Innovation and economic performance: The role of financial development

1.   Introduction

The importance of China's economic system, its development and its weight in the world scenario, have been ever increasing in the last decades. In spite of this, and the many social accounting matrices (SAM)s for China that have been built in the last twenty years, an overall SAM-based analysis of the structure of China's economy that would allow to get an in-depth knowledge of the inter-sector relationships has not been attempted yet. To the end of filling this gap, we elaborated a double-version 2015 SAM for China, the first one quite detailed and the second one more agammaegated. The elaboration of a SAM for China at this level of updating and industries, factors and institutional sectors detail is something totally new that brings a remarkable contribution to the improvement of the knowledge in the field. Then, we used the more agammaegated version to conduct a comprehensive analysis of the structure of country's economic system using the impact multiplier model, searching for the level of integration of the productive structure, factors and final demand, the interdependencies among the various industries and the role played by them in the economic growth. This analysis represents an innovative contribution to the understanding of the mechanisms that govern the functioning of the economic system as a whole and provide useful tools for government economic policy.

China's government has undertaken the reform of the indirect taxation system. With tax reform complete and the entire country under the VAT regime, it is extremely interesting to evaluate the impact of this new regime on China's economic system through a general analysis—something that has not yet been done and puts a new and enlarged light in the panorama of studies concerning the economic effects of indirect taxation changes—that allows us to study any interrelationship and the stimulus-response effects to VAT rate cuts for given productive sectors. Indeed, in its nature of indirect tax, VAT typically bears on the production structure and through it, on value added and on household expenditure. Therefore, it is reasonable at all and highly informative and effective to perform the analysis of its impact on the economic fabric in the light of the more general analysis of the overall economic system implemented by means of the SAM based impact multiplier model and of the economic reality that inspired the VAT reform.

In this paper, employing the aforementioned China's 2015 agammaegated SAM, and based on the impact multiplier model, we conduct a never so far performed comprehensive analysis of the structure of country's economic system, and address, for the first time, how to measure the effects on the economic system of cuts in VAT rate due to government policy decisions.

Specifically, the remaining of the paper is organized as follows.In Section 2 we discuss the state of the art in the elaboration of SAMs for China and illustrate the double-version SAM for China that we have developed. In Section 3, by designing the impact multiplier model, identifying the exogenous and endogenous accounts, and solving the model, i.e., estimating the endogenous coefficients and the impact multipliers, we conduct the SAM-based analysis of China's economic system. In Section 4, after illustrating and discussing the literature on the overall effects of China's tax reform, we perform the SAM-based analysis of VAT cuts on China's economic system through the impact multiplier model. Section 5 develops our conclusions.

2.   The existing SAMs for China

2.1.   The State of the art

A number of SAMs have been elaborated to represent the structure of China's economic system. Li (2008) produced a 2002 seven-industry financial SAM for China to shed light on the linkages between the real and financial sides of China's economy. Zeng and Shen (2014) constructed a macro SAM based on the forestry sector following the 2007 I-O table for China, in addition to other statistical data. They verified the accounts related to the forestry sector with reliable estimates of data not included in the Statistical Yearbook of China that could be used as a basis for analysing carbon policy based on a CGE model. A 1997 SAM for China, the source of which was not specified, was used by Li et al. (2004), to develop several policy simulation scenarios and use structural path analysis to analyse the characteristics of the Chinese industrial structure and income distribution. In his thesis dissertation, Qin (2011) constructed a SAM for Beijing, Tianjin, and Hebei based on the 2007 I-O table, accounting information from the 2008 Customs, Finance and Tax Yearbooks of China, and household and government revenue and expenditure from China Statistical Yearbook 2008. Binjian and Sakamoto (2013) compiled a 2002 macro SAM for China to calibrate a CGE model and analyse the effects of macroeconomic policies on income disparities under different assumptions about the factor market in China. Keogh-Brown et al. (2016) analysed the economic burden of Alzheimer's disease in China using a CGE model calibrated on a 2011 SAM for China that was extracted from the GTAP 9 database. To simulate various scenarios of the regional economic and environmental effects of discharge fees using a CGE model, Fang et al. (2016) developed a water resource-water-environment extended 2012 SAM for Jiangsu province. Zhang and Diao (2013), in the framework of activities of the International Food Policy Research Institute (IFPRI), constructed a 2007 SAM for China, covering 61 industries, four types of factors, and rural and urban households, with the aim of assessing the impact of the 2008–2009 global recession and the Chinese government's stimulus policy on China's economic growth. To conclude our overview of the SAMs specifically built to analyze the overall Chinese economic system, it is worth mentioning the environmentally extended social accounting matrix (ESAM) using Chinese data from 1990 proposed by Xie (2000) and Xie and Saltzman (2000).

2.2.   Our special-purpose SAM

All of the above-mentioned SAMs have merits and defects: some of them are detailed enough to be candidates to serve as a basis for our analysis; others are too agammaegated or too specifically oriented towards detailed and sectoral objectives. However, they all share a negative feature for the purpose of our analysis: they are too dated. Even those that are attractive because of their detailed structure and reliability, such as that developed by IFPRI, are far from representing the current structure of China's economic system, specifically in terms of productivity, which grew dramatically, approximately 15% on an annual basis, from 2008 to 2015 and therefore needs to be represented by a SAM that is as up to date as possible.

Consequently, we relied on the 2015 SAM for China purpose built by us at the School of Statistics of the Shanxi University of Finance & Economics and further refined at the Department of Statistics, Computer Sciences, Applications of the University of Florence. It is based on a 2012 42-industry I-O table for China at producers' prices (China Statistical Yearbook, 2016) and on the Flow of Funds Accounts (Physical Transactions) 2012 (China Statistical Yearbook, 2014).

We first elaborated a 67 × 67 2012 SAM consisting of 42 industries, 5 institutional sectors (Non-financial corporations, Financial Institutions, General government, Households and the Rest of the World, (RoW)), with a detailed specification of the operations of the primary and secondary income distributions. The data in the SAM were balanced using the cross-entropy (CE) method (Robinson et al., 2001).

We further updated the SAM to 2015 using the CE method, obtaining a 44 × 44 2015 SAM consisting of 19 industries, because in 2015, GDP data were available for only 19 industries, and such data are imposed as constraints when conducting CE extension. This SAM-version is not reported here but is available on request.

Furthermore, we conducted the 2015 CE updating of the 67 × 67 2012 SAM as follows: since the 2015 GDP data were not sufficiently detailed to be imposed as the constraints for the CE extension, we retained as benchmarks all of the 19 industries and, based on the linearity underlying the I-O model (that is, constant returns to scale and no technical progress), for the items composing each of the 19 industries, we calculated the percentage out of the total and have subdivided the GDP for the respective industry according to this percentage. Thereafter, we applied the CE method by employing as the constraints the newly calculated 42 GDPs (the conversion details are reported in Table 1). The resulting 67x67 2015 SAM-version consisting of 42 industries is not reported here but is available on request.

3.   SAM-based analysis of China's economic system

3.1.   The impact multiplier model: exogenous and endogenous accounts identification

To design the impact multiplier model, we need to identify the exogenous account as well as the endogenous ones. The exogenous account, that is, the policy instrument, is the result of the agammaegation of the following sets of accounts: (i) PA expenditure, (ii) capital accounts, and (iii) RoW. Specifically, the following eight accounts were agammaegated under the label "Government": 1) Government expenditure, 2) Capital transfers, 3) Gross fixed capital formation, 4) Changes in inventories, 5) Acquisition less disposal of non-financial assets, 6) Net lending, 7) Non-financial corporations, 8) Financial corporations, 9) General government, 10) RoW current account, and 11) RoW capital account.

The endogenous accounts, that is, the policy objectives, are the following sets of accounts: (ⅰ) value of goods and services produced (industries), (ⅱ) payment of factors (Gross Value Added (GVA), or alternatively, Gross Domestic Product (GDP)), and (ⅲ) households' income, totalling 33 in the 19-industry 2015 SAM version and 56 in the 42-industry 2015 SAM version.

3.2.   The model solution: the impact multiplier matrix

We obtained the model solution, that is, the estimation of the impact multipliers and then the total demand of the endogenous accounts resulting from an increase in the concerned exogenous account, as follows (Ferrari et al., 2018).

We refer to the 19-industry 2015 SAM and denote by ${x_{ij}}$ the payments from endogenous account j to endogenous account i (i, j = 1, 2, …, 33) or, equivalently, the receipts of endogenous account i from endogenous account j; ${z_i}$ represents the payments from the exogenous account to endogenous account i, and ${l_j}$ represents the receipts of the exogenous account from endogenous account j and from itself, which are considered leakages, because they exit the endogenous part of the economic system and do not contribute to the multiplicative process (Bellù, 2012). The row sum for endogenous account i is ${X_i} = \sum\nolimits_{j = 1}^{33} {} {x_{ij}} + {z_i}$, and that for exogenous account is ${l_{s \bullet }} = \sum\nolimits_{s = 1}^{34} {} {l_s}$. The column sum for endogenous account j is $X_{j} = \sum\nolimits_{i = 1}^{33} x_{i j}+l_{j}$, and that for the exogenous account is $Z = \sum\nolimits_{i = 1}^{33} z_{i}+l_{34}$. For $\mathrm{i} = \mathrm{j}, X_{i} = X_{j}$.

Similarly to I-O analysis, we define the endogenous accounts coefficients: $a_{i j} = \frac{x_{i j}}{X_{j}}(i, j = 1, 2, \ldots, 33)$, and the leakage coefficients, $L_{j} = \frac{l_{j}}{x_{j}}$ (j = 1, 2, …, 34).

Thus, the row sum for endogenous account i can be rewritten as follows:

meaning that the economic system can be represented by a system of simultaneous linear equations¹. In matrix form:

¹ This is equivalent to the following: (ⅰ) assume the absence of substitution between different inputs and factors for all productive sectors and between different final goods for all institutions); (ⅱ) assume that the exogenous expenditure is fully supplied by goods and services from the economic system, that is, that the economic system does not encounter constraints in terms of productive capacity (hypothesis of surplus productive capacity); and (ⅲ) assume that the prices of goods and services do not change because of the impact of changes on exogenous demand (hypothesis of fixed prices).

where X is the (33x1) vector of the 19-industry 2015 SAM industries' and sectors' total demand accounts, including endogenous and exogenous accounts, A is the (33x33) matrix of the endogenous accounts coefficients describing the structure of the economy, and Z is the (33x1) vector of the exogenous account.

Solving (2) with respect to X yields: X (I-A) = Z

and finally,

where the (33x33) (I-A)^-1 = M inverse matrix is the impact multiplier matrix.

The column vectors A_ij, i = 1, …, 21; j = 1, …, 19 of the endogenous coefficient matrix, although to a lesser extent than the technical coefficients in I-O analysis, permit the evaluation of the degree of integration of industry j within China's productive structure, the overall interdependency of industries, and the composition of its GVA and, therefore, the intensity of labour and capital employed.

Because each impact multiplier M_ij, i = j = 1, …, 33, can be interpreted as the partial derivative, $\frac{\partial \boldsymbol{X}_{i}}{\partial \boldsymbol{Z}_{j}}$, of ${X_\mathit{\boldsymbol{i}}}$ with respect to ${X_\mathit{\boldsymbol{i}}}$, it measures the variation in endogenous account i that is attributable to a unitary increase in exogenous demand for endogenous account j.

We have estimated the endogenous coefficients matrix and the impact multiplier matrix, reported in Table 2 and Table 3, respectively.

3.3.   Discussion of the results

Examining Table 2 reveals the main characteristics of industry interdependency and of GVA composition. A reading of the costs (columns) altogether reveals satisfactory average interdependency, meaning a good integration of industries, except for the somewhat lower interdependency between the group of goods industries (items 1 to 7) and some types of service industries, such as "Scientific Research and Technical Services", "Management of Water Conservancy, etc.", "Services to Households, etc.", "Education", "Health and Social Service", "Culture, Sports and Entertainments", and "Public Management, etc.". Note the good integration degree of "Manufacturing" with the other goods industries, with values of the costs that often range from 2.5% to 5% (see column 3, rows 1, 2, 6, 7, 10). In terms of entries (reading by row), there is evidence of an even better integration of "Manufacturing", with coefficient values generally ranging between 10% and 30% and a peak at 50% (see row 3). Another noteworthy finding is the degree of integration of "Construction", "Hotels and Catering Services", "Information Transmission, Software and Information Technology", "Real Estate", and "Management of Water, etc.". Finally, "Financial Intermediation" exhibits an input cost structure that indicates weak integration compared to its good integration performance on the entry side, meaning that financial activity is scarcely dependent on other industries, which viceversa are heavily dependent on it, as is typical in advanced economies.

Out of the total costs, GVA represents a substantial share for several industries. For "Education" it represents approximately 82% (rows 20–21, column 16), with 72% absorbed by compensation of employees (row 20) and 10% only to gross operating surplus (row 21); for "Agriculture, Forestry, etc." it represents more than 62% (rows 20–21, column 1), with 60%—that's practically the total—absorbed by compensation of employees; for "Real Estate" it represents 67% (rows 20–21, column 11), with 56% to gross operating surplus and 10% to compensation of employees. These examples demonstrate the clear labour orientation of the first two industries and the capital orientation of the third one. These are "high GVA producer" industries. Other industries, such as "Health and Social Services", "Wholesale and Retail Trade", "Service to Households, etc.", and "Information Transmission, Software, etc." exhibit GVA shares ranging from approximately 51% to approximately 55%, although in these cases there is no clear evidence of whether they are labour oriented or capital oriented. These are "medium GVA producer" industries. The GVAs of the other industries take 48% or less of the total cost, and there is no clear evidence in favour of them having a labour or capital orientation. These are "medium-low GVA producer" industries.

The impact multiplier coefficients reported in Table 3 confirm, on average, the evidence that emerges from the above analysis of the endogenous coefficients. The relevance of "Manufacturing" within the productive structure is definitively confirmed. Indeed, the related impact multipliers speak to this: the direct multiplier, with its very high value of 2.42 (row 3, column 3), shows the highly dynamic and reactive character of this industry, able to highly multiply self-stimulative activity, and to profoundly respond to the indirect demand from the other industries. Indeed, by column, the indirect multipliers confirm the remarkable average indirect effect of increased demand for manufacturing products on the other industries, namely on "Agriculture, Forestry, etc.", with 20% (row 1, column 3), "Mining", with 14% (row 2, column 3), and "Wholesale and Retail, etc.", and "Financial Intermediation", both with 13% (row 6, column 3, and row 10, column 3, respectively). Although this industry is not a great GVA producer, the reaction to increases in manufacturing demand on labour and capital is remarkable, with impact multiplier effects of 54% (row 20, column 3) and 43% (row 21, column 3), respectively. A notable indirect reaction is observed for variations in household expenditure, at 40%, and for income variation: for household income, the multiplier effect is 96%, whereas that of non-financial corporation income is 43%. By row, the indirect multipliers show the very high average impact of the increased demand for the products of all the other industries on "Manufacturing" (see row 3, where its response to unitary increases in demand for products is in several cases higher than 100%). The responses of "Manufacturing" to increased labour demand, social security payments, allowances, household income households' expenditure are also remarkable (row 3).

3.4.   Impact multiplier decomposition

The nature of the links in China's economy can be better analysed if the impact multipliers are decomposed. Decomposition also creates more transparency of the exogenous shock effects represented by a variation in a given exogenous demand on the output structure and functional and institutional distribution, redistribution and use of income (Pyatt and Round, 1979). We decompose the 19 industry SAM 2015 coefficient matrix, S, as follows: S = Q + R

where

The three matrices contain: S, the SAM direct coefficients, Q, the blocks of the diagonal, and R, the off-diagonal blocks. Regarding the sub-matrices: A = matrix of technical coefficients; V = matrix of VA coefficients; Y = matrix of VA distribution coefficients; C = matrix of expenditure coefficients; and H = matrix of institutional and household distribution coefficients.

Then, we can write the supply and demand balance equation as follows:

where X = vector of sector supply; V = vector of VA by categories; Y = vector of household incomes; ex = vector of exogenous commodity demand; ev = vector of exogenous VA; and ey = vector of exogenous household incomes. Equivalently:

where $(I-S)^{-1}$ represents the matrix of SAM impact multipliers.

We rework (4) using matrix decomposition. After simple algebra, we obtain

where $T = (I-Q)^{-1} R$.

Let us now rewrite (5) as follows:

where

We multiply (8) through by T and then substitute for Tx from (8) in the same equation. After some algebraic manipulations, we obtain:

Again, we multiply through (7) by T and substitute for Tx and obtain:

where

In matrix form:

Thus, matrix M has been decomposed into three multiplicative components, M₁, M₂ and M₃, which contain own (intragroup or direct effect), extra-group (indirect or open loop), and inter-group (cross or closed loop) multipliers, respectively.

Because of the key role it maintains in China's economic structure as stressed above, in Table 4, the decomposition of "Manufacturing" industry vector, j = 3, M_i3 (i = 1, …, 33) = M_1i3 (i = 1, …, 19) M_3i3 (i = 20, …, 33) M_3i3 (i = 1, …, 19), is shown as an example.

Sub-vector M_1i3 shows how the effects of the exogenous shock represented by increased demand for "Manufacturing" products are passed on to other industries. In other words, this vector represents the "within account" effects, that is, the multiplier effects that the exogenous shock to the block of accounts of these industries has on the block itself. Thus, an injection of 1 euro to demand for "Manufacturing" products leads to a 12.34% increase in "Agriculture, Forestry, etc." demand, a 11.59% in crease in "Mining", a 101.94% increase in "Manufacturing", and so forth, up to a 0.3% increase in "Public Management, etc." demand. This effect concerns only industriy accounts, and therefore, M_1i3 = 0, i = 20, …, 33. Sub-vector M_2i3 captures "cross" or "spillover" effects, showing the effects of the exogenous shock that are transmitted to endogenous accounts of blocks other than the industry block, due to the circular flow of income. Thus, the same 1 euro injection increases "Compensation of Employees" by 9.05%, "Gross Operating Surplus" by 8.85% (thus, overall, GVA is increased by 17.90%), and "Household Expenditure" by 7.13%, "Households", "Non-financial Corporations", and "General Government" income by 17.18%, 8.82%, and 7.56%, respectively.This effect concerns only factors' and institutions' accounts, and therefore M_2i3 = 0, i = 1, …, 19. Sub-vector M_3i3 shows, for each industry, the induced portion of the closed-loop effect, i.e., the circular structure of the system from exogenous to endogenous accounts. The economic interpretation of M_3i3 is that the variation in "Manufacturing" demand has travelled out from the original block that first felt the shock, through all the other blocks, and back to "Manufacturing". Consequently, the effect of the above 1 euro shock to "Agriculture" that travels out of it to return through the blocks of factors and institutions to yield variation in "Manufacturing" demand is 1.95%, the effect of the shock to "Industry" that travels out of it to return through the factors and institutions blocks to variation in "Manufacturing" demand is 0.69%, and so forth, up to the 0.03% effect that travels out of "Other Services" to return through factors and institutions blocks to variation in "Manufacturing" demand. As the "Manufacturing" accounts belongs to the industry block, no M_33i coefficient exists other than those of the industry block (that is, M_3i3 = 0, i = 20, …, 33).

4.   SAM-based analysis of VAT cuts on China's economic system

4.1.   China's tax reform: an appraisal of its economic and fiscal effects

4.1.1.   The story

By May 1, 2016, China had entirely abandoned its former indirect double taxation system and implemented a new value-added tax (VAT) system that is collected centrally. It replaced the business tax (BT) system that had been collected by local governments, moving to a unified VAT system applicable to all goods and services. This reform more closely aligned China's tax system with international standards, although it still possesses its own unique characteristics and complexities (Shira and Associates, 2016).

China's indirect tax system was implemented through a bifurcated regime in which a VAT applied to the sale and importation of goods, and a BT applied to the services, and this system had been in place since the country's opening up to the world economy in 1978 and subsequent bold reforms. The system underwent a major overhaul in 1994, as the VAT was expanded to include the sale of goods, and processing and repair services.

The latest phase of the reform began in 2012 in Shanghai with the aim of reducing the tax payments of smaller mainland firms when the transportation industry and modern service industries were included. In 2014 railway transport, postal services and the telecommunications industry were included. In 2016 the VAT was comprehensively implemented as the country's only indirect tax, finally including construction, real estate, finance² and life services, effectively replacing the BT system. The tax reform, which is currently being tested, is part of government's efforts to restructure the Chinese economy from one driven by labour-intensive manufacturing to one that is service-oriented by easing the tax burden on service industries, which have historically paid a disproportionate share.

² As affirmed by Wolfers et al. (2016), ―China's VAT system contains three key features that represent breakthrough policies among VAT/GST systems around the world‖. One of these features is the application of VAT to the financial service sector, specifically, to: ⅰ) interest income, ⅱ) fee and commission-based income, ⅲ) net gains from trading in financial products, and ⅳ) general insurance products.

Indeed, in 2015, services represented more than half of China's Gross Domestic Product (GDP) for the first time, and were growing at a faster rate than any other sector of the economy. Accordingly, the Chinese government fully included the entire service sector under the VAT regime to further propel growth in services and consumption as the country pivots away from low-value-added industries. The broader introduction of the VAT was also designed to encourage low-end manufacturers to upgrade their technology and capabilities, and invest in research and development to move up the value chain.

China's fiscal reforms can be recognized as having helped maintain sustainable growth, thereby achieving the goal of the so-called New Normal, that is, shifting from quantity to quality. This means shifting economic growth from high-speed to middle-to-high speed and shifting the balance of growth away from heavy-industrial investment and toward domestic consumption.

4.1.2.   Economic and fiscal effects

Since the reform was fully implemented approximately a year and a half ago, it is currently too early to evaluate its economic effects, except for the short-term ones. Nonetheless—for the sake of the analysis we intend to conduct—here are interesting and useful, albeit somewhat few, studies in this vein.

Gourdon et al. (2014) focused on export VAT rebates and reported that a 1% rebate results in a 7% increase in exports. A VAT refund is equivalent to a decrease in the VAT rate, and thus, the inverse multiplier effect is very high. Since exports are almost entirely manufactures, we would not be that far from the truth if we concluded that a 1% decrease in VAT results in a 5–6% increase in manufacturing products. For international shippers importing from and exporting to China, VAT implementation entails a 6% rise in shipping costs, as claimed by Maersk et al. (2013), an integrated transport and logistics company, with multiple brands, and a global leader in container shipping and ports. Thus, as the VAT for shipping increased by 6%, a 1% increase in VAT should result in a parallel 1% increase in costs. The same result is reached by Cavas (2013) regarding China-US import-export. Again, Yue (2015), reported the same finding for the transport industry (where the VAT rate is 11%): each percentage point increase in the VAT results in an approximately one-percentage-point increase in costs. Moreover, the results of his regression analysis indicate that the regression coefficient between enterprise performance and the VAT policy is 4.35. This represents the direct VAT multiplier for enterprise performance, which roughly confirms, mutatis mutandis, what has been indirectly obtained for manufacturing.

Du (2015), on the basis of the input-output (I-O) table and Urban Household Survey data, analysed the impact of China's VAT expansion reform on the joint income redistribution effects of VAT and BT on urban households, by comparing the Gini coefficient and general entropy indexes before and after the reform. He finds that the reform improved the redistribution effects of the two taxes by lowering the average tax burden and narrowing the income gap "within" the low-income household group. Since this reform did not deliver a considerable tax cut to the expenditure items that are particularly important for the low-income households, the income gap "between" the household groups with different income levels is nearly unaffected.

Positives and negatives—with the former being predominant—can be traced if we limit the analysis to the first period of the reform, that is, up to and including 1994. As highlighted by Gordon and Li (2002), while the 1994 fiscal innovations in the tax law, in accounting procedures, in banking, and in tariff rates, that led China to have a fiscal structure much more similar to those of developed economies, did initially generate very rapid growth, it created its own problems. As argued by Gordon and Li (2002), for example, tax competition undermined the tax base, and local governments had an incentive to protect local firms facing high tax rates, while the national government still had a competing incentive to protect large firms. While this system may not be quite as supportive of entry and growth as the previous system, it should deal well with many of the problems that arose under the previous system.

To analyse the overall effects of the tax reform up to 1994, Toh and Lin (2004) applied a computable general equilibrium (CGE) model. The results of the simulations indicate small agammaegate welfare gains. However, household groups are worse off because of the redistribution of resources from the household to the government sectors. There will be a substantial increase in government revenue, and the prudent and productive use of the increased revenue could improve household welfare. This result also suggests that the statutory rates introduced in 1994 may be too high from an equal yield perspective. Their analysis also suggested that further improvements in the tax system could be obtained by extending a consumption-type VAT to other sectors that were not included in the reform as of their writing.

A number of studies have analysed and discussed the reform in itself, including its effects on taxation and the fiscal advantages and disadvantages for specific economic sectors. For example, Zhang and Lu (2017) analysed the influence of VAT on the air transportation industry, while Wan (2016) estimated that more than 97% of tax payers will pay less tax under the VAT regime, saving a total of over 300 billion yuan, and that sectors such as manufacturing that already use VAT would also save over 300 billion yuan in tax payments due to an increased number of deductible items. The transition from the mixed to the purely VAT system has been studied, with respect to the minor tax burden on the industrial and services sectors, and the consequent development opportunities, as well with respect to the household income distribution.

The widspread finding is that the reform resulted in a reduced tax burden, except for small service entrepreneurs in Shanghai (as we will see in a moment) (Zheng and Shu, 2017, South China Morning Post, 2016). It is then evident that these positive effects transfer to the whole economy.

A comprehensive appreciation of the reform, that is, an examination of the entire period until 2016, is provided by Xinhua (2017), with a focus on services. Xinhua (2017) stressed the value of the reform because China is relying on services, particularly high-value-added services in finance and technology, to reduce the economy's traditional reliance on heavy industry and investment, extending the VAT reform across all industries will encourage the development of the service sector, support industry upgrading, stimulate consumption and support supply-side structural reform. The reform has streamlined the tax system, reduced and standardized double central and local government fiscal responsibilities and eiminated taxation barriers between manufacturing and service sectors to increase the weight of the service sector in China's economy. This view is definitely shared by the Financial Tribune (2017), which affirms that the VAT reform has given impetus to economic growth and boosted entrepreneurship. Moreover, the reform has the potential to stimulate mass innovation and create a favourable climate for private enterprises by reducing corporate burdens.

A discordant voice is represented by Ren's (2016) view that the reform, which has intended to reduce the tax payments of smaller mainland service firms, appears to have had the opposite effect on many entrepreneurs, who contend that they have become victims of the new system. As small service entrepreneurs in Shanghai claim, while before 2012, the tax was imposed on service companies based on their sales revenues, after the reform the amount levied was calculated based on the value they added to their products and services, and the pilot scheme proved unsuccessful³. As a result, two-thirds of Shanghai-based companies ended up paying more tax under the new system, as demonstrated by a survey. In stark contrast, Zhao Yang, chief economist with Nomura, Asia's global investment bank, said, "while the VAT reform could initially result in a higher tax burden owing to ineffective enforcement, appearing to have had no immediate positive impact on businesses, taking a long view, it will benefit the development of privately owned companies".

³ Under the VAT system, the value of invoices issued by suppliers of raw materials is measured against the value of the sales invoices that service companies issue to their customers.

4.2.   The economic effects of a VAT rate reduction government policy

China's government, with the aim of stimulating the economy, may implement a balanced policy consisting of reducing the VAT rates of industries that are important in terms of production volume flanked by industries with low production volume. Actually, this intervention can consist of cutting the VAT for "Manufacturing", "Construction", and "Transport, Storage, etc." by 1.5%, the VAT for "Information, Transmission, etc.", and "Financial Intermediation" by 1%, and the VAT for "Real Estate" by 0.5%. These cuts result in increased exogenous demand for the products of the respective industries that may be quantified based on the conclusions on the economic effects of the fiscal reform reported in Sub-paragraph 4.1.2, particularly those obtained by Gourdon et al. (2014) on exports. Moreover, Barrell and Weale (2009), in discussing the theoretical effects of the 2008 VAT reduction in the UK, found that a temporary 1% cut in VAT resulted in an approximately 0.95% increase in household consumption per quarter. A rough linear extension over the year would lead to an expected 3.9–4.0% increase. As a result, an average inverse multiplier of a 1% VAT rate cut/demand increase equal to 5% for China seems appropriate.

Therefore, the exogenous demand for the above industries' products will modify as follows (hundred million yuan): "Manufacturing": 229, 990; "Construction": 176, 542; "Transport, Storage, and Post": 11, 659; "Information, Transmission, etc.": 11, 473; "Financial Intermediation": 1, 713; and "Real Estate": 12, 227. The exogenous account vector adjusts accordingly (see column 3 of Table 5).

The total demand vector resulting from equation (3) due to the VAT shock is reported in column 5 of Table 5. Columns 6 and 7 report the percentage and absolute changes of total demand after the VAT cut shock.

There is evidence that a 1.5% VAT cut for "Manufacturing", "Construction", and "Transport, Storage and Post" yields increases in demand for the products of these industries of 6.4%, 7.4% and 5.9%, respectively. A 1% VAT cut for "Information, Transmission, Software, etc." and "Financial Intermediation" yields increases in demand for the products of these industries of 5.0% and 5.5%, respectively. A 0.5% VAT cut for "Real Estate" yields a 4.3% increase in the demand for its products (column 7, figures in bold).

It is worth stressing the absolute increase in "Manufacturing", which accounts for 59, 977 hundred million yuan, or 1, 874.0 billion PPP USD, as well as those of "Construction" and "Transport, Storage and Post", accounting for 12, 978 hundred million yuan, or 405.6 billion PPP USD and 4, 289 hundred million yuan, or 134.0 PPP USD, respectively (column 8).

Turning from the above industries, directly affected by the shock, the remainder of the productive sector responds, on average, to the shock with a 3.9% increase in demand, although this varies, sometimes greatly, across industries. These range from 6.2% increase for "Mining", which is very close to that of "Manufacturing", to the very small 1.7% and 1.4% observed for "Education" and "Conservancy, Environment, etc.", respectively. The overall increase for China's productive sector is 5.4%, in absolute terms equal to 108, 844 hundred million yuan, or 3, 401 billion PPP USD.

Outside the productive structure, very significant increases are recorded by the two GVA components, "Compensation of Employees", with 4.9%, and "Gross Operating Surplus", with 5.4%. Noteworthy are the percentage increasings recorded by "Household Expenditure", and for the income categories, by "Household", "Financial Corporations", and "General Government", all between 4.5% and 5.0% (column 8).

In addition, we conducted detailed complementary analysis of the absolute value increases of the six industries targeted by the fiscal intervention.

Let us begin with "Manufacturing". As shown in Table 6, where in column 22 is reported the total exogenous output, in column 23 its absolute increase, in column 24 the household expenditure and in column 25 its absolute increase, the most relevant share of its 59, 977 hundred million yuan total demand increase is represented by the 38, 991 hundred million yuan increase (65.0%) recorded by total intermediate output (see second cell of the column "Absolute increase"). If we consider the further 4940 hundred million yuan (8.2%) represented by the increase in "Household Expenditure", making the total of the endogenous share of the total demand equal to 73.2%, we can conclude that little is left to the increases in the exogenous share of total demand. A similar and more pronounced case is observed for "Transport, Storage and Post", with endogenous share accounting for 80.5%, and "Financial Intermediation", with endogeus share accounting for 98.1% and practically no share for the exogenous demand. Also for "Real Estate" the endogenous share of the total demand is very high (approximately 86%), but here approximately 60% is taken by the "Household Expenditure" increase. Conversely, we observe an entirely different behaviour for "Construction", where only 5.1% of the total increase in demand is taken by the increase in total intermediate output, and no increase is taken by "Household Expenditure", making the exogenous share absolutely dominant (approximately 95%), confirming the uniqueness of this industry.

Regarding the supply side depicted in Table 7, the large absolute value, 42, 099 hundred million yuan, of the increase in the total input of "Manufacturing", and that of "Construction", equal to 9507 hundred million yuan, are particularly notable when compared to the very low value (661 hundred million yuan) of total output.

Note the absolute value increases in the GVAs of "Manufacturing" and "Construction" (see again Table 7), at 10, 735 and 3, 050 hundred million yuan, respectively.

5.   Conclusion

In this paper, we concentrated on the analysis of China's economic system and in the measurement of VAT reform effects on the syestem itself. We used a 19 industry 2015 SAM for China, identified the exogenous and endogenous accounts and conducted an impact multiplier analysis. The industry-level endogenous coefficients provide evidence of a considerable average degree of integration among the industries, excluding a lower interdependency between the group of industries producing goods and some types of industries producing services, meaning a weak "communicability" that likely still exists in the relationships among the state-owned enterprises. "Manufacturing" stands out among the other industries, thus confirming its key role in China's economy, In terms of GVA, there is evidence that 21% of industries is high GVA producer, divided into half labour-oriented and half capital-oriented, whereas another 21% is medium GVA producer, with no clear labour or capital orientation definition. The remaining industries are medium-low GVA producers. All these results are reaffirmed, and strengthened by the evidence from the impact multipliers, which confirms the dynamism and reactivity of "Manufacturing", and by their decomposition into three components, showing that the direct and the intergroup effects concern the industry accounts only, whereas the indirect effect relates to all the other endogenous accounts. This is the economic reality that inspired the tax reform.

Afterwards, we have discussed how the longstanding efforts at reforming China's indirect tax system concluded in 2016 represented a turning point in the country's economic and social equilibrium and have demostrated that making the central government the sole operator of indirect taxes is a general advantage because it reduces the tax burden borne by enterprises.

To measure the effects of the exogenous shock represented by VAT cuts, we employed a model, with the government cutting the tax to different rates for a group of selected industries. The consequent total demand increases for these industries have a multifaceted impact on the economic system. First, there is evidence that the industries have an inversely proportionate positive reaction to the VAT cut through the mediation of the aforementioned total demand increase: the greater the cut, the greater the increase in industry total demand. For example, a 1.5% VAT cut generates increases in total demand in three crucial industries, i.e. "Manufacturing", "Construction" and "Transport, Storage and Post", in a range from approximately 6% to 7%. This is a strong response, which can be better appreciated in absolute terms: in the largest case of "Manufacturing", little less than 2000 million PPP USD, an absolutely remarkable and meaningful value, even for a large economy as big as that of China. A 1% VAT cut prompts a little smaller response from the "Information, Transmission, Software, etc." and "Financial Intermediation" industries: approximately 5.0% and 5.5%, respectively. A 0.5% VAT cut yields an increase in total demand for "Real Estate" industry of more than 4%.

Overall, there is evidence of a strong average response by the whole economic system (nearly 4%) to an average VAT cut of 1.2% for 6 industries. An analogous strong positive reaction is observed for GVA, whose labour and capital components respond on average by more than 5%, as well as by private consumption and institutions' income, for which the rates of increase are only slightly lower.

Considering the increases in absolute values for the 6 industries targeted by the exogenous shock reveals that over two-thirds of the total demand increase is due to endogenous increases, with the most relevant share taken by total intermediate output, except for the "Real Estate" industry, which exhibits a much higher share taken by final expenditure. The case of "Construction" is very peculiar, as there is evidence of an increase almost entirely due to exogenous factors.

To conclude and summarize, the picture that emerges from our analysis is that of a highly integrated economic system, in which the manufacturing sector plays a crucial role. This system is highly reactive to exogenous shocks represented by VAT cuts, meaning that the tax lever represents a powerful tool for the government to make economic policy.

Acknowledgements

This paper was prepared during the six month visit of Mi Zichuan and Ma Kewei to the Department of Statistics, Computer Science, Applications of the University of Florence held during the second half of 2017 and early 2018 and further discussed at seminars at the Department of Statistics and at the School of Statistics of Shanxi University of Finance & Economics. We thank the participants for their valuable comments that helped us to further improve the paper.

This research has been supported by the Program for the Innovative Talents of Higher Education Institutions of Shanxi, 2017. Funded by the China National Study Abroad Fund (Local Cooperation Project: 201808140226).

Conflict of interest

The authors declare no conflict of interest in this paper.