Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
Money growth and inflation rate: evidence from Angola from 2014 to 20231
Alexandre Ernesto da Costa Antonio2
Recibo: 20.06.2024
Aceito: 27.06.2024
Publicado: 30.09.2024
Abstract: This paper investigates the impact of the money supply on inflation in Angola from 2014 to 2023, by assessing
the effect of money supply on inflation in Angola and to examine how the effects of money growth on inflation rate has evolved over
time. For this purpose, we use money aggregate M2, and national price index data published by Banco Nacional de Angola (BNA,
hereafter), that it is analysed using Vector Autoregressive model (VARM, hereafter) and rolling VARM. The results suggest that
firstly, money supply does affect inflation in Angola and its impact is positive, in other words, a 1% increase in money supply leads
inflation to rise by 0.11%. Thus, the greater the money supply the greater the inflation. Secondly, money supply and past inflation
may explain nearly 67% of current inflation in Angola. Third, the time-varying elasticity of inflation rate with respect to money
supply has changed dramatically over time and its value per period is mostly positive, confirming that throughout the period money
growth affect constantly inflation. These findings have considerable implications for the role of monetary policy in the short term to
control inflation rate in Angola once it reinforces the views to BNA deliver a discretionary monetary policy during the periods of
increase in government activity. For the future, further analyses is requires, using a set of data much more robust and oil markets
impacts on inflation through exchanges rates.
Keywords3: money growth; money supply; inflation; VAR; time-varying analysis.
Crescimento monetário e taxa de inflação: evidências de Angola de 2014 a 2023
Resumo: Este artigo investiga o impacto da oferta de moeda na inflação em Angola de 2014 a 2023, avaliando o efeito da
oferta de moeda na inflação em Angola e examinando como os efeitos do crescimento da moeda na taxa de inflação evoluíram ao
longo do tempo. Para esse propósito, usamos o agregado monetário M2 e os dados do índice nacional de preços publicados pelo
Banco Nacional de Angola (BNA, a seguir), que são analisados usando o modelo Vector Autoregressivo (VARM, a seguir) e o
VARM contínuo. Os resultados sugerem que, em primeiro lugar, a oferta de moeda afeta a inflação em Angola e seu impacto é
positivo, em outras palavras, um aumento de 1% na oferta de moeda leva a inflação a aumentar em 0,11%. Assim, quanto maior a
oferta de moeda, maior a inflação. Em segundo lugar, a oferta de moeda e a inflação passada podem explicar quase 67% da inflação
atual em Angola. Terceiro, a elasticidade variável no tempo da taxa de inflação em relação à oferta de moeda mudou drasticamente
ao longo do tempo e seu valor por período é principalmente positivo, confirmando que ao longo do período o crescimento da moeda
afeta constantemente a inflação. Essas descobertas têm implicações consideráveis para o papel da política monetária no curto prazo
para controlar a taxa de inflação em Angola, uma vez que reforça as visões de que o BNA deve entregar uma política monetária
discricionária durante os períodos de aumento da atividade governamental. Para o futuro, análises adicionais são necessárias, usando
um conjunto de dados muito mais robustos e os mercados de petróleo impactam a inflação por meio de taxas de câmbio.
Palavras-chave: crescimento monetário; oferta de moeda; inflação; VAR; análise de variação temporal.
Crecimiento monetario y tasa de inflación: datos de Angola de 2014 a 2023
Resumen: Este trabajo investiga el impacto de la oferta monetaria sobre la inflación en Angola desde 2014 hasta 2023,
evaluando el efecto de la oferta monetaria sobre la inflación en Angola y examinando cómo los efectos del crecimiento del dinero
sobre la tasa de inflación han evolucionado con el tiempo. Para este propósito, utilizamos el agregado monetario M2 y los datos del
índice de precios nacionales publicados por el Banco Nacional de Angola (BNA, en adelante), que se analizan utilizando el modelo
vectorial autorregresivo (VARM, en adelante) y el VARM móvil. Los resultados sugieren que, en primer lugar, la oferta monetaria
afecta la inflación en Angola y su impacto es positivo, en otras palabras, un aumento del 1% en la oferta monetaria hace que la
inflación aumente un 0,11%. Por lo tanto, cuanto mayor sea la oferta monetaria, mayor será la inflación. En segundo lugar, la
oferta monetaria y la inflación pasada pueden explicar casi el 67% de la inflación actual en Angola. En tercer lugar, la elasticidad
variable en el tiempo de la tasa de inflación con respecto a la oferta monetaria ha cambiado drásticamente con el tiempo y su valor
por período es mayoritariamente positivo, lo que confirma que a lo largo del período el crecimiento del dinero afecta
constantemente a la inflación. Estos hallazgos tienen implicaciones considerables para el papel de la política monetaria en el corto
plazo para controlar la tasa de inflación en Angola, ya que refuerzan la opinión de que el BNA debe aplicar una política monetaria
discrecional durante los períodos de aumento de la actividad gubernamental. Para el futuro, se requieren más análisis, utilizando un
conjunto de datos mucho más robustos y los impactos de los mercados petroleros en la inflación a través de los tipos de cambio.
Palabras clave: crecimiento monetario; oferta monetaria; inflación; VAR; análisis variable en el tiempo.
1 DOI: https://dx.doi.org/10.4314/academicus.v2i2.3
2 Instituto Superior Politécnico de Tecnologias e Ciências (ISPTEC). E-mail: alexandre.antonio@isptec.co.ao
3 JEL codes: C01, C12, C32, E31, E51
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
Introduce
Highly volatile inflation in Angola since 2016 has forced profound changes in the path of
monetary and exchange rate policies in the country. Available evidence suggests that the more
volatile the inflation, the greater the effect of money supply on the inflation rate (Papadia and
Cadamuro 2021). However, it is unclear if it is true for Angola, particularly in last 9 years
characterized by great inflation, combined either with low economic growth or economic
depression. To the best of my knowledge, there is no public published evidence of how money
supply has affected inflation in Angola in last 9 years, and whether its effects have changed
dramatically throughout the last 9 years.
Available evidence in the literature suggests that there is a positive relationship between
money supply and inflation, that is, as money supply increases, inflation increases, but it may not be
entirely true when inflation is less volatile (Papadia and Cadamuro, 2021). In SSA economies, such
as South Africa, Nigeria and Tanzania is observed similar patterns, that is, money supply has a
strong positive relationship with inflation (Mpofu, 2011; Mbongo et. al., 2014; Evans, 2019).
Evidence for Angola is lacking, but theoretically, is argued that money supply is correlated with
inflation. However, the following question remains uncertain: how money supply has affected
inflation in Angola and how its effects has changed dramatically over time?
This paper aims to bridge this gap by assessing the effect of money supply on inflation in
Angola and to examine how the effects of money growth on inflation rate have evolved over time.
To do so, the following hypotheses are tested: H1 - money growth does affect inflation rate in
Angola; H2 - money growth does not affect inflation rate in Angola; H3 - money growth effect on
inflation rate in Angola has changed dramatically over time and; H4- money growth effect on
inflation rate in Angola has not changed dramatically over time.
For this purpose, money aggregate M2, and national price index data published by Banco
Nacional de Angola (BNA, hereafter) are used, that it is analysed using Vector Autoregressive
model (VARM, hereafter) and rolling VARM. The rationale to study Angola’s case is given by the
following reasons. First, availability of new economic facts provided by the new exchanges rates
regime and discretionary monetary policy since 2018. Second, its massive economic reliance on oil
revenue brings several challenges to monetary authorities.
The rest of the paper proceeds as follows. Section 2 reviews literature review. Section 3
presents an overview of inflation in Angola, and recent developments. Section 4 present empirical
strategy, data, and presents and discusses our empirical results, respectively. Section 5 summarizes
the paper and draws policy implications.
Literature review
Available evidence from the impact of money supply on inflation rate is generally
supportive that money supply does affect levels of prices, although there might exist differences in
the degree of its effects between developing countries, commodity-rich countries, and developed
countries. Borio et., al. (2023) examines the link between money growth and inflation using a crosscountry analysis (United States, Euro area, United Kingdom, Canada, Brazil, and Thailand) and
their study indicates that, for the period 2015-2022, money growth and inflation have been closely
linked, despite the absence of robust empirical evidence. This result reached by Borio et., al. (2023)
is in line with what was found years ago in a study carried out by Roffia and Zaghini (2007)
focusing on 15 industrialised economies, and by Bozkurt (2014) focusing on Turkiye.
A study carried out by Papadia and Cadamuro (2021) concluded that in countries in which
inflation is volatile, money supply does help to explain inflation, but when inflation is not volatile,
there is no evidence of money supply effects on inflation. Thus, this result implies that the
relationship between money supply and inflation is no constant or may change over time. For
43
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
instance, Roshan (2014) examines the inflation and money supply growth in Iran, using VAR
models and finds evidence a bidirectional causality between money supply and inflation,
concluding that inflation may have a feedback impact on money growth which generates more
inflation.
Many other studies are supportive to impact of money supply on inflation. However,
recently, there has been some controversy about it. There is a debate about money exogeneity
versus money endogeneity and it has gained force around the world (Tobin, 1970; Fontana and
Setterfield, 2009; Thwaini and Hamdan, 2017; Sieron, 2019). Some defendthat money is indeed
endogenously created (Fontana and Setterfield, 2009), in contrast to monetary view (money
exogeneity). This may explain why central banks have been using massively quantitative easing to
boost economic growth.
Inflation in Angola
Before we proceed, it seems necessary to explain the main characteristics of Angola’s
economy. Angola experienced a strong economic growth between 2002 and 2014, with average
growth rates around 10%, and in 2022 it grew only 2.8%, after reached nearly 0% in 2021 and five
(2016-2020) years of economic recession (IMF, 2023b). The oil sector has been the engine of
economic growth accounting for about 29% of GDP and about 95% of exports in 2022 (BNA,
2023). Government accounts is also dependent heavily on oil sector. For instance, non-oil fiscal
deficit in percentage of GDP, is estimated in -7.7% in 2022, while in 2014 it was -25.4%. (MINFIN,
2022).
From the monetary side, specifically inflation, it is valuable to explain that historically,
Angola economy has faced periods of great problem of price level instability. Inflation has
fluctuating, mainly around a double digit throughout recent Angolan economic history. The
exception is only for the years 2012, 2013 and 2014 in which inflation rate was 9%, 7.7% and 7.5%,
respectively, as illustrated in Fig 1, panel a (IMF, 2023b). High volatility is one of the issues of
Angolan inflation over the years. As shown in the Fig. 1 panel (a), while from 2012 to 2014 the
economy faced a relative price stability, from 2015 to 2022 it was profoundly volatile wherein
reached a peak of 41% in 2016.
Figure 1. Money supply, GDP, and inflation (year-on-year, in percentage)
𝑃𝑎𝑛𝑒𝑙 (𝑎): 𝑀𝑜𝑛𝑒𝑦 𝑔𝑟𝑜𝑤𝑡ℎ 𝑣𝑠. 𝐼𝑛𝑓𝑙𝑎𝑡𝑖𝑜𝑛
50%
40%
50%
40%
30%
30%
20%
20%
10%
10%
0%
0%
- 10%
- 10%
- 20%
- 20%
𝑃𝑎𝑛𝑒𝑙 (𝑏): 𝑅𝑒𝑎𝑙 𝐺𝐷𝑃 𝑣𝑠. 𝑅𝑒𝑎𝑙 𝑚𝑜𝑛𝑒𝑦 𝑔𝑟𝑜𝑤𝑡ℎ
- 30%
- 30%
2012
2014
2016
Inflation
2018
2020
2022
2012
2014
2016
2018
Real money growth
Money growth (M1)
2020
2022
Real GDP
(Source: BNA, 2023; IMF, 2023b)
In terms of economic context, a relative price stability between 2012 and 2014 was due to a
favourable oil sector (and responsible for nearly 95% of Angolan exports) which permitted a
monetary policy anchored on foreign reserves to control inflation through exchanges rates
44
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
manipulation4. Conversely, since 2015, the economy has performed poorly, Angolan oil market fell
considerably due to continuing decline in oil production and constant negative shocks in oil prices.
For a country heavily dependent on oil sector to gain foreign reserves, exchanges rates pressures
take place and directly affect price level. Monetary authority starts to liberate exchanges rates
markets in 2018 and reached a full flexible exchanges rate in 2020, and inflation responded
positively dropping to nearly 14% in 2022. Panel (a) also shows that money growth (measure as
year-to-year banknotes and coins held by the public and transferable deposits) was highly volatile in
the period.
There are three major trends in money growth in Angola. Firstly, is observed a significant yearto-year growth from nearly 5% in 2012 to 20% in 2014. Secondly, it fell profoundly until reach a
minus growth of about 2% in 2017. Third, it increased consistently, reaching a peak of nearly 29%
in 2021. Before 2015, monetary authority in Angola was less cautious in controlling money supply
in the economy once to control inflation was much more effective using exchanges to control price
level burning foreign reserves. After that, BNA starts to deliver a tight monetary policy. Despite
some exceptions fig. 1 tells us that there might exist a positive correlation between money growth
and inflation in Angola, which is in line with hypothesis 1. This assumption is mainly true during
the periods of less volatile inflation rate.
Considering that money growth was highly volatile between 2012 and 2022, a real money
growth using national consumer price index to convert it from nominal to real (taking 2012 as basic
year) shows that real money growth in 2013, 2014, 2019 and 2020 was higher than real GDP, as
observed in Fig. 1 panel (b), which might have contributed to accelerate inflation rate, particularly
in 2019 and 2020.
Search Results
To assess the impact of money growth on inflation rate in Angola this studiey use a Vector
Autoregressive Model (VARM). A VARM is used when there is no confidence that a variable
really is exogenous, each variable has to be treated symmetrically (Asteriou and Hall, 2011). In
other words, some variables are not only explanatory variables for a given dependent variable, but
are also explained by the variables that they are used to determine5.
As pointed out by Asteriou and Hall (2011), the VARM approach has some very insightful
characteristics. Firstly, it is very simple, once an econometrician does not have to worry about
which variables are endogenous or exogenous. Secondly, estimation is also very simple, in the
sense that each equation can be estimated separately with the usual ordinary least squares method.
Finally, forecasts obtained from VARM are in most situations better than those obtained from the
far more complex simultaneous equation models (Mahmoud, 1984; McNees, 1986).
The VAR model was implemented with two key variables has been imposed. The variables
are national consumer price index and money supply (M2). To study the evolution of relationship
4
Considering that Angolan imports its main goods and services, it was able to control inflation supplying foreign
currencies to artificially appreciate local currency and consequently deliver a stable price level (lower inflation).
For instance, the imagine the time series 𝑦𝑡 that is affected by current and past values of 𝑥𝑡 and, simultaneously, the time
series 𝑥𝑡 to be a series that is affected by current and past values of the 𝑦𝑡 series. Thus, VARM is given by:
5
𝑦𝑡 = 𝛽10 + 𝛽12 𝑥𝑡 + 𝛿11 𝑦𝑡−1 + 𝛿12 𝑥𝑡−1 + 𝜇𝒚𝒕
𝑥𝑡 = 𝛽20 + 𝛽21 𝑦𝑡 + 𝛿21 𝑦𝑡−1 + 𝛿22 𝑥𝑡−1 + 𝜇𝒙𝒕
45
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
between money growth and inflation rate over time and/or assess whether money growth has
affected differently inflation rate through time we use time-varying estimates of our VAR. This is
justified by the presence of structural breaks on the data, which might affect the intensity of money
growth effect on inflation.
The general form of our VAR can be illustrated by the following expression:
∆𝑧𝑡 = 𝑐0 + ∑𝑛−1
𝑖=1 𝛤i 𝑧𝑡−i + 𝑢𝑡
(1)
Where 𝑧𝑡 denotes a vector of variables. In our case, 𝑧𝑡 = (𝐶𝑃𝐼𝑡 , 𝑀𝑡 ), where 𝐶𝑃𝐼𝑡 is (log)
national consumer price index and 𝑀𝑡 is (log) money supply M2. Further detail such as data
definitions and sources are provided below in Table 1. 𝑐0 is a vector of constant terms. 𝑛 denotes
the VAR lag-order. The short-run dynamics of the model are captured by the matrix of
coefficients 𝛤i , this includes the coefficients for 𝐶𝑃𝐼𝑡 and 𝑀𝑡 . 𝑢𝑡 is a vector of error terms, Normally
and Independently Distributed (NID).
The empirical strategy proceeds in four steps. The first two are aimed at understanding the
properties of our data and model it appropriately. First, a test for unit roots in 𝑧𝑡 using the standard
unit-root stationary test (Augmented Dickey-Fuller and Phillips-Perron) and KPSS6 test. Due to the
presence of large shocks in the sample a unit-root test with breaks Zivot and Andrews (1992) is
used. We test if cointegration vectors exist among 𝑧𝑡 variables by testing for the rank (𝑟) of the
matrix 𝛱 using the Engle-Granger approach.
Second, we estimate a VAR model focusing on 𝐶𝑃𝐼𝑡 equation for the full-sample and use
our estimates of 𝛤i to assess the effects of money growth on inflation rate in Angola. The
coefficients for the lags of 𝑀𝑡 contained in 𝛤i , provides the elasticities of 𝐶𝑃𝐼𝑡 to money growth.
Considering the underlying economic theory, we expect a positive relationship between 𝐶𝑃𝐼𝑡 and
𝑀𝑡 . Three, to examine how the effects of money growth on inflation has evolved over time we reestimate 𝐶𝑃𝐼𝑡 equation using rolling windows. In our case, after experimenting with several
window-sizes, we find that a window of 12 observations (quarterly data), that provides 24 rolling
estimations or subsamples, is statistically satisfactory and economically meaningful. And lastly, we
complement this evidence with estimates of impulse response functions of inflation for innovations
on money supply.
Table 2 provides definitions and sources for each variable in 𝑧𝑡 . Data expands over the
period 2014:Q4-2023:Q2. This covers the post-global 2007/2008 financial crisis and the era of
significative monetary policy reforms in Angola. It is well suited to study money growth and
inflation rate as it contains periods of fixed exchange rates regime, as well as flexible exchange rate
after.
Fig. 2, presents the evolution of variables in 𝑧𝑡 over the sample period. As we can see, in the
panel (a), 𝐶𝑃𝐼𝑡 follows an upward trend, but from 2016 it accelerates considerably, reflecting the
effect of lower capability to deliver discretionary monetary policy by National Bank of Angola to
stabilize prices through exchange rates manipulation before the implementation of a fixed
exchanges rates regime. Angola depends entirely on oil sector to accumulate foreign reserves and it
imports majority of all essentials goods and services needed thereafter, constant oil prices shocks
and reduction in oil production since late 2014 imposed a huge restriction on country’s capability to
accumulate foreign reserves to support external demand. Thus, with less foreign reserve, monetary
policy is less effective and need to be restrictive. Panel (b), 𝑀𝑡 , exhibits an upward-trend that fell
6
We run, KPSS, firstly, to cross-check results from ADF, PP tests, which have a unit-root null hypothesis, against those of
a test where the null is of stationarity (Kwiatkowski, et al., 1992). Secondly, it appears that, 𝐶𝑃𝐼𝑡 might not be normally
distributed. As noted by Hadri (2000), KPSS test is not affected by the non-normality of variables, as in our case. Hence, it
seems appropriate to use this test.
46
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
profoundly from mid of 2017, reflecting a week economic performance in that period in Angola.
And in 2022 it turns down sharply in consequence of National Bank of Angola response to counter
inflation rate. We seasonally adjusted 𝐶𝑃𝐼𝑡 and 𝑀𝑡 to deal with monetary and seasonal effects using
X-13-ARIMA, as proposed by Balcilar et al. (2013).
Figure 2. Evolution of 𝐶𝑃𝐼𝑡 and M𝑡 2014: q4 - 2023: q2
𝑃𝑎𝑛𝑒𝑙 (𝑎): 𝐶𝑃𝐼𝑡
6.4
𝑃𝑎𝑛𝑒𝑙 (𝑏): 𝑀𝑡
9.6
9.4
6.0
9.2
5.6
9.0
5.2
8.8
4.8
8.6
4.4
2023:q1
2022:q1
2021:q1
2020:q1
2019:q1
2018:q1
2017:q1
2016:q1
2015:q1
2023:q1
2022:q1
2021:q1
2020:q1
2019:q1
2018:q1
2017:q1
2016:q1
2015:q1
8.4
(Source: Banco Nacional de Angola, 2023)
Table 1. Data
Variable
𝐶𝑃𝐼𝑡
𝑀𝑡
Description
National consumer price index (in
logs).
Money supply M2 (in logs).
Data period
Source
[1]
Quarterly
[2]
[1] IMF (2023a). [2] BNA (2023)
We use M2 (which is the entire stock of currency and other liquid instruments, that includes
cash, coins and balanced held in checking and saving accounts) as measured to money supply
growth because it contains more than just physical money and it is prominent instrument to measure
money growth, and it is frequently used to evaluate liquidity.
Unit-roots
We start our analysis by testing the unit-root properties of our variables, 𝑧𝑡 = (𝐶𝑃𝐼𝑡 , 𝑀𝑡 )′.
Table 2 presents the results of tests without breaks, i.e., ADF, PP and KPSS tests. Overall,
comparing the results for ADF and PP suggests some contradictions once it is not possible to
assume stationarity and/or non-stationarity jointly in each one of the tests. Thus, we run a KPSS
which is much more robust than ADF and PP. KPSS tests rejects stationarity at level and concludes
that 𝐶𝑃𝐼𝑡 and 𝑀𝑡 are stationary when imposed I(1).
Zivot-Andrews’s test, reported in Table 3, does not reject the null of a unit-root on the first
differences for all three specifications of the test. Further, Zivot-Andrews’s test identifies significant
structural breaks for 𝐶𝑃𝐼𝑡 in 2016q1 coinciding with massive currency depreciation and
consequently an inflation peak in Angola. In sum, evidence from Table 2 suggests that we should
treat variables in 𝑧𝑡 as 𝐼(1) and that there are breaks in the data, which justify our time-varying
VAR approach.
47
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
Table 2. Unit-root tests
Variables
Level
𝐶𝑃𝐼𝑡
𝑀𝑡
1st diff.
∆𝐶𝑃𝐼𝑡
∆𝑀𝑡
Const.
ADF
Const. &
trend
Phillips-Perron test
No
Const. &
Const.
trend
Const.
KPSS test
Const. & trend
1.48
0.13
-2.17
-3.83*
7.99**
-2.65*
-1.07
-2.00
1.39**
1.34**
0.16*
0.13
-2.07
-2.79
-2.18
-2.46
-0.61
-2.29*
-2.75
-2.91
0.16
0.07
0.07
0.07
Note: The number of lags selected by AIC are 8 𝐶𝑃𝐼𝑡 , 5 𝑀𝑡 and 1 𝐼𝑅𝑡 . ADF critical values at 1% and 5% with intercept
are: [-3.74/ -2.99], and with intercept and trend: [-4.37/ -3.60]. Phillips-Perron critical values at 1% and 5% with intercept are: [-2.65/
-1.95]. KPSS test critical values at 1% and 5% with intercept are: [0.74/ 0.46], and with intercept and trend: [0.22/ 0.15].
* Rejection of the null hypothesis at the 5% level/ ** Rejection of the null hypothesis at the 1% level.
Table 3. Unit-root tests with breaks, Zivot-Andrews’s test
Variables
Zivot-Andrews’s test
Model A,
with break in const.
Model B,
with break in trend
Break
Break
t-statistic
t-statistic
Model C,
with break in
const. and trend
Break
t-statistic
Levels
𝐶𝑃𝐼𝑡
𝑀𝑡
2016q1
2019q3
-5.64**
-4.28*
2015q2
2016q1
-4.83*
-3.45
2016q1
2019q3
-5.48*
-4.17
∆𝐶𝑃𝐼𝑡
∆𝑀𝑡
2015q4
2021q1
-4.40
-3.36
2016q1
2022q4
-4.40
-3.30
2015q4
2021q3
-4.40
-3.81
1st Diff.
Note: Critical values for A at 1% and 5% are: [-5.34/ -4.80]. For B at 1% and 5%: [-4.93/ -4.42]. For C at 1% and 5%: [5.57/ -5.08]. * Denotes rejection of null hypothesis at 5% level and ** at 1%.
Next, we test if our variables are cointegrated using Engle-Granger cointegration approach.
Table 4 reports our results for the full sample. The result indicates that there is not long-run
relationship among our key variables at the 5% between CPIt and Mt. In sum, considering the
results from cointegration analysis, we should employ a VAR model in first differences to assess
the effect of money growth on inflation rate in Angola to avoid a spurious regression.
Table 4. Engle-Granger Cointegration test between CPIt and Mt
z(t)
Test
statistic
Critical value
1%
5%
10%
-3.16
-4.75
-4.01
-3.64
Money growth impact on inflation rate
Next, we evaluate the effect of money growth on inflation rate in Angola for the whole
sample in our selected model and focusing the analyses on ∆CPIt equation. Following the
appropriate standard selection criteria (AIC/SBC), it delivers n=1 for our VAR model, see
Appendix A. Conversely, to resolve autocorrelation problems, we adopt a VAR model with two
lags (𝑛=2). Table 5 presents our results. Our regression for ∆𝐶𝑃𝐼𝑡 , explains a large proportion of the
48
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
variation in the data, and the adjusted R-squares (𝑅̅ 2 ) is 67.3%. For ∆𝑀𝑡 , the fitted values explain
32.9% in this variable. All regressions pass the corresponding serial correlation, homoscedasticity,
and fell slightly the normality diagnostic tests, at the 5% significance level. Further, Eigenvalue
stability condition is satisfied suggesting that the model is stable and satisfactorily specified, see
Appendix B.
Table 5. VAR estimations
∆𝐶𝑃𝐼𝑡−1
Equation (1): ∆𝐶𝑃𝐼𝑡
Equation (2): ∆𝑀𝑡
0.968***
(6.30)
0.159
(0.26)
∆𝐶𝑃𝐼𝑡−2
-0.313**
(-2.06)
0.003
(0.06)
0.080
(0.42)
∆𝑀𝑡−2
0.108**
(2.31)
0.265
(1.42)
𝐶0
0.014***
(2.61)
0.094***
(4.26)
𝜎̂
0.011
[0.000]
∆𝑀𝑡−1
𝑅2
𝑅̅2
P-value
-1.626***
(-2.69)
0.715
0.673
0.415
0.329
0.043
[0.000]
Observations=32 in all regressions. ( ) stands for t-statistics. * Indicates significance at 10%, ** at 5% and *** at 1%. 𝜎̂
Stands for the standard error of the residuals.
In equation (1), we find that lagged inflation rates do affect current inflation in Angola as the
corresponding elasticities at lag 1 and 2 of ∆𝐶𝑃𝐼𝑡 are significant at the 5% test. This result is
expected and in line with the inertial inflation principle. The results also indicates that the
coefficients of ∆𝑀𝑡 on ∆𝐶𝑃𝐼𝑡 equation have the correct sign, implying that the greater the money
supply the greater the inflation. Note that current inflation is not only affected by its previous levels
but that lagged money supply movements have significative impact on current inflation rate in
Angola. At lag 2 (∆𝑀𝑡−2 ), a 1% increase in money supply leads inflation to rise by 0.11%. At lag 1
(∆𝑀𝑡−1 ), the underlying coefficient (0.003%) on ∆𝐶𝑃𝐼𝑡 is not significantly different from zero.
To sum up, the result from VAR indicates firstly, that money supply does affect inflation in
Angola and its impacts is positive. Secondly, there is inertial inflation once past inflation does affect
current inflation rate. And finally, money supply and past inflation may explain nearly 67% of
current inflation. These results are in line with what was found by Roshan (2014), and Papadia and
Cadamuro (2021) wherein countries whose inflation is volatile, money supply does help to explain
inflation, and considering that inflation in Angola is highly volatile, the evidence suggests that
money supply does affect inflation in Angola. Thus, hypothesis 1 - money growth does affect
inflation rate in Angola is not rejected
Time-varying analysis: money growth impact on inflation rate
Following our results, we can argue that money growth does affect inflation rate in Angola.
Additionally, it is important to assess how money growth has affected inflation rate throughout the
period and to do so, Fig 3 below presents the time-varying coefficient of money growth on inflation
using rolling windows in our VAR model. Panel (a) and (b) in Fig. 2 shows that time varying
elasticity of ∆𝐶𝑃𝐼𝑡 with respect to ∆𝑀𝑡−𝑖 have changed dramatically over time and its value per
period is mostly positive (excepting at the end of 2018 and beginning of 2019, and end of 2021),
confirming the hypothesis 3.
49
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
Figure 3. Rolling elasticity of ∆𝐶𝑃𝐼𝑡 with respect to lags of ∆𝑀𝑡 .
𝑃𝑎𝑛𝑒𝑙 (𝑎): ∆𝑀𝑡−1 𝑜𝑛 ∆𝐶𝑃𝐼𝑡
.3
𝑃𝑎𝑛𝑒𝑙 (𝑏): ∆𝑀𝑡−2 𝑜𝑛 ∆𝐶𝑃𝐼𝑡
.5
.2
.4
.1
.3
.0
.2
-.1
.1
-.2
.0
-.3
-.1
-.4
-.2
2017
2018
2019
2020
2021
2022
Rolling
2023
2017
Full sam ple
2018
2019
2020
2021
2022
2023
CI 95%
Note: Windows size=12 obs. (24 subsamples) dates in x-axis denote last quarter of estimation window.
Observing Fig 3 panel (b), whose full sample elasticity of ∆𝐶𝑃𝐼𝑡 with respect to ∆𝑀𝑡−2 is
significantly different from zero it indicates that there are three different patterns for money growth
effects on inflation. Firstly, from 2014 to 2019 the elasticity of ∆𝐶𝑃𝐼𝑡 with respect to ∆𝑀𝑡−2 shows
a declining trend. Secondly, from 2019 to 2021 shows a stable trend, and finally, after the last
quarter of 2022 indicates an increasing trend. There could be several reasons behind such behavior,
but it is certain that the following ones may help to understand what happened in each period. The
reason for declining trend before 2019 may be linked to the use of foreign reserves to guarantee
artificially local currency value and consequently prices levels stability once majority of essential
goods and services consumed in Angola are imported. Thus, in that period, the exchange rates
position played a key role in controlling inflation. The stable trend between 2019 and 2021 can be
explained by a tight monetary policy to control inflation combined with introduction of flexible
exchanges rates regime in Angola, whereas an increasing trend from 2022 might be linked to
government expenditures growth (general election in 2022) and oil prices increment.
Impulse response function
This section presents the impulse response function of inflation for innovations on money
supply. As illustrated in Fig. 4 panel (b), a standard deviation shock on money supply does not
affect inflation at the beginner, but in the following period the inflation rises profoundly during two
quarters and then starts to decelerate slightly throughout the period. However, it appears that the
effects of money supply shocks on inflation does disappear completely after sixteen quarters.
Figure 4. Impulse response function of ∆𝐶𝑃𝐼𝑡 with respect to shocks in ∆𝐶𝑃𝐼𝑡 and ∆𝑀𝑡 .
±
.016
𝑃𝑎𝑛𝑒𝑙 (𝑎): 𝑅𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑜𝑓 ∆𝐶𝑃𝐼𝑡 𝑡𝑜 ∆𝐶𝑃𝐼𝑡
𝑃𝑎𝑛𝑒𝑙 (𝑏): 𝑅𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑜𝑓 ∆𝐶𝑃𝐼 𝑡𝑜 𝑀𝑡
.016
.012
.012
.008
.008
.004
.004
.000
.000
-.004
-.004
-.008
-.008
-.012
-.012
2
4
6
8
10
12
14
16
18
20
2
4
6
8
10
12
14
16
18
20
The results from impulse response functions tend to support the view that, money supply
shocks in Angola does affect systematically inflation rate and it may last for at least 4 years if a
50
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
discretionary monetary policy is not introduced to control its spreads effect throughout the period.
Conclusion
The purpose of this study was firstly to assess the effect of money supply on inflation in
Angola, secondly, to examine how the effects of money growth on inflation rate has evolved over
time. To do so, a vector autoregressive model was used, after assessing the underlying mythological
conditions.
The findings suggest that money supply does affect inflation in Angola and its shocks may
last a certain period. And finally, the time-varying elasticity of inflation rate with respect to money
supply has changed dramatically over time and its value per period is mostly positive. These
findings have considerable implications for the role of monetary policy in the short term to control
inflation rate in Angola once it reinforces the views to BNA deliver a discretionary monetary policy
during the periods of increase in government activity.
Moreover, it is required, in the future, further analyses about the effect of money supply on
inflation under a set of data much more robust in order to take into consideration the role of oil
market sector on inflation through exchange rates.
References
Asteriou, D. & Hall, S. 2011. Applied Econometrics. 2nd Edition, Palgrave Macmillan.
Balcilar, M., Gupta, R. & Miller, S. M. (2013). Regime switching model of US crude oil and
stock market prices: 1859 to 2013. Energy Economics. 49(2015), 317-327.
Banco
Nacional
de
Angola.
(2023).
Consulta
https://www.bna.ao/#/pt/estatisticas/consultar-dados
de
dados.
[Online].
Bozkurt, C. (2014). Money, inflation, and growth relationship: the Turkish case.
International Journal of Economics and Financial Issues, 4(2), 309-322.
Borio, C., Hofmann, B. & Zakrajsek, E. (2023). Does money growth help explain the recent
inflation surge? BIS Bulletin (67)
Evans, O. (2019). Money, Inflation and Output in Nigeria and South Africa: Could
Friedman and Schwartz Be Right? Journal of African Business, 20(3).
Fontana, G. & Setterfield, M. (2009). Macroeconomics, endogenous money and the
contemporary financial crisis: a teaching model. International Journal of Pluralism and Economics
Education, 1(2), 30 - 147. https://dx.doi.org/10.1504/IJPEE.2009.028970
International Monetary Fund. (2023a). Angola IMF office in Angola. Latest High Frequency
Macroeconomic Data. [Online]. https://www.imf.org/en/Countries/ResRep/AGO.
International Monetary Fund. (2023b). World economic outlook database. [Online]
Government of Angola, Ministério das Finanças. (2022). Relatório de Fundamentação do
OGE 2023.
Mahmoud, E. (1984). Accuracy in forecasting: a survey. Journal of Forecasting, (3), 13959.
Mbongo, J. E., Mutasa, F. & Msigwa, R. E. (2014) The Effects of Money Supply on
Inflation in Tanzania. Economics. 3(2), 19-26. https://dx.doi.org/10.11648/j.eco.20140302.11
McNees, S. 1986. Forecasting accuracy of alternativa techniques: a comparison of US
macroeconomics forecasts. Journal of Business and Economic Statistics, (4), 5-15.
51
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
Mpofu, R. T. (2011). Money supply, interest rate, exchange rate and oil price influence on
inflation in south Africa. Corporate Ownership & Control 8(3), Continued – 6.
Papadia, F., & Cadamuro, L. (2021). Does money growth tell us anything about inflation?
Working Paper (11), Bruegel.
Roshan, S. A. (2014). Inflation and money supply growth in Iran: empirical evidence from
cointegration and causality. Iran Econ. Rev. 18(1).
Roffia, B., & Zaghini, A. (2007). Excess money growth and inflation dynamics. Working
Paper Series (749), European Central Bank.
Sieron, A. (2019). Endogenous versus exogenous money: does the debate really matter?
Research in Economics, 73(4), 29-338. https://dx.doi.org/10.1016/j.rie.2019.10.003 1090-9443
Sainani KL. (2012) Dealing with non-normal data. American Academy of Physical Medicine
and Rehabilitation. 4(12), 1001-1005. https://dx.doi.org/10.1016/j.pmrj.2012.10.013
Thwaini, F. H. & Hamdam, A. A. (2017). Money supply. Endogenous or exogenous variable
with reference to Iraq. Banks and Bank Systems, 12(4) https://dx.doi.org/10.21511/bbs.12(41).2017.03
Tobin. (1970). Money and income: Post Hoc Ergo Propter Hoc? Quarterly Journal of
economics, (84).
Zivot, E., Andrews, D., (1992). Further evidence on the great crash, the oil-price shock, and
the unit-root hypothesis. Journal of Business Economic Statistics (10), 251–270.
https://doi.org/10.2307/1391541 .
Como citar: Antonio, A. E. da C. (2024). Money growth and inflation rate: evidence from Angola from 2014 to
2023. Academicus Magazine, 2(2), 42–54. DOI: https://dx.doi.org/10.4314/academicus.v2i2.3
Appendix A. Selection-order
Table A1. Lag-order selection criteria
Lag
1
2
3
4
5
6
7
8
9
LR
AIC
SBIC
43.169
7.7816
.32127
2.3809
5.4582
5.3195
.88769
11.056
28.557*
-10.3722*
-10.3634
-10.0563
-9.83151
-9.72983
-9.62261
-9.33812
-9.46035
-10.2826
-10.0796*
-9.87587
-9.3737
-8.95392
-8.65722
-8.35498
-7.87547
-7.80268
-8.42993
Appendix B: VAR Post-estimation tests
∆𝐶𝑃𝐼𝑡 fails the normality test. This persisted despite experimenting with several alternatives, e.g., using more lags and
correcting outliers with dummies. Hence, given that this issue invalidates the t-test statistics but has no impact on the efficiency of
estimates, we decided to proceed acknowledging that we cannot use the t-test to validate results when observation is larger.
52
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
Table B1. Lagrange-multiplier test for serial correlation
Lag
1
2
Chi2
3.5349
1.6960
df.
4
4
Prob. > chi2
0.47249
0.79144
We conclude that residuals are not serially correlated.
Test for normally distributed disturbances, reported in Table B2. The null hypothesis is that disturbances are normally
distributed. When p-value higher than 5% we do not reject null and rejected otherwise
Table B2. Test for Normality of residuals
Equation
∆𝐶𝑃𝐼𝑡
∆𝑀𝑡
All
Equation
∆𝐶𝑃𝐼𝑡
∆𝑀𝑡
All
Equation
∆𝐶𝑃𝐼𝑡
∆𝑀𝑡
All
Jarque-Bera test
Chi2
df.
Prob. > chi2
19.621
6.334
25.954
0.00005
0.04214
0.00003
2
2
4
Skewness test
Skewness
Chi2
1.225
8.003
0.87183
4.054
12.057
Kurtosis test
Kurtosis
Chi2
5.9518
11.617
4.3076
2.280
13.897
df.
1
1
2
df.
1
1
2
Prob. > chi2
0.00467
0.04407
0.00241
Prob. > chi2
0.00065
0.13107
0.00096
Table B3. Homoscedasticity of residuals
Sample: 35
Included observations: 32
Joint test:
Chi-sq
df
Prob.
26.6477
24
0.3211
Individual components:
Dependent
R-squared
F(135,23)
Prob.
Chi-sq(135)
Prob.
res1*res1
res2*res2
res2*res1
0.436033
0.154649
0.264701
2.222819
0.525954
1.034975
0.0641
0.8248
0.4392
13.95307
4.948764
8.470438
0.0830
0.7630
0.3889
White heteroscedasticity test indicates that we cannot reject the null hypothesis i.e., we have reasons to believe that the
data is homoscedastic once the p-value for the joint test are higher at the 5% test level.
53
Versão Online - ISSN: 3005-3633
Vol. 2, N˚2, pp.42-54, 2024
http://www.revista.academicuspro.ao
Figure B1. Eigenvalue Stability Condition Test
-1
-.5
0
.5
1
Roots of the companion matrix
-1
-.5
0
Real
.5
1
Eigenvalue Stability condition test yields that all roots lie inside the unit circle, thus the VAR system is stable.
54