty of estimation by reducing the degrees of freedom. Moreover, the way the index
is constructed makes the variable stationary because it uctuates around a reference point.
Anyway, it is unclear to what extent this type of indicator captures expectations as theorized
by Keynes, since radical Keynesian uncertainty cannot be measured, as discussed above.
18. e specic statistics of trace and eigenvalues are not reported here. e complete set of
auxiliary statistics can be obtained from the authors upon request.
19. e model assumes a long-term stationary equilibrium relationship (cointegration) between
non-stationary variables, with the trajectory of the variables having corrections or adjustments
in the short term. e discussion regarding the extent to which this traditional concept of
equilibrium and adjustment process can be consistently applied to the theory of Kalecki and
Keynes is beyond the scope of this work.
Acknowledgement
e authors thank the anonymous referees for comments and suggestions. All remaining errors are
the authors’ responsibility.
Disclosure statement
No potential conict of interest was reported by the authors.
References
Alexiou, C. 2010. “A Keynesian-Kaleckian Model of Investment Determination: A Panel Data
Investigation.” Journal of Post Keynesian Economics 32 (3): 427–444.
Arestis, P. 1996. “Kalecki’s Role in Post-Keynesian Economics: An Overview.” In An Alternative
eory: e Kaleckian Model and Post-Keynesian Economics, edited by J. King, 11–34. Norwell,
MA: Kluver Academic Publishers.
Asimakopulos, A. 1971. “e Determination of Investment in Keynes’s Model.” e Canadian Journal
of Economics 4 (3): 382–388.

16 E. MALDONADO FILHO ET AL.
Baddeley, M. 2003. Investment: eories and Analysis. London: Palgrave MacMillan.
BCB (Brazilian Central Bank). 2016. Séries Temporais [Time Series], May 4. http://www.bcb.gov.br.
Bielschowsky, R., G. Sque, and L. F. Vasconcellos. 2015. “Evolução dos investimentos nas três frentes
de expansão da economia brasileira na década de 2000 [e evolution of investment during the
Brazilian economic growth in the 2000s].” Texto para Discussão nº2063/IPEA, Brasília: IPEA.
de Carvalho, F. C. 1988. “Keynes on Probability, Uncertainty, and Decision Making.” Journal of Post
Keynesian Economics 11 (1), Fall: 66–81.
Crotty, J. R. 1992. “Neoclassical and Keynesian Approaches to the eory of Investment.” Journal of
Post Keynesian Economics 14 (4), Summer: 483–496.
Davidson, P. 1972. Money and the Real World. New York: Wiley.
Davidson, P. 1994. Post Keynesian Macroeconomic eory: A Foundation Successful Economic Policies
for the Twenty-rst Century. Aldershot: Edward Elgar.
Duménil, G., and D. Lévy. 2012. “Modeling Monetary Macroeconomics: Kalecki Reconsidered.”
Metroeconomica 63 (1): 170–199.
Eichner, A. S., and J. A. Kregel. 1975. “An Essay on Post-Keynesian eory: A New Paradigm in
Economics.” Journal of Economic Literature 13 (4), December: 1293–1314.
Fazzari, S., and T. Mott. 1986–87. “e Investment eories of Kalecki and Keynes: An Empirical
Study of Firm Data, 1970–1982.” Journal of Post Keynesian Economics 9, Winter: 171–187.
Holt, R. P. F., and S. Pressman, eds. 2007. Empirical Post Keynesian Economics: Looking at the Real
World. Armonk, NY: M.E.Sharpe.
IBGE. 2016. Indicadores [Indicators]. Accessed May 4. http://www.ibge.gov.br.
IPEADATA. 2016. Séries Históricas [Historical Series]. Accessed May 4. http://www.ipeadata.gov.br.
Johansen, S. 1995. Likelihood-Based Inference in Cointegrated Vector Autoregressive Models. Oxford:
Oxford University Press.
Kalecki, M. 1937. “e Principle of Increasing Risk.” Economica 4 (New Series): 440–447.
Kalecki, M. 1969. eory of Economic Dynamics: An essay on Cyclical and Long-Run Changes in
Capitalist Economy. New York: Augustus M. Kelley.
Keynes, J. M. 2007. e General eory of Employment, Interest and Money. London: Palgrave
Macmillan.
Lavoie, M. 2014. Post-Keynesian Economics. Cheltenham: Edward Elgar.
Luporini, V., and J. Alves. 2010. “Investimento privado: uma análise empírica para o Brasil [Private
investment: an empirical analysis to Brazil].” Economia e Sociedade 19 (3), dezembro: 449–475.
Minsky, H. 1975. John Maynard Keynes. New York: Columbia University Press.
Ramey, G., and V. Ramey. 1995. “Cross-country Evidence on the Link between Volatility and Growth.”
American Economic Review 85 (5), December: 1138–1151.
Santos, C. H. M., A. M. Modenesi, G. Sque, L. Vasconcelos, M. Mora, T. Fernandes, T. Moraes, R.
Summa, and J. Braga. 2015. “Revisitando a dinâmica trimestral do investimento no Brasil: 1996–
2012 [Revisiting the quarterly investment dynamics in Brazil: 1996–2012].” Texto para Discussão
nº5/IE-UFRJ. Rio de Janeiro: IE-UFRJ.
Sawyer, M. 1985. e Economics of Michał Kalecki. London: Macmillan.
Skott, P. 2012. “eoretical and Empirical Shortcomings of the Kaleckian Investment Function.”
Metroeconomica 63 (1), February: 109–138.
Studart, R. (1993). O sistema Financeiro e o nanciamento do crescimento uma alternativa pós-
keynesiana à visão convencional [e nancial system and nancing growth: a post Keynesian
alternative to the conventional view]. Revista de Economia Política, vol.13, 1(49), Janeiro-Março:
101–118.

INTERNATIONAL REVIEW OF APPLIED ECONOMICS 17
Annex: Macroeconomic data and stationary tests
Table A1.Selected macroeconomic indicators for the brazilian economy.
Indicators/years 1995 1996 1997 1998 1999 2000 2001 2002 2003
Inflation rate (%) 22.41 9.56 5.22 1.66 8.94 5.97 7.67 12.53 9.3
Growth rate (GDP) (%) 4.2 2.1 3.4 0.0 0.3 4.3 1.3 3.1 1.2
Average interest rate
(SELIC) (%)
54.5 27.5 25.0 29.4 26.1 17.6 17.5 19.1 23.3
Average exchange rate
(R$/USD)
0.92 10 1.08 1.16 1.81 1.83 2.35 2.93 3.08
Trade balance (USD
Billion)
−3.5 −5.6 −6.8 −6.6 −1.2 −0.7 2.6 13.1 24.8
Current account (USD
Billion)
−18.4 −23.5 −30.5 −33.4 −25.3 −24.2 −23.2 −7.6 4.2
Foreign reserves (USD
Billion)
51.8 60.1 52.2 44.6 36.3 33.0 35.9 37.8 49.3
Fiscal surplus/GDP (%) 0.2 −0.1 −0.9 0.0 3.2 3.5 3.6 3.9 4.3
Net public debt/GDP (%) 29.1 29.6 30.4 35.4 44.5 45.5 48.4 50.5 52.4
Indicators/years 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Inflation rate (%) 7.6 5.69 3.14 4.46 5.9 4.31 5.91 6.5 5.84 5.91 6.41
Growth rate (GDP) (%) 5.7 3.1 4.0 6.0 5.2 −0.2 7.6 3.9 1.8 2.7 0.1
Average interest rate
(SELIC) (%)
16.2 19.1 15.3 12.0 12.7 10.1 9.9 11.75 8.63 8.29 10.96
Average exchange rate
(R$/USD)
2.92 2.43 2.17 1.92 1.83 2.0 1.76 1.67 1.95 2.16 2.36
Trade balance (USD
Billion)
33.6 44.7 46.5 40.0 24.7 24.6 20.3 29.8 19.4 2.6 −3.9
Current account (USD
Billion)
11.7 14.0 13.6 1.5 −28.3 −24.3 −47.5 −52.6 −54.2 −81.4 −90.9
Foreign reserves (USD
Billion)
52.9 53.8 85.8 180.3 193.8 238.5 288.6 352.0 373.1 358.8 363.8
Fiscal surplus/GDP (%) 4.8 4.3 4.0 3.91 4.1 2.1 2.8 3.1 2.4 1.9 −0.6
Net public debt/GDP (%) 47.0 46.5 44.7 43.0 36.0 43.0 39.1 36.5 35.1 33.8 35.3
Source: BCB (2016), IBGE (2016) and IPEADATA (2016).
Table A2.Stationarity tests.
Variable (period)
DF-GLS test with linear
trend (number of lags
in parenthesis)
DF-GLS without
linear trend (number
of lags in parenthesis) Stationary?
Ln real GFCF (1994–2013) Fail to reject H0 at 10% (1) Fail to reject H0 at 10% (1) No
Ln real GFCF−1 (1995–2013) Fail to reject H0 at 10% (1) Fail to reject H0 at 10% (1) No
Ln real GDP (1994–2013) Fail to reject H0 at 10% (1) Fail to reject H0 at 10% (1) No
Ln real GDP−1 (1995–2013) Fail to reject H0 at 10% (1) Fail to reject H0 at 10% (1) No
Ln real profits (1994–2013) Fail to reject H0 at 10% (1) Fail to reject H0 at 10% (1) No
Ln real profits−1 (1995–2013) Fail to reject H0 at 10% (1) Fail to reject H0 at 10% (1) No
Ln real interest rate (1994–2013) Fail to reject H0 at 10% (1) Fail to reject H0 at 10% (1) No
Real interest rate−1 (1995–2013) Fail to reject H0 at 10% (1) Fail to reject H0 at 10% (1) No
Credit to the private sector/GDP (1994–2013) Fail to reject H0 at 10% (1) Fail to reject H0 at 10% (1) No
(Credit to the private sector/GDP)-1 (1995–2013) Fail to reject H0 at 10% (1) Fail to reject H0 at 10% (1) No