Thực trạng dự báo thuế giá trị gia tăng trường hợp tại Việt Nam

Nội dung text: Thực trạng dự báo thuế giá trị gia tăng trường hợp tại Việt Nam

1. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 CURRENT SITUATION OF VALUE ADDED TAX FORECASTING THE CASE IN VIETNAM THỰC TRẠNG DỰ BÁO THUẾ GIÁ TRỊ GIA TĂNG TRƯỜNG HỢP TẠI VIỆT NAM Nguyen Thi Hoai Phuong, Ha Kieu Oanh National Economics University phuongnh@neu.edu.vn ABSTRACT In many countries’ current structure of tax revenue, value added tax (VAT) is the largest contributor due to its simplicity, efficiency and stability. In France, VAT is the most important source for the national treasury, accounts for 45% of total tax revenue. In Vietnam, the proportion of VAT revenue compares to total state budget revenue increases rapidly and become the most important source of state budget: increases from 22.36% in the period 2006 - 2010 to 23.78% in the period 2011 - 2015, continued to increase to 24.48% in 2016, and 25.49% in 2018 (Ministry of Finance, 2018). Therefore, the forecasting of VAT has great impact on the preparation process of budget revenue, as well as the state budget plan. The main subject of this article is an analysis on the method and current situation of forecasting revenue from VAT, from which some solutions and recommendations for VAT forecasting in Vietnam will be proposed. Derived from analysis of the current situation of VAT policy, VAT revenue and influencing factors; bases on the collected data, the authors propose to apply four models of VAT revenue forecasting, including: ARIMA model, macroeconomic model, forecasting method based on effective tax rate (ETR) model and monthly revenue forecasting model. The research results indicate that the use of monthly revenue forecasting model and macroeconomic model are the most suitable in the case of Vietnam. Từ khóa: Value added tax, tax revenue forecast, monthly revenue forecast, elasticity. TÓM TẮT Trong cơ cấu nguồn thu thuế hiện nay ở nhiều quốc gia thì thuế giá trị gia tăng (VAT) chiếm tỷ trọng lớn nhất do tính đơn giản, hiệu quả và ổn định. Tại Pháp, VAT là nguồn thu quan trọng nhất đối với kho bạc nhà nước, chiếm khoảng 45% tổng doanh thu thuế. Tại Việt Nam, tỷ lệ thu VAT so với tổng thu ngân sách nhà nước tăng nhanh và trở thành nguồn thu quan trọng nhất của ngân sách nhà nước: tăng từ 22.36% trong giai đoạn 2006 - 2010 lên 23.78% trong giai đoạn 2011 - 2015, tiếp tục tăng lên 24.48% trong năm 2016, và 25.49% trong năm 2018 (Bộ Tài chính, 2018). Vì vậy, việc dự báo thuế giá trị gia tăng có tác động lớn đến quá trình chuẩn bị nguồn thu ngân sách cũng như kế hoạch ngân sách nhà nước. Đối tượng nghiên cứu chính của bài viết là phân tích về phương pháp dự báo và thực trạng dự báo nguồn thu từ thuế GTGT. Trên cơ sở đó, đề xuất một số giải pháp và khuyến nghị cho công tác dự báo thuế GTGT tại Việt Nam. Xuất phát từ phân tích tình hình hiện tại của chính sách VAT, doanh thu VAT và các yếu tố ảnh hưởng; dựa trên dữ liệu thu thập được, các tác giả đề xuất áp dụng bốn mô hình dự báo doanh thu VAT, bao gồm: mô hình ARIMA, mô hình kinh tế vĩ mô, phương pháp dự báo dựa trên mô hình thuế suất hiệu quả (ETR) và mô hình dự báo doanh thu hàng tháng. Kết quả nghiên cứu chỉ ra rằng việc sử dụng mô hình dự báo doanh thu hàng tháng và mô hình kinh tế vĩ mô là phù hợp nhất trong trường hợp của Việt Nam. Keywords: Thuế giá trị gia tăng, dự báo thu thuế, dự báo thu tháng, độ co dãn. 1. Introduction In the market economy, forecasting is extremely important as it provides necessary information to identify and efficiently allocate resources in the future. For the state budget, budget revenue forecast is the starting and most important point. In the current tax revenue structure of many countries, value added tax is the largest distributor due to its simplicity, efficiency and stability. Therefore, the forecasting of value added tax has great impact on the preparation process of state budget plan. The process of forecasting VAT revenue also establishes a reliable information network and database, from which to predict a country’s potential tax revenue evaluate the impacts of tax policy 269
2. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 changes such as tax rate, tax base, tax exemption on tax revenue. Information on tax collection efforts and compliance rate has also been collected, helping to improve tax administration and tax reforms towards an effective tax system. In addition, the forecasting of VAT revenue also contributes to evaluate the impacts of macroeconomic policy changes on tax policy and tax revenue, helping to develop and implement fiscal policy effectively. Specifically, estimate the effects of such factors as growth rate of the economy or of economic sectors on VAT collection amount, evaluate the impacts of macroeconomic policy changes on tax policy and tax revenue, for example the limitation of import tariff barriers, trade liberalization. Therefore, the authors focus on analyzing VAT revenue forecasting methods and the current status in Vietnam. On that basis, some solutions and recommendations will be proposed to improve the forecast of VAT revenue. To accurately forecast VAT revenue, the forecasting agency must use complex models to forecast the impact of Vietnamese economy on the collected revenue. Base on the current data and tax policies in Vietnam, the authors' results indicate that using the monthly revenue forecasting model and macroeconomic model are much more appropriated. In the coming time, it is necessary to build a database system, especially for the macroeconomic model, to simulate the impact of tax policy changes (such as tax rates, preferential conditions, exemptions) on the tax payable or tax liability incurred by a typical taxpayer or industry. 2. Literature review and theoretical basis for VAT revenue forecasting methods Tax revenue forecast has long been studied and applied, especially in developed countries. Such organizations as the IMF and the OECD have done a great deal of researches and produced a number of reports to support countries, especially developing countries, in applying models for forecasting revenue. Theoretical and empirical researches have identified 5 methods for forecasting VAT revenue, including auto regressive model, effective tax rate model, monthly revenue forecasting model, macroeconomic and microeconomic models. King and John (1995) have shown that the simplest and unconditional forecasting method is the extrapolation of linear trend formed by revenue over the years. Tt = f (Ti-1, Ti-2 ). Legeida and Sologoub (2003) used the Auto Regressive Integrated Moving Average (ARIMA) to forecast VAT revenue in Ukraine. Accordingly, VAT revenue in the current period correlated with that in the previous period (self- correlation), and the disturbance (random shock) in the previous period affected that in the current period. Legeida and Sologoub (2003) also used the method of forecasting VAT revenue basing on effective tax rate (ETR). To forecast tax revenue in future periods, we calculate effective tax rate by dividing the tax collection amount by the estimated tax base. Usually effective tax rate is lower than statutory tax rate. This difference may come from tax exemption or taxpayer compliance problem. Then tax revenue forecast is calculated by multiplying tax base forecast in the following period by effective tax for the current period. Under effective rate approach, the major challenge is to estimate potential VAT base. Chan Yan Koo (2000)’s study on methods of forecasting tax revenue with current tax policies and amendment proposals on China’s tax law includes VAT forecasting models. He proposed monthly revenue forecast model for VAT to develop a short term forecast for tax management. Accordingly, to forecast VAT revenue in the current year, it is essential to calculate the growth rate compared to the same period of previous year, basing on monthly collection of the last 12 months and current year, as well as expected GDP growth rate for current year. The forecasting method applying microeconomic model is based on the fundamental principle that if the tax law is not changed, the increase in VAT revenue will be due to the increase in tax base over time and the elasticity of tax revenue with respect to tax base. According to Mackenzie (1991), tax base must be identified to calculate VAT forecast. Since VAT follows the destination principle and applies on 270
3. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 final expenditure on goods and services in the domestic economy, the VAT base must correspond to expenditure on final consumption and government expenditure on goods and services. To reflect the impact of such other factors as policy on tax revenue, the dummy variable model can be used (according to Singer, 1968). While macroeconomic models approach the issue from the perspective of the national economy, micro simulation models focus on actions or behaviors of sectors and individuals affected by relating public policies. Specifically, micro models simulate impacts of tax policy changes (such as tax rate, preferential conditions, tax exemption) on the tax payable amount or the arising tax debt of a typical taxpayer or sector. Then, the model sums up the data of each taxpayer, sector and calculate arising VAT from current tax laws and regulations. The micro simulation models can be considered by approaching the I/O table. According to Le Minh Tuan (2007), to calculate VAT using I/O model, we should begin with the final consumption of goods and services (both domestics and imports) reflected via the final demand matrix. After that, this consumption is adjusted to tax system and tax policy issues, including: VAT (included tax in the retail price), duty-free goods (via tax rate), compliance rate, etc. Specifically, to accurately forecast the tax revenue for the following year, we need to develop a final consumption input- output table (I-O) of all sectors and exempt industries of the previous year, from which to calculate the potential tax of each sector of the current year (with corresponding tax rate) and that of the following year (basing on consumption growth rate forecast and policy changes on tax rate and tax base). The potential tax of the previous year will be compared with actual revenue to determine compliance rate. Compliance rate will be adjusted with potential VAT revenue of the following year to provide a forecast to actual VAT. 3. Current situation of forecasting value-added tax (VAT) in Vietnam The estimated revenues for different types of VAT shall be made as follow: The General Department of Taxation will use the database and forecast the collection amount of VAT inclusive of the domestic VATs; on the other hand, The General Department of Customs is responsible for managing and reporting the VAT receipts related to the purchase of the goods. Forecast for next year is mainly based on the growth rate of current tax revenue compared to the same period of last year. Appropriate growth rates are adjusted in line with the pace of growth of relevant industries, sectors and regions of the economy to provide the ultimate level of growth. 3.1. Methodology The method used by tax authorities to forecast revenue is the extrapolation method combined with the expert method. The extrapolation method is based on the assumption that the VAT revenue will continue to follow a trend similar to the preceding period using the information of a short-term nature and especially based on experts with substantial experience in the field of collecting VAT in order to assess growth rate, revenue forecast, etc. This method is quite simple in terms of technics and easy to calculate. In this method, the tax authorities use aggregate forecasting indicators to calculate and determine VAT revenue in a given period (short-term forecasts, monthly and quarterly and yearly results). Tax revenue is forecasted based on the assumption that the revenue procedure is not broken by the impact of the mechanism, policy or structural change of the economy. Two steps to forecast VAT collection include: Step 1: Before making the forecast, the forecasting agency must carry out the analysis and assessment of the situation of revenue in the period in order to adjust the extraordinary revenues arising out of the norm and have the basis to forecast growth indicators for next year. This analysis is made of actual data from years and opinions of experts in the industry. Step 2: Use the indicators to conduct the forecast. The most frequently used indicators are the ratio of revenue, growth rate and adjust ratio on GDP. 271
4. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 - Earnings growth index: The average annual growth rate (usually five years) is calculated according to the average formula of the growth rate over the years. Tax of year t ^(1/n) Average annual growth rate of the year = Tax of Year (t-n) Forecasted tax amount = Previous year tax amount x Average growth rate - Tax revenue ratio/density: The indicator is used to forecast the amount of tax collected in a year. Tax revenue density is the percentage (%) of the tax collected in the assessment period over the annual tax rate determined by the formula: Amount of tax collected in n months (Year X) Revenue of n months (Year X) = x 100% Amount of tax collected in Year X - Adjusted ratio on GDP: Previous year tax amount Proposed tax amount = x GDP forecast Previous year GDP In addition, the tax authorities are applying microeconomic -based method to forecast the tax revenue including VAT. This method forecasts revenues based on analyzing, calculating the collection amount from businesses, the taxpayers (accounting for 80% of the total revenue), from which to estimate the overall. The traditional forecasting method applied by the tax authorities has not met the management requirements, requiring the forecasting task to be innovated and modernized so that the forecast can be quick and accurate to effectively serve the macro management of the state. Since 2006, the European Technical Assistance Project (ETV2 / PTF2) has been implemented by the European Technical Assistance (EU) project to the Ministry of Finance (MOF). From 2009 up to now, the General Department of Taxation has continued updating data, running model tests, analyzing and evaluating forecast results based on state budget revenues and experts' consultations and comparing with reality to gradually adjust the model. Up to now, the General Department of Taxation has started to pilot some revenue forecasting models such as monthly forecasting model for monthly revenue forecast for each type of tax and total tax revenue; the model of value-added tax forecasts based on the IO balance sheet in 2000. However, the selection and application of the most optimal VAT forecasting models have not yet been done. 3.2. Declaration of results of forecasting value added tax The work of forecasting tax revenue including VAT has initially made some changes, namely: - The revenue forecasting now has a staffing structure for data collection and tax collection. - Assessment of the level of completion of consolidated plans and details of VAT; Identification of the objective and subjective causes that affect the process of implementing the revenue estimate. 272
5. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 - Analysis of the pros and cons in the process of managing the organization of tax collection by tax authorities at all levels; influencing factors and lessons learned to improve the organization of revenue management and develop an appropriate process of managing collection. - Accurate determination the ability of taxpayers to pay taxes as well as the tax contributions of different regions and economic sectors. However, the VAT forecasting has some limitations, such as: - Up to now, the analysis, forecasting and construction of revenue estimates have not been fully and correctly acknowledged. Therefore, the investment in the analysis and forecasting is limited as it does not apply modern technology to the analysis; and does not improve the capacity and facilities to serve the task of collecting, processing, synthesizing, archiving and providing information to the statistical system to create a source database for analysis and projection. - Slow adoption of the new method of predictive analysis and estimation of revenue. Up to now, the taxation industry has been using traditional analytical methods, mainly based on the trends of previous years and the determination of economic indicators such as output, price, sales, cost, income of a number of products and services that have large tax payments. The current VAT forecast does not really apply to either a micro- or macro-economic model but only to the trend of revenues. Although VAT is a fairly stable tax in terms of tax policy and tax base, in the situation of current tax reform requirements, the simple forecasting method is no longer appropriate. - The quality of the statistical reporting system is still low and does not meet the requirements of modern forecasting analysis. The database used for forecasting analysis has not been prioritized, invested, reviewed, archived systematically and is not abundant. In addition, the identification of an effective forecasting model is not available, so the data input for forecasting is not sufficient and updated, many statistical indicators of the tax sector do not comply with national statistical standards and not in accordance with international practice. On the other hand, the unification and integration of databases on tax calculations and underlying economic parameters affecting tax collection efforts are not available. - The capacity and experience of forecasting staff are lacking. The number of analytical and forecasting staff is limited. Most analysts and forecasters have not been trained in finance and taxation, and have not yet had access to quantitative analysis and forecasting methods. - Theoretical basis for the development of analysis and forecast model has not been invested in proper research, but instead, is in the initial stage of research. - The results of the forecast analysis are generally local, sketchy, emotional, not quantifiable to determine the relationship between the components of GDP and the revenue results as well as their mutual impacts. The standard deviation between the state budget and actual settlement is large over the years. In the period 2003-2016, the difference between the amount collected in the balance sheet and the estimate was about 9.87%. In the post-crisis period from 2010 -2013 the level of bias was up to 18.61%; specifically, in 2010, the draft budget for VAT were 123,977 billion dong, but the final settlement of VAT was 155,022 billion dong; and in 2013 the VAT in the estimation is 258,494 billion but in the settlement report is 208.535 billion dong. This may be due to the fact that revenue forecasts do not reflect the economical and political changes. In addition, the actual VAT collected in the finalization reports is lower than the estimated VAT amount over many years, which may be due to tax incentives, tax reductions, tax compliance and tax refunds in export and investment activities. 273
6. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 Figure 1: VAT estimated and settled in period 2003 -2015 Source: Ministry of Finance, 2016 4. The application of VAT revenue forecasting model in Vietnam Derived from the analysis of the actual status of VAT receipts and based on available data from the General Department of Taxation, the General Statistics Office, the Ministry of Finance and the Ministry of Planning and Investment; the research team proposes the application of four models for VAT revenue forecasting, including the Autoregressive Integrated Moving Average (ARIMA) model, the macroeconomic based model, the effective tax rate model, and the monthly revenue forecasting model. 4.1. Models used in forecasting 4.1.1. ARIMA model The ARIMA model forecasts future VAT revenues derived from its past values. The amount of VAT collected in the current period is believed to correlate with the amount collected in the previous period (self-correlation); and the disturbance (random shock) in the previous period affected the disturbance in the present period. To forecast ARIMA-based VAT revenues, the team used a series of 44 quarterly VAT observations from 2005 to 2016 from the Ministry of Finance budget report. The data string used in the ARIMA model is integrated. So to predict VAT revenues for 2017 with the ARIMA model, we need to consider whether the chain is an integrated sequence. The Unit Root Test for the differential first-order value chain of VAT resulted in a p-value of 0 (below the critical 1%, 5%, 10%), which demonstrates the rejection of the hypothesis H0: variable with unit tests; and the data we are considering has integrated. Also, note the p-value of C and Trend, if this p- value is statistically significant, then the data has an even and trending coefficient, where the resulting table is correct. Therefore, the VAT data collected in step 1 differential has been stopped. Table 1: ADF test results of the VAT data chain Data series Level ADF test Prob Critical value at 5% level VATq Deviation 1 -9.563293 0.0000 -3.596616 Source: Author's data processing results Identification of the ARIMA model (p, d, q) is to find the appropriate values of p, d, q. The above tested VAT data chain indicates that the chain stops at level 1, we have d = 1. The determination of p and q will depend on the SPAC graphs = f (t) and SAC = f (t). Through the self-correlation graph and the partial correlation of the first-order variable VAT value chain, there is a statistically significant (PAC1) correlation coefficient, which may be appropriate for the model AR (1). As such, p can have a value of 1. Similarly to the definition of p, we see that q can have a value of 1. Thus we have the ARIMA model (p, 1, q). 274
7. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 Figure 2: Self-correlation graph and partial correlation of the sequence Source: Author's data processing results In fact, the VAT rate is less variable, so VAT revenues often depend on the potential tax base as the final expenditure of the individual and the government. While final consumption usually has seasonal factors, in quarter (Q) 1 and Q4 consumption levels tend to increase. Therefore, the team will consider adding seasonal factors when making predictions using the ARIMA (SARIMA) model. To test the suitability of the models we are based on AIC / SBC / HQ standards. Table 2: Model selection according to AIC / SBC / HQ standards Model AIC SBC HQ AR(1)MA(1) SAR(4) 19.88138 20.04358 19.94153 AR(1) MA(1) SMA(4) 19.88138 20.04358 19.94153 MA(1) 19.91448 20.03613 19.95959 Source: Author's data processing results According to the AIC standard, the model chosen for the current VAT collection is self-correlated with VAT revenues and shuffles (random shocks) in the previous period and a fourth-grade delay. According to the SBC standard, the model chosen for the current VAT collection is self-correlated with VAT revenues and shocks (random shocks) in a previous seasonal period and 4th grade delay. According to the HQ standard, the selected model is the VAT collection in the current period only correlates to random shocks in the previous period. After estimating the SARIMA models obtained according to the AIC/SBC/HQ standards, we have a residual stop sequence and most of the latencies are in terms of white noise. Therefore the selected models are suitable, unmotivated and have a small variance. Proceeding from the models according to the selection criteria, we have forecasted the VAT revenue for 2017 as follows: 275
8. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 Table 3: Results of the 2017 VAT forecast from the self-regression model Unit: Billion VND Model 2017 Q1 2017 Q2 2017 Q3 2017 Q4 AR(1)MA(1) SAR(4) 72629.35 73997.82 74366.28 73734.74 AR(1) MA(1)SMA(4) 71887.24 72617.86 72861.02 72207.14 MA(1) 71760.06 70577.71 71893.12 72241.59 Source: Author's data processing results In the selected models, when forecasted according to the HQ standard, the error of the forecast is the lowest: 4.77%. Total forecasted VAT for 2017 according to the model is VND 286,472.48 billion. Table 4: Forecast error of self-regression models Root Mean Mean Mean Model Squared Error Absolute Error Abs. Percent Error AR(1)MA(1) SAR(4) 4124.044 3592.031 5.72175 AR(1)MA(1)SMA(4) 4096.548 3571.552 5.692394 MA(1) 3557.33 2934.791 4.768390 Source: Author's data processing results 4.1.2. Model based on macroeconomic variables The forecast of VAT revenues may be derived from establishing a stable empirical relationship between the growth of VAT revenues and the corresponding increase in the tax base. In this method, the team chooses to consider the relationship between VAT and GDP. Data is taken as a series of data on VAT and GDP representing the quarterly tax base for the period 2005 - 2016 from the Ministry of Finance and the General Statistics Office. Data entered into the model is the natural log of VAT and GDP. In addition, according to the analysis of tax policy change and VAT collection, the team added further dummy variable D to the model to examine the change in other factors on VAT revenues. Specifically D assumes a value of 0 representing the pre-crisis period of 2009 and D assumes a value of 1 for the post-2009 period. The team used the following log-linear model: Ln VAT = α + β * ln GDP + γD + ui. Linear regression Number of obs = 44 F( 2, 41) = 128.73 Prob > F = 0.0000 R-squared = 0.8205 Root MSE = .26456 Robust lnvat Coef. Std. Err. t P>|t| [95% Conf. Interval] lngdp .4956721 .1048775 4.73 0.000 .2838676 .7074766 d .8392019 .1099357 7.63 0.000 .6171823 1.061222 _cons 3.997765 1.208623 3.31 0.002 1.556903 6.438628 Figure 3: Results of regression model estimation Source: Author's data processing results 276
9. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 After the hypothesis testing of the model the estimate was obtained without deviation and the coefficients were statistically significant at α levels of 5% \ Ramsey RESET test using powers of the fitted values of lnvat Ho: model has no omitted variables F(3, 38) = 1.03 Prob > F = 0.3893 Figure 4: Regression model test results Source: Author's data processing results The model's results show that average VAT revenues increased by 0.495% on a 1% increase in GDP, given that other factors remained unchanged, and that the 2009 economic crisis had an impact on the VAT. We have a model for forecasting VAT revenues by GDP as follows: LnVAT = 3.997765 + 0.4956721 LnGDP + 0.8392019 D + ui Based on the forecast of the world and domestic economic outlook for the period 2016-2020 from the National Center for Socio-Economic Information and Forecasting, we have forecasted GDP growth, from which forecasted VAT receipts as follows: Table 5: Results of the VAT forecast for 2016-2018 from the regression model Low-end High-end Year GDP (%) VATf (billion dong) GDP (%) VATf (billion dong) 2016 6.2 267,195.54 6.67 267,799.46 2017 6.35 275,605.57 6.83 276,865.65 2018 6.75 284,826.74 7.62 287,322.92 Source: Author's data processing results 12.0 Forecast: LNVATF 11.5 Actual: LNVAT Forecast sample: 2005Q1 2015Q4 11.0 Included observations: 44 Root Mean Squared Error 0.255385 10.5 Mean Absolute Error 0.206444 Mean Abs. Percent Error 2.042470 10.0 Theil Inequality Coefficient 0.012330 Bias Proportion 0.000000 9.5 Variance Proportion 0.049425 Covariance Proportion 0.950575 9.0 8.5 05 06 07 08 09 10 11 12 13 14 15 LNVATF ± 2 S.E. Figure 5: Forecast errors of the regression model Source: Author's data processing results 277
10. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 Thus, according to the forecast model, VAT revenue in 2017 is estimated at 275,605.57 billion dong with a 6.35% GDP growth scenario; and 276,865.65 billion dong with the GDP growth scenario of 6.83%. The predicted error in the average sample is 2.04%. 4.1.3. Forecasting Method Based on Effective Tax Rate (ETR) To forecast tax revenue in the near future, we need to calculate effective tax rates by dividing the tax revenue by the estimated tax base. Then, we forecast the tax revenue by multiplying the tax base forecast for the following period with the effective tax rate for the current period. The data taken was the VAT amount for the period of 2005 - 2016 according to the Ministry of Finance's final settlement report of the state budget; the potential tax base to be used is the final consumption expenditure of individuals, households and the government for the period 2005-2016 from the World Bank database. Table 6: ETR of the VAT period 2003 – 2016 Year VAT Final Expenditure ETR 2003 33,130 512,454.87 6.46% 2004 38,814 621,501.67 6.25% 2005 45,878 635,765.00 7.22% 2006 55,148 725,509.00 7.60% 2007 69,822 922,651.00 7.57% 2008 91,506 1,246,796.00 7.34% 2009 108,549 1,324,379.00 8.20% 2010 155,022 1,564,832.00 9.91% 2011 192,064 2,067,736.00 9.29% 2012 174,056 2,247,562.00 7.74% 2013 208,536 2,550,787.55 8.18% 2014 241,103 2,752,258.80 8.76% 2015 259,229 3,003,264.80 8.63% 2016 277,192 3,342.54 8.87% Source: Author's data processing results Calculate the ETRs for the years 2003 to 2016 by taking the actual VAT divided by the final consumption for goods and services. It can be seen that the ETR of VAT is rather high and is increasingly improving, reflecting a good level of tax compliance. Given the assumption that the VAT rate is unchanged in the next period, we can obtain the ETR of 2016 at 8.87%. According to the Center for Economic Analysis and Forecast, the final consumption forecast for 2017 rose 7.2 percent. VAT in 2017 is forecasted at 277,893.5 billion dong. 4.1.4. Monthly revenue forecast model With the data from the Internal Revenue report of the General Department of Taxation in the first 4 months of 2017 and 12 months of 2016, we can forecast the VAT revenue of the last 8 months of 2017. 278
11. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 Actual growth rate of total VAT revenue for the 4 months of 2017 compared to the same period of 2016 is 18.76%. The growth factor for the subsequent months of the period from 2016 to 2017 is estimated at 10.38% assuming a GDP growth rate of 6.2%; and 10.7% assuming a 2016 GDP growth rate of 6.67%. In fact, VAT is a very high effective tax rate (ETR), which can reach up to 8.63% in 2016. The ETR is the ratio between the actual and potential tax revenue. Or else the ratio C (equal to the effective tax rate divided by the standard rate, here is 10%) is 86.3%. This means that tax exemptions are low and compliance is high. In the monthly revenue forecast model, the author does not adjust the change in effective tax rates in the following months of 2017. According to the model of monthly revenue forecast, in the last 8 months of 2017, the domestic VAT amount is 141,017 billion dong; and the total domestic VAT in 2017 is 207.537 billion dong with the GDP growth rate of 6.67%. Table 7: Forecast of local VAT in 2017 2017 2016 GDP growth rate 6.2% GDP growth rate 6.67% Month 1 15914.244 22840.07885 22840.07885 Month 2 13740.86039 16864.69926 16864.69926 Month 3 12444.93438 13495.68306 13495.68306 Month 4 13913.30023 13319.58115 13319.58115 Month 5 13962.41669 15412.53012 15456.27903 Month 6 13462.62017 14860.82557 14903.00844 Month 7 12988.87184 14337.87452 14378.57299 Month 8 11982.98833 13227.52162 13265.06831 Month 9 16085.95672 17756.61749 17807.02016 Month 10 22188.42783 24492.88113 24562.40487 Month 11 19817.21775 21875.40111 21937.49506 Month 12 17260.66191 19053.32562 19107.40902 Source: General Department of Taxation, 2017 and author calculations 4.2. Comparison of models for forecasting value added tax in Vietnam With actual data of 2017 VAT revenue is 255,724 billion dong (Dang Thi Han Ni, 2019), the following models of VAT prediction can be compared as follow: The ARIMA model will help to forecast with greater confidence from traditional econometric modeling approaches, especially for short-term forecasting and future forecast environments with little variation. Modeling tools and model data are simple and easy to collect. 279
12. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 In practice, however, VAT revenues depend on changes in the tax base, standard tax rates, tax compliance, and changes in policy and economic environment. Therefore, the application of the ARIMA model is just a reference to consider the evolution of VAT from the past to speculate for the future rather than to show the impact of other factors on the actual tax revenue. Applying the ARIMA model, VAT in 2017 is forecasted at 286,472.48 billion dong. The ETR-based model forecasts the VAT revenues from potential tax bases, and the effective tax rates. Therefore, this approach takes into account taxpayer issues or exemptions. But in this approach, the difficulty is to estimate the potential VAT base, not available in the tax authorities database. The identification of potential tax bases is limited to household and government final consumption, which has to be adjusted to the added value of the exempt industry and of the duty-free establishments. Applying the model, VAT in 2017 is forecasted at 277,893.5 billion dong. Monthly forecasting model is implemented with monthly revenue generation rather than actual tax potential. Monthly forecasting is a useful tool for allocating targets for taxation and monitoring, made for tax officials and for forecasting seasonal revenues. Applying the model is also simpler when data requirements are very small. Just calculate the monthly revenue of the last 12 months of the year, the monthly revenue this year and the GDP growth rate expected for the current year. However, this model is not suitable for analyzing impacts on tax revenue and is limited to the expected monthly revenue forecast. According to the model of monthly revenue forecasting, the total domestic VAT in 2017 is 207,537 billion dong with the GDP growth rate of 6.67%. The macroeconomic model allows forecasting the VAT revenues in terms of elasticity with GDP and can produce fairly predictable results and allow for longer-term forecasts (1 to 3 years). Specifically: Firstly, VAT revenues depend largely on the tax base because VAT is stable and less subject to tax rate and compliance. Especially, since the introduction of the VAT law in 1997 till now, the VAT policy has not changed much in terms of calculation method, taxable price and tax rate. After the VAT law was amended in 2003 up to now, the standard VAT rate is only 10% and the preferential rate is 5% and 10% respectively. In addition, the average C rate (equal to the effective rate / standard rate) of the period 2005- 2016 is more than 81.6%, meaning that the deduction is not high and the actual tax loss rate is not high; especially, because the taxation is bound between the stages of production, circulation and consumption of goods and services. Secondly, GDP as a proxy for economic activity has a theoretical macroeconomic background. GDP is capable of explaining specifically to tax revenues, and may eliminate other macroeconomic factors that do not have statistical significance in the model. Furthermore, the use of one or more explanatory variables in the model minimizes or eliminates the problem of multi-collinear. Thirdly, the macroeconomic model also takes into account some of the effects of policy changes in certain periods by including dummy variables into the model. However, according to regression results, R2 is not good (82%), the model has some disadvantages: Firstly, the determination of the VAT base derived from GDP is not sufficient. The VAT base must come from the government, individual and household end users; adjusted for total value added in tax- exempt areas such as finance, insurance, agriculture. Secondly, the change in the taxpayers attitudes, or in other words, the level of compliance is not reflected in the model. 280
13. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 Thirdly, revenues and GDP have mutual impacts. Here, both tax revenues and GDP are no longer exogenous factors but have joined together within the model. Finally, in the coming time, VAT reform focuses on one issue: revising and supplementing to reduce the number of groups of goods and services not added to VAT; groups of goods and services are subject to a tax rate of 5%. This means increasing the number of goods and services subject to a tax rate of 10%. The provisions on turnover thresholds for the application of VAT declaration forms are compatible with the market economy regime managed by the State and international practices. There will be changes in the tax base of the VAT, and the tax rates of some industries, which will have an impact on the effective tax rate and the level of VAT compliance. This will change potential VAT bases, potential VAT amounts and actual VAT amounts. The application of a simple macroeconomic model between GDP and VAT will not reflect all of these changes. According to the macroeconomic model, VAT revenue in 2017 is estimated at 275,605.57 billion. The research results indicate that the use of monthly revenue forecasting model and macroeconomic model is most suitable in the case of Vietnam. 5. Solutions and recommendations Based on the analysis of the current situation of VAT forecasting and the application of forecasting models, the research team proposed some solutions to enhance the forecasting of state budget revenue in Vietnam in three steps, including: developing a logical forecasting model, building database, and finally implementing the model. Firstly, regarding forecasting model: The application of simple model like monthly forecasting model is meaningful in monitoring and allocating targets in tax collection management, whereas ARIMA model is only for reference when impacts of factors on VAT revenue cannot be evaluated. To provide an accurate VAT revenue forecast, the forecasting agency needs to use a variety of complex models so as to forecast the impact of Vietnam’s typical economy on the collected amount. The microeconomic model allows the forecasting of potential VAT revenue over a 1-3 year period for the state budget estimation, nevertheless, it does not accurately reflect the impact of each policy on VAT revenue. The application of this model brings effective forecast of VAT collection amount since it allows considerately accurate calculation of tax base and changes of such factors as tax rate, tax base, compliance rate on VAT revenue. However, in Vietnam, the current I/O table has only been updated by 2012, and the incompatibility between data on budget revenue from taxation by sectors and the division of economic sectors in the I/O table of General Statistics Office has caused difficulties for forecasting. Therefore, the forecasting agency must develop its own model, in particular the model for calculating tax base and arising VAT in each sector, from which to forecast VAT revenue when there are changes in tax rate and tax exemption in specific industries. Secondly, regarding building database for the model: The accuracy and connectivity of data will determine the final quality of the models and their results. Therefore, the forecasting agency needs to develop a consistent, complete and timely database on budget revenue and other indicators in the forecasting model. Database needs to be collected from three following sources: - Available data at tax sections, divisions, departments such as VAT categorized by taxpayers, VAT refund on a monthly, quarterly and annual basis must be collected and reviewed. - The data collected from relevant agencies must be compatible with forecasting models’ input data such as GDP, final consumption, value added by sectors, VAT on imports, VAT refund on exports. 281
14. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 - The new data corresponding to variables designed by the forecasting agency to put into the forecasting model that has not been collected, for example: VAT base, VAT by sectors, etc. These information can be obtained from the management of taxpayers, VAT declarations, invoices and documents at all tax district departments and tax departments across the country. Accordingly, each invoice has its own code on the electronic declaration system. We can base on the fact that this invoice code does not appear in the VAT deduction ( final consumption) to calculate VAT. Likewise, VAT base and VAT number by sector or region can also be determined based on this invoice code. So when there is a change in tax rate and tax base policy, we can analyze the fluctuation of VAT revenue. Thirdly, regarding implementing the forecasting model Currently, VAT forecast and state budget forecast are both performed by the State Budget Department. This will lead to inevitably limitations in forecasting, since tax administration agency is the one who directly implements and deeply understands VAT policy. Meanwhile, the Department of Estimation of Taxation under the General Department of Taxation only estimates the collection of VAT in the domestic market, not including VAT on imports and VAT refund on exports, which is managed by the Customs Department. According to the study, the forecasting agency must agree on functions and duties of related units. The General Department of Taxation should organize a specialized unit for forecasting different types of taxes, including VAT. This unit is under the State Budget Department and responsible for analyzing and evaluating the changes in the economy and tax policy, thus introducing an appropriate forecasting model for each period. To determine an appropriate model, the forecasting agency also designs variables into the model and develops statistical indicators for data collection. The Customs Department, the General Statistics Office and other relevant agencies are responsible for supporting and providing statistical data for forecasting. On the other hand, staff training must incorporate theories on forecasting models and practical application in order to improve the quality of forecasting. The General Department of Taxation’s staff should be trained in both econometrics, such as understanding and applying E views, SPSS soft wares, and specialized knowledge in taxation and economics, such as analyzing tax policy. Human resources training is not limited to staff engaged in the work of forecasting but also to statisticians. Because without knowledge of VAT forecasting will lead to the misunderstanding of indicators and variables in the model, thus distorting the data and producing unexpected forecast. REFERENCES [1] Chan Yan Koo (2000), "Estimation of Tax Revenue and Tax Capacity", JDI Executive Programs, No. 2000-08. [2] Dang Thi Han Ni (2019), “Tax gaps in the field of flat tax for business households”, Proceeding of the scientific conference of young officials and lecturers and graduate students of the University of Economics and Law 2019. [2] Gangadha Prasad Shukla, Pham Duc Minh, Engelschalk, M. & Le Minh Tuan (2011),“Tax Reform in Vietnam Gearing Towards a more Efficient and Equitable System”, Economic Management and Poverty Reduction in Asia and the Pacific, World Bank. [3] General Department of Taxation, General report on domestic revenue in 2003 – 2016. [4] King, J. (1995), "Alternative Methods of Revenue Forecasting and Estimating", Tax Policy Handbook, P. Shome. Washington, DC: International Monetary Fund, 254-257. [5] Legeida, N. & Sologoub, D. (2003), "Modeling Value Added Tax (VAT) Revenues and Transition Economy: Case of Ukraine". Working paper No.22, Institute for Economic Research and Policy Consulting. 282
15. INTERNATIONAL CONFERENCE FOR YOUNG RESEARCHERS IN ECONOMICS & BUSINESS 2019 ICYREB 2019 [6] Le Minh Tuan (2007), "Estimating the VAT Base: Method and Application", Tax Notes International, No. 2, Volume 46, 203-2010. [7] Mackenzie, G.A. (1991), "Estimating the Base of the Value-Added Tax (VAT) in Developing Countries: The Problem of Exemptions", IMF Working Paper, WP/91/21. [8] Ministry of Finance (2003-2018), Finalization of state budget. [9] Ministry of Finance (2003-2018), State Budget estimates. [10] Singer, N.M. (1968), "The Use of Dummy Variables in Estimating the Income-Elasticity of Income-Tax Revenues", National Tax Journal, No. 2, Episode 21, 200-204. [11] World Bank (2016), The Databank from World Development Indicators . 283