thesis degree bachelor major actuary topic the determinants and prediction of cpi in vietnam using arima and ecm models

There are two primary aims of this study: to forecast CPI values using the Autoregressive Integrated Moving Average Model and analyse the relationship between the gasoline price, money s

NATIONAL ECONOMICS UNIVERSITY FACULTY OF MATHEMATICAL ECONOMICS

THESIS

DEGREE: BACHELOR MAJOR: ACTUARY

Topic: The Determinants and Prediction of CPI

in Vietnam using ARIMA and ECM models

Student : Pham Minh Tam

Student’s ID : 11184358

Class : Actuary 60

Supervisor : Ph.D Nguyen Manh The

Hanoi, 06/2022

National Economics University Faculty of Mathematical Economics

DECLARATION OF AUTHORSHIP

I] herewith formally declare that I have written the submitted thesis independently | did not use any outside support except for the quoted literature and other sources mentioned at the end of this paper I clearly marked and separately listed all the literature and all other sources which I employed producing this academic work, either literally or in content

Hanoi, June 2022 Pham Minh Tam

ACKNOWLEDGEMENT

1 am grateful for the valuable guidance of my supervisor, Ph.D Nguyen Manh The He has given me worthy advice and support with full encouragement and enthusiasm to help me complete my work

1 am fortunate to be a student of Faculty of Mathematical Economics (MFE), National Economics University I] am grateful to all of my teachers for continuously motivating

me with their wealth of knowledge during my studying time

1 would like to send my sincerest thanks to all the members of the Institute of Economics and Finance They have provided motivation and chances for me in professional development Without their help, this research would not have been achieved

Last but not least, 1 want to thank my family and friends for always being there for me

as well as giving me continuous support

Thank you all for your support and encouragement

Hanoi, June 2022 Pham Minh Tam

ABSTRACT

This study provides the forecasting method for CPI and examines the determinants of inflation in Vietnam There are two primary aims of this study: to forecast CPI values using the Autoregressive Integrated Moving Average Model and analyse the relationship between the gasoline price, money supply and inflation using the Error Correction Model The research data in this thesis is collected from three main sources: the General Statistics Office releases, the database of the State Bank of Vietnam and the Institute of Economics and Finance The time interval of the monthly dataset is from January 2010 to February 2022, while the annual CPI data in the 1998-

2021 period is collected With an acceptable RMSE, the Seasonal ARIMA model gives the most accurate prediction of monthly CPI in Vietnam Based on these forecast values, the rise and fall of CPI will be forecasted This study reveals that the change in the current money supply will produce a negative effect on CPI values in the second month ahead Gasoline price also has an influence on inflation, which is a multi-stage impact The results of the causality test show that the money supply M2 Granger causes inflation Moreover, the CPI in Vietnam needs more than a month (adjusting about 88% per month) for the short-term adjustments to reach the long-term equilibrrum Taken together, these findings provide some suggestions not only for controlling inflation uncertainty risks but also for policy amendments

Keywords: Consumer Price Index, gasoline price, money supply, inflation forecasting, ARIMA, ECM

TABLE OF CONTENT

1.2.2 Relationship of macroeconomic indicators and inflation 19

CHAPTER 2 CONTEXTUAL ANALYSIS 22

LIST OF FIGURES

Figure 1 Annual CPls in Vietnam from 1998 to 202 Ì - 22c 2 222cc ssss 22 Figure 2 Money supply M2 growth in Vietnam from January 2010 to February 2022 G1111 111 111111111 1111 11 T111 11 111 111111111111 11111 111111111110 111111101101 10 11111611101 1111111110171: 26 Figure 3 Vietnam’s RON-95 price ffom January 2010 to February 2022 27 Figure 4 CPI and gasoline price growth from January 2018 to February 2022 29 Figure 5 CPI and money supply growth from January 2018 to February 2022 30 Figure 6 Monthly CPls in Vietnam from January 2010 to February 2022 33 Fieure 7 PACF plot of the second differenced data - 2c c2 ccccs2 42 Figure 8 ACF plot of the second differeneed data - 0 2c 2222 szcses 43 Figure 9 ACF plot of the first differenced data - 2c 2222222222 44 Fieure I0 PACF plot of the đrst differenced data 2c 2c 2222222 45 Figure 11 Forecast plot by using seasonal model in train set - - 45 Figure 12 Forecast plot from [[I[[] 0(013)(200 (12) once 2g 46 Figure 13 CPI and M2 growth with a two-month là ccc c2 sex 49 Fieure I4 Weiehts for CPI compilation from 2006 to 2025 ĩc 60 Figure 15 Time series plot of the second differenced annual CPI data 61 Fieure I6 Forecast plot ffom ARIMA(1,2,) - c2: 2221132113211 1521 rse 62 Fieure L7 Time serles plot of the imitial monthÌy data - 525cc 5+2 65 Fieure I8 Time serles plot of the first differenced monthly data 65

LIST OF TABLES

Table 1 Variables deserIptIi0n 5 2 2222111211 11211 1121112211111 1101112211811 31 Table 2 Descriptive statistlcs for monthly dataset 5222522225552 32 Table 3 Foreeast values by using[[[ILILII21] - -¿- 5:2 5:22+52222<zx+2cxs<s2 43 Table 4 Foreeast values from [[JJ [I01,3(200112Ị - -:¿- 5:52 5522 c5>ss3 46 Table 5 Cointegration test result using trace fest S2: cà cv 2222222 47 Table 6 Cointegration test result using eleenvalue test - c5 5225-52 48 Table 7 Optimal lag selection reSuÏIt - 5 5c 2 2211221112111 1221 1111152211111 2⁄2 48 Table 8 ECM estimation result e 48 Table 9 Granger test result 50 Table 10 Weiehts for CPI compilation for the period 2021-2025 - 59 Table L1 Estimated coefficients of[][[I[I[Í1/21) - - ¿5:2 52+25<5zx+2<zx+ss2 61 Table 12 Accuracy performance of [[[[J]f2 1) - -: +5:+52+5zx2szszzsss2 61 Table 13 Coefieients of[[LLILIÍ100) with non-zero mean - 2222 sez s2 s22 63 Table 14 Accuracy performanee øf [[[[[ILũŨ) ¿522222 2£2£2£zEz£z£zzzzzzzs+z 63 Table 15 Forecast values from [[[[[I1,0} - - 5-5: ¿522 25+2222<2*2x+ex+zzzzs+2 64 Table 16 Coeffictlents of ARIMA(Œ,Ì,2) 2 2221222111211 1 1211152111211 1 1122 tk 65 Table 17 Accuracy performance of ARIMA(2,1,2) c c2 222222 66 Table 18 Forecast values from ARIMA(2,,2) - 2 1 2222221122221 2222 66 Table 19 Coeffieients of[[JI[[I(01 3J200)/12| -¿ 5-5552 5< sc<<cczxz< 67 Table 20 Accuracy performance of [[[T[Jf.13)200112] -:- ¿5-5 +5: 67 Table 21 ECM estimation result 69 Table 22 VECM estimatfion reSUÏK - 221121121121 1111 1111111101 111 11 112g hưu 70 Table 23 Forecast values from VECM - - Án HT TS 1 HH1 H11 kết 71

Augmented Dickey-Fuller test

Akaike Information Criterion

Autoregressive Conditional Heteroskedasticity

Autoregressive Integrated Moving Average

Bayesian Information Criterion

Consumer Price Index

Error Correction Model

Foreign Direct Investment

Generalized Autoregressive Conditional Heteroskedasticity Gross Domestic Product

Gasoline (RON-95) Price

General Statistics Office

Harmonized Index of Consumer Prices

Hannan—Quinn Information Criterion

Kwiatkowski-Phillips-Schmidt-Shin test

Long Short-term Memory

Mean Absolute Error

Mean Absolute Percentage Error

Mean Absolute Square Error

National Income

Neural Network Auto Regressive

Official Development Assistance

Partial Autocorrelation Function

Producer Price Index

Seasonal Autoregressive Integrated Moving Average

Inflation is a major subject that affects both economic and social life CP] measures the average changes in prices that consumers pay for a basket of goods to use in their daily life It 1s also the measurement to determine inflation Furthermore, knowing about inflation can assist policymakers in making a better decision given the current scenario Citizens do not feel uncertain about future inflations and trust the government’s guidelines and directions In the situation that salaries cannot keep up with inflation, for instance, people face up to poverty and instability, especially when the inflation may be up to double digits It also leads to the possibility of excessive increasing prices, which raise high prices of common goods and make households’ life harder

Inflation forecasting is a common topic of forecasting economic indicators Many forecasts used numerous models in different countries, both developed and developing countries Due to the change in the CPI calculation method, the forecasting model for CPI values in Vietnam needs to be modified to be suitable for each period In the past, predicting the price level was based on analysis from Central Institute for Economic Management (CIEM) or General Statistics Office (GSO) With the introduction of univariate models, time series models, such as ARIMA, are becoming the most popular method of forecasting inflation However, both the selection of the appropriate predicting model and the model accuracy for long forecast periods are generally open to discussion

It is clear that changes in the economy including management policies or price adjustments will influence the price level In the past, several studies on this topic have been conducted for other developed economies, such as the European Union (EU) or the United States (US) However, little research on this topic has been found for developing economies, particularly Vietnam's economy

Recently, some researchers have done an analysis of the link between economic variables and inflation in Vietnam For example, they measured the weight of gasoline price in CPI calculation and the composition of other industries and then described the relationship between those variables In addition, much of the research up to now has been descriptive of the impacts of macroeconomic indicators on inflation It is known that the change in money supply causes fluctuations in CPI However, there are limited studies that investigate the relationship between these variables by using econometric models

For all of these reasons, the topic “The Determinants and Prediction of CPI in Vietnam using ARIMA and ECM models” was chosen to study

RESEARCH QUESTIONS

This study will answer:

- How will the CPls of Vietnam be forecasted in the period 2022-2025?

- Whether and how do the gasoline price and money supply M2 affect Vietnam’s inflation?

The following sub-questions will be considered:

- How long the lag period is when using gasoline price and money supply M2? to predict Vietnam’s CPI?

- What are the policy recommendations to control the prices that follow the inflation targeting set by Vietnam’s National Assembly?

- What measures should the government consider when developing policies? RESEARCH SCOPE

This study will use several econometrics models to forecast CPIs for the 2022- 2025 period and examine the relationship between some macroeconomic indicators in predicting inflation in Vietnam Excepting from descriptive analysis, this thesis uses ARIMA model to forecast the inflation rate and evaluate the impacts of economic variables on inflation by ECM

For annual and monthly CPIs, ARIMA model is used as a predicting model to determine the inflation rate in Vietnam The relationship between macroeconomic indicators is determined using ECM models Following that, the thesis will discuss policy recommendations related to the topic and suggest appropriate solutions THESIS STRUCTURE

This thesis is divided into four main chapters, excluding the introduction, references and appendices:

Chapter 1 Theoretical framework and literature review

This chapter is split into two sections The first section outlines the fundamental ideas

of inflation and other economic measures as well as explains how these measures are computed in Vietnam The second section discusses previous studies on forecasting the inflation rate and the influence of other economic indicators on inflation Studies are categorized into groups based on their major area of research or main technique of study to make it easier to follow

Chapter 2 Contextual analysis

This chapter presents real statistics of inflation and the economic situation in Vietnam from 2000 to 2022 to give the big picture of Vietnam’s economy Furthermore, not only this chapter illustrates Vietnam's system for petroleum price adjustment and management but also it analyzes monetary policies for controlling the money supply Finally, this chapter examines the historical growth of gasoline price, money supply and CPI values in Vietnam to form a picture of the relationship

Chapter 3 Data and methodology

This chapter introduces datasets and methods that used in this study The data source and descriptive statistics from datasets are also presented In addition, the two econometric models used to forecast and analyze the macroeconomic relationships are Autoregressive Integrated Moving Average (ARIMA) and Error Correction Model (ECM)

Chapter 4 Data analysis and results

This chapter describes and analyzes findings from two models In addition, it gives the predicting inflation results

To determine CPI, we need to determine:

- A basket of goods includes all retail goods and services for customers to use in their daily life during a particular period This basket is used to reflect the average changes in prices that customers pay for their cost of living

- Weights for CPI compilation are the percentages of spending on each group in the list These proportions may change over time due to economic development and household expenditures

Therefore, CPI depends on the price changes, list of goods and their corresponding weights in the consumer basket

In the rapid development of the economy, the basket of goods and the weights have been updated to be suitable for the change in the real life, because there are a large number of goods and services provided in the market Customers will change the products they use in their daily life because they have many options to choose from

On the other hand, when people’s life quality and the economy are improved, customer spending structure may differ between periods It requires GSO to update the list of representative goods and the corresponding expenditure of each group This may lead to the volatility of CPI over time

In Vietnam, CPI was calculated for the first time by GSO in 1998 with 300 items in the representative goods list After some updates, the calculation method for CPI in

the 2020-2025 period is determined with the base year 2019 and the basket of representative goods and services including 754 items.

At present, CPI is calculated based on 11 major groups Overall, based on Table 10 (Appendix A), the weight of the food and foodstuff group is the highest, which is 33.56% This group accounts for a large amount of spending in household expenditure because food is the goods that families consume in their daily life The second greatest

is the housing and construction materials group, which has a weight of 18.82% The proportions of other groups, such as education, entertainment, medicine and healthcare, are quite small, approximately 4-7%

Throughout the period, as shown in Appendix A, Figure 14, the chart illustrates the changes in the weights of each group when calculating CPIs Category I (Food and foodstuff) accounted for the largest proportion, which was nearly 50% This group showed a downward trend from 42.85% in the 2006-2010 period to 33.56% in the 2021-2025 period In contrast, growth was recorded in group IV, which is housing and construction materials Other groups were changing but the changes here were slight They shared the same figures that the weight was less than 10% for each group According to GSO, Vietnam’s CPI is calculated by applying the Laspeyres formula, which measures the change in the prices based on the base year weighting

where ['~*: CPI value at reported time p compared with base time 0;

De: the price of goods } at time J];

[}*: the price of goods pat base time 0;

DỆ: weights for CPI compilation in the base year

Equation | is to calculate long-term CPI because it compares the CPI with the base time 0 However, this formula may create difficulties when choosing representative goods to replace the old goods that no longer exist in the market To tackle this problem, Vietnam’s CPI now is determined by the transformed Laspeyres formula, which is known as comparing the price to the short-term base period

where

where Ï ”Š: CPI value at reported time J, compared with time 0;

re ™*: individual CPI value of goods | in the previous month, compared with the time 0; I" 'C- individual CPI value of goods | inthe reported month, compared with the previous month;

DỆ: weights for CPI compilation in a base year

In each area, CPI is computed based on 8 economic zones, from the reports of each province in the region Different areas, such as urban and rural areas, have different weights of the representative goods These CPIs are used to determine the regional CPIs for both rural and urban areas After that, the CPI for the whole country is calculated from the regional ones

1.1.2 Inflation

According to GSO, core inflation is a statistic which indicates long-term price changes It means the changes in the cost of living of residents in one country In Vietnam, inflation is estimated from CPI values, after removing food and food products, energy, and state-managed commodities such as healthcare and education from the basket of representative goods If the CPI increases from 100 to 110 in a year, it is known that the price level has risen by 10% In other words, the annual inflation rate is 10%

While GSO announced multiple CPI datasets, a definition of inflation is varied by each researcher, based on the purpose, the method of study and the way that the dataset is used Some analysts choose the method to determine the inflation level by using the base year or month In this study, inflation is known as the change of price level in that period compared to it in the previous period, with the previous value equaling 100

In Vietnam, some CPI datasets were published to calculate the inflation rate, as follows:

- Monthly CPI compared to the previous month

- Monthly CPI compared to December of the previous year

- Monthly CPI compared to the same period of the previous year

- Monthly CPI compared to the base period (commonly the year 2014 or 2019)

- Annual CPI compared to the previous year (with previous year = 100), that announced at the end of the year

However, provided that there are many ways to publish the CPI dataset, the topic of determining the inflation rate is open to discussion This raises the question of which the most suitable data series to use to forecast CPI values is This is because the price

of one month cannot be used to predict the price for the whole year Moreover, the time before important occasions, such as New Year's Holiday, and summer vacation, is typically a period of price volatility To put it another way, monthly CPIs may be fluctuated and are affected by the seasonal trend

Basically, because one of GSO’s releases is the change in prices of representative goods compared to the previous year, this value is represented inflation for the whole year Its advantage is that annual CPls cancel out the seasonal and trending patterns

On the other hand, according to Tu N.H (2007), the average prices of goods in that year should be used to calculate CPI and then determine inflation This study will compare the forecasting accuracy in calculating the inflation rate of both two approaches Those are predicted from historical annual CPIs and the monthly CPIs’ average

In addition, the relation between inflation and other indicators (oil price, money supply) is still controversial, so in this study, this relation is tested to give a conclusion about Vietnam’s situation ECM framework is used to determine the impact of gasoline price and money supply M2 on CPIs in both the short and long term While many studies show that there is no relation between inflation and macroeconomic measures, other scholars confirm that these criteria have a significant influence on inflation

1.2.1 CPI prediction

Most research on forecasting CPI has been carried out using univariate models including ARIMA Nyoni, T (2019) forecasted UK consumer price index using ARIMA models for annual CPI data from 1960 to 2017 This study recommended policymakers use strict fiscal and monetary policy tools to control the UK’s inflation

In this study, only AIC was considered an indicator in choosing the best model Although AIC is the most common indicator, the study would have been more interesting if it had included other indicators, such as SIC and HQIC, to compare the accuracy of forecasting models

Studying China’s economy, which has similarities with our economy, Zhang, F W., Che, W G., Xu, B B., & Xu, J Z (2013) chose the ARMA model in forecasting because of its stmple approach and high accuracy However, the ARMA models are able to predict the inflation rate in the short-term period In the long run, the forecast values are underestimated ARMA models have also required a stable time series, so this model is more suitable for predicting annual CPIs, especially for the next one- or two-year CPIs

Not only used in the economic research but ARIMA ts also taken into account in the study in other fields Borucka, A (2018) provided a mathematical description of the risk of road accidents in Poland as well as future accident prediction based on SARIMA model The predicted error was 30%, which is considered an acceptable finding The results might be used not only to identify possible threats but also to educate about road safety

In their detailed study of forecasting inflation, Pufnik, A., & Kunovac, D (2006) concluded that a CPI prediction generated by aggregating forecast values of the index’s components is more accurate than a direct estimate from the basket of goods The problem in Croatia’s data was the insufficient length of the data and the changes

in the CPI calculating method This issue is similar to Vietnam's

situation due to the changes in policy and approach methods in determining and calculating CPI values In addition, the paper suggest how seasonal ARIMA model becomes the appropriate forecasting model This result is due to the way the model deal with seasonal patterns

In the same way, for monthly inflation data, the Seasonal ARIMA model 1s the most accurate forecasting in the short term, according to Omane-Adjepong, M., Oduro, F T., & Oduro, S D (2013) in the study of Ghana’s inflation This seasonal model performed as the appropriate method because of the small values of the forecast

accuracy measures

In their work, Almasarweh, M., & Alwadi, S (2018) presented that ARIMA provided the best performance in short-term prediction Although banking time series data was non-stationary and high in the variance of the residuals, the forecasting price index results by using ARIMA were helpful for the investment decisions However, the forecasting accuracy was gradually reduced from period to period, so this model is not suitable for predicting in the long run In contrast, Vo, N., & Slepaczuk, R (2022) found that the hybrid ARIMA-SGARCH model would give better performance when predicting S&P500 log returns in creating algorithmic investing strategies over the long term These differences in conclusion here can be attributed to the combined approach

One of the drawbacks of the traditional ARIMA model is that it has a considerable variance when forecasting time series With the advancement of technology, Li, Z., Han, J., & Song, Y (2020) integrated ARIMA with deep learning model to forecast the monthly financial time series This combined approach not only improved the ARIMA model's forecasting accuracy but also minimized the complexity of a deep learning model It might be used to reduce the investment risks of a stock index The ARIMA- LSTM model had better performance and stability compared to the single ones The authors also stated that the combined model is appropriate for low-frequency time series

Using data on a monthly basis, Karadzic, V., & Pejovic, B (2021) compared the accuracy of the three types of statistical models, such as ARIMA, Holt-Winters and Neural Network Auto Regressive (NNAR) models in predicting Harmonized Index of Consumer Prices (HICP) The finding of this article is that ARIMA models provide the most accurate forecast for EU countries Jere, S., Banda, A., Chilyabanyama, R., & Moyo, E (2019) compared forecast accuracy between ARIMA and multi- cointegration approach based on monthly data in Zambia They

showed that ECM was the more accurate CPI predicting model because of its smallest

errors

Collectively, while the model is easy to use and interpret, it gives the most accurate performance when forecasting the next one or two periods A number of papers outline a vital role in choosing ARIMA as a CPI forecasting model One major drawback of this approach is that its accuracy may reduce when exists shocks, so it is suitable for short-term predicting Therefore, the forecasting results might have been more convincing if other complicated methods were adopted

1.2.2, Relationship of macroeconomic indicators and inflation

A number of authors have considered the influences of macroeconomic variables and the inflation rate by using multivariate and cointegration models Nhu L.T.Q (2015) used ECM model to determine the impact of gasoline price on Vietnamese inflation with the monthly data from 2005 to 2014 This model showed the dependence of CPI

on fuel price in short-term In the long run, this relationship was quite weak However, the author did not offer an explanation for the speed at which CPI converges to long- term equilibrium

Using Vector Autoregression (VAR) analysis, the relationship between the CPI and the fuel price index in Turkey was examined by Celik, T., & Akgil, B (2011) They revealed that the change in fuel prices was the one-way Graner cause for the change in CPI because with one year lag, a 1% increase in oil prices led to a CPI growth at 1.26% Nevertheless, throughout the 2005-2010 period, there were fluctuations in Turkey's oil prices Although this condition has increased the possibilities of rising oil prices influencing inflation rates, it is seen as one of the study's weaknesses

Hossain, A A (2005) suggested that the CPI and money supply M2 form a weakly relationship using ECM framework It also indicated that there had been a long- term relationship between inflation and money supply growth Likewise, Hung, L V., & Pfau, W D (2009) found no strong relation between money supply and inflation by using VAR analysis

As noted by Richard, E O., & Victor, O (2013), the insurance practice had a direct impact on the growth of the Nigerian economy in both the short-term and long- term These findings from ECM implied that the msurance business would make a significant contribution to the growth of the Nigerian economy in the long run

Thus, more efforts were needed to help the msurance business, particularly in risk management and product innovation

In the same vein, Ogbulu, O M., & Eze, G P (2016) used ECM analysis to find out that the performance of financial institutions was influenced by credit risk measures such as return on equity (ROE), return on total assets (ROTA), and return on shareholders’ funds The authors recommended that banks in Nigeria pay close attention to their credit risk management strategies to improve their operational performance

In 1993, Manning, L M., & Andrianacos, D used VAR model to find the long- run equilibrium relationship between dollar movement and the United States price level The study implied that there had been no long-term relation between CPI and money supply M2 In the short run, shocks in CPI have the predictive power for M2 However, because these data were based upon data from over decades ago, the result may be sensitive to the period and the structural changes in the economy from 1973 to

1991 in the US

In a similar approach, Kim, K H (1998) studied the relationship between the US producer price index and other variables such as dollar exchange rate, money supply M2, aggregate income, and interest rate VECM model was used to test the cointegration between variables The paper found that all the variables have an equilibrtum relation More specifically, inflation and money supply are positively related However, the study used PPI instead of CPI to avoid the impact of non- traded goods According to the author, using CPI also leads to the same conclusion Using three statistical models, Uko, A K., & Nkoro, E (2012) tested the performance

of models in different periods They mentioned ARIMA as an example of a benchmark model in inflation forecasting in Nigeria Although VAR is suitable for short-term predicting, it is revealed that M2 plays a minimal role in forecasting price levels in a long-run period by using ECM

Macro factors determining inflation in Vietnam in the period 2000-2010 formed the central focus of a study by Hang N.T.T and Thanh N.D (2010) In their study, the authors concluded that inflation is affected by the monetary policy However, they offered no explanation for the lag period when the policy takes effect

While most studies in the field of using multivariate forecasting models have only focused on predicting values, Salmanpour, A., & Bahloli, P (2011) used ARCH- GARCH model to investigate the relationship between inflation and inflation

uncertainty In addition, they concluded that the money supply has positive effects on the inflation rate Money supply growth is one of the main factors affecting inflation

in Iran

Chou, W L., Denis, K F., & Lee, C F (1996) compared the hedging performance of using ECM and the conventional model in Japan’s Nikkei Stock Average (NSA) index and the NSA index futures data They implied that the hedge ratios estimated by ECM outperformed in eliminating the risk of the cash position Furthermore, this work showed that temporal aggregation influenced the magnitude of the hedge ratios Taken together, ECM is recognised as one of the practical approaches to investigating both short- and long-term relationships between variables It seems that it is necessary

to examine the link between macroeconomic factors and the inflation rate, but there are many conflicts about this topic Selection bias is one of the potential concerns because each author applied different research approaches to investigate varied economic situations It is also useful if there have studies that focus on calculating the lag period to forecast CPI by using historical data series in Vietnam

Figure 1 Annual CPls in Vietnam from 1998 to 2021

In the 2000-2020 period, CPIs showed an upward trend While the inflation in the period 2000-2003 slightly increased, the 2006-2011 period saw a jump in the price level There were many reasons for the high inflation rate in this period, such as government spending, and expansionary monetary policy to maintain the goal of economic development Besides, there were many unpredictable fluctuations, which were the change in calculating approach, the financial crisis of 2007-2008, and the impacts of the COVID-19 pandemic - the newest event In these years, inflation was high as the CPI rose sharply Overall, the increase in inflation was curbed and achieved the target set by the National Assembly each year This is because of the Government's and Prime Minister's directions and guidelines, together with the collaboration among line ministries and localities

In addition, Vietnam’s inflation has seasonal patterns This is because of the change in the households’ spending in a year For example, families usually spend a large amount of money at the end of the year, which is the occasion to decorate their house and purchase new furniture and appliances for the becoming year At this time, a large sum of money is spent on food and foodstuff, which is accounted

for the largest proportion in CPI compilation Another occasion that witnessed a high volume in consumption expenditure is before the summer occasion Many families decide to have a vacation, sometimes a luxurious vacation, after the long and stressed working time In contrast, the time after the Tet holidays, shows a reduction in families’ expenditure The reason for this is that people do not spend their money on pricey expenditures, but they mainly need to spend money on daily living expenses Because the future values are mainly affected by the recent past values, in this study, the economic situations of recent years will be mentioned to illustrate an overview of CPls and related issues 2020 had been predicted as a successful year with the new leaps in economic development Our country is deeply integrated with other nations

by signing new trade agreements and cooperation frameworks Moreover, Vietnam held the role of ASEAN Chair 2020 and the hosted country of SEAGAMES 31 It was forecasted as new opportunities for developing our economy However, the COVID-

19 pandemics and the lockdown periods had created many challenges to the socio- economic situation

In the face of such difficulty, Vietnam put in a great deal of flexible effort to overcome these obstacles in adapting to the fast change in regional and global situations In

2020, CPI increased by 3.23% compared to that in 2019 It achieved the target of under 4% that was set by the National Assembly At the beginning of the year 2021, CPI was forecasted to be rapidly increasing because the pandemics had been controlled and the “normal life’ was coming back However, according to GSO, CPI in

2021 increased by 1.84% compared with that in the same period in 2020 The CPlIs increased due to the reason that the price of domestic goods and services in Vietnam’s market strongly depended on the global prices of fuels and raw materials It means that when the world market prices fluctuate, the domestic prices are also volatile The period 2020-2022, particularly 2021, witnessed the outbreak of the pandemic, which affected all aspects of socio-economic life This resulted in a decreasing trend

in domestic goods prices over January 2020 - April 2020 period, but these prices recovered from May 2020 to October 2021 At the beginning of the pandemic, the food and foodstuff prices rapidly rose because of the fear of shortages of basic goods causing residents to hoard essentials The prices of medicine and medical equipment grew because the pandemic spread led to the high demand for these products Additionally, in 2021, countries provided better

responses to the disease such as many countries stepping up vaccination to prepare for safe reopening, and global economies and financial markets were gradually recovered Meanwhile, the supply chain disruption had not fully recovered and did not keep up with the sudden increase in demand, leading to higher transportation costs and fuels and raw materials prices

Surprisingly, due to the impacts of natural disasters and epidemics, while many countries saw high inflation rates, the Vietnamese inflation over these two years was controlled During the lockdown, residents cannot go out of their homes and tighten their spending because of the risk of prolonged disease Thus, the total demand for consumer goods and services was not high as expected Similarly, the spending on tourism and travelling slumped because of the spread of new virus variants The postponement in transportation activities caused the prices of fuels and raw materials

in the global market to fall In addition, the government implemented incentive packages for people who were affected by the pandemic such as a reduction in electricity prices The price stabilization policies ensured the balance of supply and demand in the market These policies helped to reduce social uncertainty and prevent the negative impacts of the pandemic

In 2022, there is two inflation predicting scenarios in Vietnam Although the pandemic

is controlled and the economy is recovering, it is possible that the whole year inflation will be higher than 4% compared to the plan The reason for this is that the prices of goods rise because of the increasing demand in the market and the price of petroleum surge because of the war rage in Ukraine The roadmap of increasing tuition fees, according to Decree No.81/2021/ND-CP, results in the growing expenditure on the education service group Nevertheless, the increase or decrease of 1% is acceptable because of economic fluctuations on both local and global scales Vietnam’s economy

is also small-scale, developing and vulnerable

The second scenario is the better scenario and the more reasonable one as well At the market and pricing seminar, experts predicted that inflation would remain at a low level, and the inflation target of 4% remained achievable There exists pressure on annual CPI in 2022, but the inflation rate will be increased between 2-3%, compared

to this in 2021 Goods prices will rise due to the spread of new variant Omicron, war rages in Ukraine, the lookdown of China’s economy and global political conflicts However, the global supply chain recovery and the reopening of several industries such as tourism are the highlight points that help to reduce the risk of inflation

where [] is the money supply;

[] is the velocity of money;

[] is the average price level for transactions in the economy;

[] is the total quantity of goods and services produced;

it is clear that the money supply has a direct impact on the price level

In addition, M1 is known as the medium of exchange but M2 is the store of value M2 includes everything in M1, however, it adds other types of deposits Money supply

MI contains highly liquid assets such as cash, checkable deposits, and traveller's checks, which are coins and cash in circulation M2 is a less liquid money supply that contains M1 plus savings deposits, certificates of deposit, and money market funds

To achieve the economic development goal, the government may adopt an expansionary monetary policy It involves lowering interest rates and expanding the money supply in order to boost economic activity If the money supply increases too fast, it will create a liquidity trap This means that residents decide to keep their assets

in cash and monetary policy becomes powerless In recent years, the State Bank of Vietnam (SBV) has announced two targets for money supply growth and inflation each year In 2010, for example, SBV set a money supply growth targeting of 20% and an inflation targeting of 8% It is obvious that increasing the money supply and reducing inflation cannot happen at the same time The policy of setting money supply growth at a fixed rate will reduce its flexibility in controlling inflation In addition, setting the money supply growth target at 20% was unreasonably high at that time because the economy had just come out of the financial crisis Pursuing excessively high money supply growth for long period will lead to hyperinflation and create price bubbles in the market Because of this, there is a need to understand the dependence of the money supply on inflation in Vietnam

[ Vh My cll NOW poe ly phil

|

Figure 2 Money supply M2 growth in Vietnam from January 2010 to February

2022

Several factors contribute to the expansion of the money supply, resulting in economic inflation, but in the case of Vietnam, macroeconomic management policies and investment capital flows are considered the primary contributors

First, many measures, based on expansionary fiscal and monetary policies, were implemented to achieve economic development goals These measures are, for instance, keeping a low reserve requirement ratio; purchasing USD to stabilize the exchange rate; expanding public investment through different development programs and supporting the state-owned business sector However, because of the increasing accumulation of money over time and the inefficiency in public investment, these policies have the potential to produce inflation

Secondly, the money supply expanded too rapidly, but there was no corresponding action to keep the money in circulation at a reasonable amount Furthermore, low investment efficiency from state-owned firms and an increase in capital inflows from outside sources contribute to the high rate of money supply growth Every year, huge amounts of foreign money enter Vietnam in the form of Official Development Assistance (ODA), Foreign Direct Investment (FDI) and remittances

While these factors create opportunities for economic development, these approaches may also contribute to inflation and cause negative effects on the economy It indicates that the changes in money supply and monetary policies may

affect the prices and inflation rate of an economy Thus, it is concise when using money supply M2 to determine inflation

2.3 GASOLINE PRICE

Because petrol is a necessary commodity for other production industries, changes in gasoline price can affect the prices of other commodities in the basket of goods Moreover, fuel plays an important role in the economy It is also a vital input source for other industries such as transportation, heating, and operating In the structure of CPI, because transportation accounts for a considerable proportion, it is clear that petrol prices can affect inflation

by the fuel prices in the global market This change in the calculating method not only results in faster reactions of petroleum enterprises but also boosts their ability to set their selling prices within a defined framework

Furthermore, petrol prices are announced on the 1“, 11", and 21 of each month by the Ministry of Industry and Trade This adjustment schedule makes the time gap between global and domestic gasoline prices smaller than it was It means petrol prices reflect immediately any fluctuations in the international market

Although Vietnam is a big oil exporter, our country is an even bigger importer This result is due to the way of consuming crude oil Crude oil after exploitation is exported, and then we import crude oil from other countries to refine and consume domestically This is common because the imported crude from the Middle East is more suitable for the production in Dung Quat Refinery The price when importing crude and producing it also brings more profit than its price when exploiting domestically In addition, since the launch of Nghi Son Refinery in 2018, the domestic petroleum supply is expected to be guaranteed Therefore, the domestic oil price will

be more stable

In such context, this policy change in the petroleum business can be both beneficial and harmful in various aspects The advantage is that retail oil prices in Vietnam are likely to be brought closer to world prices than that was in the past Domestic fuel prices will respond quickly to price shocks in the worldwide market By contrast, with the shorter adjustment period, the gasoline price is expected to be more fluctuate, compared with the prices in the previous period Furthermore, petroleum prices are modified under the Government’s control; for example, with the help of the Petroleum Price Stability Fund, the calculating price approaches determined by Ministry of Finance or the measures to stabilize the price These actions help to balance the benefits of both customers and petrol sellers The State does not need to invest a big sum of money to fill the gap between local and international gasoline prices This implies that these pricing policies are going in the nght direction

Because petroleum costs account for around 3.52% of total production costs in the economy, Nguyen Bich Lam, former head of the GSO, said that a 10% increase in its price might cut 0.5 percentage point off GDP growth Moreover, oil consumption contributes greater than the individual contributions of several key industries in the economy, such as construction and real estate This indicates that for such a huge contribution, a change in petroleum prices must have a considerable influence on macroeconomic performance, on CPI, for instance

2.4, FLUCTUATIONS OF GASOLINE PRICE, MONEY SUPPLY AND CPI

Figure 4 CPI and gasoline price growth from January 2018 to February 2022

It can be seen from the chart above that the relationship between gasoline price growth and the CPI was quite weak As explained earlier, the price of gasoline has an impact on inflation, which is predicted as a long-term effect Changes in gasoline price will immediately affect the prices of other commodities, thereby exerting an impact on the price level The chart also shows that CPI does not change much although there have great fluctuations in oil prices Many previously published studies stated that gasoline had a direct impact on the CPI However, the adjustments in petrol price management policies recently are affecting the way gasoline price influence the CPIs

Regarding Figure 4, since the start of the pandemic, in 2020, all activities in transportation were delayed This led to the plunge in gasoline prices It is clearly obvious in June 2020, the price was 11,930VND per litre, which was the cheapest price in the last few years, while at the time pre-pandemic, the price of gasoline usually fluctuated around 18,000-20,000VND per litre The year 2022 witnessed an economic recovery after two years of the pandemic Tourism and transportation bounced back leading to the high demand for petroleum Moreover, in January 2022, domestic petroleum supply was scarce after Nghi Son Refinery reduced its production capacity Although taxes on oil account for 40% of its price, authorities had not made any decision about cutting off these taxes These factors have increased the gasoline price

In spite of the fact that petroleum prices witnessed a period of erratic behaviour, a slight rise was recorded in CPI When gas prices fluctuate, the CPI often responds slowly In mid-2020, while oil selling price hit its trough, inflation declined by 2% In the following months, CPI values remained stable at around 100 It means the inflation was stabilized, maybe just a few percentage points When CPI increased at the beginning of 2021, there was seen modest growth in gasoline prices

While M2 have a direct effect on inflation, this impact takes effect after several months Hoa N.T.L (2013) mentioned that the money supply created an impact on inflation after 6 months because of the demand-pull and cost-push factors in Vietnam’s economy Additionally, Figure 5 also illustrates that the M2 growth and CPI lines are in the same fashion

CHAPTER 3

DATA AND METHODOLOGY

This section presents the descriptive analysis of two datasets and introduces ARIMA, ECM methods that used in this study

3.1 DATA

In this study, there are two datasets which include CPI and two economic indicators values

Table 1 Variables description

Variable names Indicator names Definitions

CPI Annual/Monthly CPI Consumer Price Index

depend on the dataset) in Vietnam

RON-95 selling price

GP Gasoline price that announced each

adjustment period Money supply is M2 Money supply M2

ysupply selected as M2

The first dataset is Vietnam’s annual CPIs from 1998 to 2021, based on the General Statistics Office announcements It means that the time series includes 5 periods of calculation CPI, which are based on the changes in weights for CP] compilation They are 1998-2000, 2001-2005, 2006-2010, 2011-2015, 2016-2020 The

forecasted period is the 2022-2015 period

The second dataset contains three variables, which are monthly CPlIs, gasoline price and money supply M2 The first variable is monthly CPls, published by GSO Data on gasoline price (RON-95) was from the Institute of Economics and Finance (IEF), while money supply data was collected from the database of SBV The data was collected from January 2010 to February 2022

In both datasets, the CPI value for the next period is calculated based on the requirement that the previous period’s CPI is 100

As shown in Figure | in the previous chapter, the chart illustrated a plot of annual Vietnam’s CPIs on a yearly basis between 1998 and 2021 The mean annual CPI was 106.03 In recent years, from 2016 to the present, the CPIs showed fluctuations

around 102 What can be seen in this chart is the high inflation rates of the year 2008 and 2011, at 122.97 and 118.58, respectively It is indicated that inflation rates in Vietnam were high and fluctuated, particularly, some years had double- digit inflation

Table 2 Descriptive statistics for monthly dataset

| | | | - | | Ì

Figure 6 Monthly CPls in Vietnam from January 2010 to February 2022 Briefly, monthly CPI values were increased with fluctuation Most values are around 100.4, which means inflation increase 0.4% per month CPI reached a peak of 103.32

in April 2011, whereas its minimum value was 98.46 after nine years, in April 2020 However, the CPI is not volatile much, sometimes owing to shocks that cause the monthly CPI to significantly rise or fall

Two datasets contain some high values in 2008 and 2011, but there is no considerable evidence that these outliers are aberrant The 2020-2022 period witnessed unpredicted shocks due to the pandemic and war rage These shocks affect CP] values in the same way as the financial crisis in 2008 Moreover, the number of high values accounted for 10% of the number of observations It is not suitable that remove all of these Therefore, these values are not considered outliers and the author decides to keep these in the time series

3.2 METHODOLOGY

3.2.1 ARIMA

3.2.1.1 Differencing

ARIMA requires the time series data to be stationary If not, the most popular method

is to differentiate it It means subtracting the next values from the previous values Depending on the series, multiple differencing may be required Therefore, instead of predicting the initial time series, ARIMA model predicts the differences between the series from one time step to another time step The order of differencing times is the parameter [] in the model, usually 0, 1 or 2 If the series

contains positive autocorrelations out to a high number of lags, a higher level of differencing is required

3.2.1.2 Testing autocorrelation

To model the time series data accurately, it is vital to check for autocorrelation Testing autocorrelation means testing how values in the same series relate It can be tested in two methods The informal way is a graphical method The second is using the autocorrelation test, which is called Durbin Watson test

This test is used for testing the autocorrelations of residuals of the statical model The null hypothesis of the test is Ho: no first-order autocorrelation The DW statistic is

where Pp: are residuals from a regression

If the test statistic is not 2, the Ho is rejected It means there are autocorrelations

On the other hand, the Ljung-Box Test is one of the approaches to test for the absence

of autocorrelation in the series It examines residuals from a fitted time series model to check whether all underlying error autocorrelations are 0 In this study, this test is focused on checking all the errors are white noise This means the errors are identically and independently distributed

The test statistic is

r&"

where }) is the accumulated sample autocorrelations;

Pp is the time lag

The null hypothesis will be rejected if p (p | is larger than(1 — o quantile This shows that the time series are not autocorrelated In other words, the observation today is uncorrelated with that of previous days

3.2.1.3 Testing stationary

Time series needs stationary data for the precise forecast Theoretically, the stationary series means that it follows the same pattern forwards as it does

backwards Stationary means the series has constant mean and variance; moreover, the covariance of }- and (| + p}'- terms is not time-dependent

An ADF test is used to determine if the series is stationary The null hypothesis of the test is that the series has a unit root In other words, the series is nonstationary If the p-value is smaller than 0.05, Ho is rejected, so the data is considered stationary data

In contrast, Kwiatkowski-Phillips-Schmidt-Shin Test (KPSS) also test the stationary

of time series, but the null hypothesis is that the series is stationary The p-value is used to determine the rejection of the null hypothesis

3.2.1.4, ACF and PACF

An ACF reflects the correlations between the values of the time series at two different points in time The order of lags is related to different lag lengths For instance, the first lag correlation is the correlation between the values at time Pand those at time []

— 1 The correlation at the second lag, similarly, is measured by the correlation of the values between time [] and [] — 2 ACF plot is the visual representation of the autocorrelation of the time series

A PACF is similar to the ACF, except that the PACF is a conditional correlation It is also the correlation between different pomts in a time series, after considering the effect between the values of other points of time In a stmple way, PACF captures the correlation between the time series and lagged version of itself At the second lag, for example, the PACF measures not only the correlation of the values between time [] and[] — 2 but also those between time [] and [] — 1

3.2.1.5 Box-Jenkins (ARIMA) model

ARIMA is Auto Regressive Integrated Moving Average This model is based on the idea that past values can alone be used to predict future values The time series must

be stationary so that this series has the same autocorrelation for any lags regardless of time ARIMA model is used to analyze and forecast data by simply requiring a series

of historical values

Auto Regressive (AR) model is a linear model that forecasts dependent variables depending on their past values The parameter p is the order of the AR term and refers

to the number of lags when forecasting

where []: is the time series where the model will be applied;

py is the AR coefficients;

pr is the error term

Moving Average (MA) model shows that past errors affect current observation The number of lagged errors is the []': order inthe model

where P is the time series where the model will be applied;

py is the MA coefficients;

[) 1s the past error

Autoregressive terms, moving average terms, and differencing operations are three components in ARIMA model Therefore, the ppp{][](]) J, []} is formed

(O00 0 ok = [[ + pe + he to t+ blo

th — Pic — [[l‹* — *!: — [bho (8) where [}, [], [} and [) are all as defined above

A seasonal ARIMA is an extension of ARIMA model that add the seasonal component

to remove additive seasonal effects SARIMA’s parameters are determined in the same way as in ARIMA model The notation for this model is

ĐDP |p P, Pip p, Pipl

where [],[], [] are non-seasonal terms as defined above;

(| 0 are seasonal autoregressive, moving average, and differencing terms, respectively;

Pis the number of time steps in a single season

To build an ARIMA model, first, start with the regression with no explanatory variables, then add AR and MA terms

ACF and PACF are used for choosing parameters p and p for ARIMA model The two should be considered together All number of lags that has a value above the critical value are considered To reduce the model complexity, the minimal lag number 1s chosen

In ARIMA, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) or Schwarz Information Criterion (SIC) are used to determine the most