Value-added-tax rate increases: A comparative study using difference-in-difference with an ARIMA modeling approach

The paper investigates whether Saudi Arabia’s July 2020 VAT rate increase from 5% to 15% affected key financial indicators of non-financial firms listed on the Saudi stock exchange (Tadawul). Motivated by Saudi Vision 2030’s diversification agenda and the limited empirical evidence on VAT impacts in GCC economies, the study addresses the research question: Does imposing a 15% VAT affect firms’ total assets, equity, liabilities and equity, total income, total revenue, total expenses, net income, changes in operating, investing, and financing activities, and cash at period end? The context includes VAT’s global diffusion, debates about its regressivity/progressivity, and unique GCC tax structures. The study aims to fill a gap by using disaggregated firm-level data, a DiD design enhanced with ARIMA time-series modeling, and sectoral comparisons across two periods (2019 vs. 2020) to isolate short-run impacts amid COVID-19.

The review covers foundational principles of tax progressivity (Smith, 1776) and concerns about VAT regressivity, often mitigated by exemptions or zero-rating essentials (Cnossen, 1992; Alavuotunki et al., 2019). It surveys global VAT adoption drivers, including policy diffusion via neighbors, EU accession requirements, and IMF programs (Keen & Lockwood, 2010; Ufier, 2014, 2017; Čížek et al., 2017). Studies report mixed effects of VAT on efficiency and distribution; while general equilibrium analyses suggest potential efficiency gains, real-world design features (exemptions, multiple rates) can erode benefits (Ballard et al., 1987; Bye et al., 2012; Ormaechea & Morozumi, 2019). Empirical evidence shows VAT reforms can influence exports, innovation, and compliance (Gourdon et al., 2022; Cao et al., 2022; Irawati et al., 2022; Kim et al., 2022). In Saudi Arabia and GCC, low direct taxes and cautious indirect tax introduction aim to avoid regressive outcomes and investor deterrence (Ezenagu, 2021). Prior Saudi-focused work at 5% VAT indicates modest balance sheet and income statement changes in banks (Alhussain, 2020) and broader socioeconomic effects (Bogari, 2020). The review underscores that VAT’s growth and distributional impacts hinge on design details, enforcement capacity, and country context.

Design: A comparative sectoral analysis combined with a Difference-in-Differences (DiD) framework augmented by time-series modeling (ARIMA) to address endogeneity and serial correlation. - Sample and periods: Non-financial firms listed on Tadawul (192 companies across 11 sectors; financial firms excluded). Periods include pre-increase quarters (Q2–Q3 2019) and post-increase quarters (Q3–Q4 2020), corresponding to pre-15% VAT (and pre-COVID-19 discovery) and post-15% VAT (during COVID-19). - Data sources: Quarterly financial statements publicly available via Tadawul. - Outcomes/controls: 11 firm-level indicators grouped by statement: Balance sheet (Total Assets, Shareholders’ Equity, Total Liabilities and Shareholders’ Equity), Income statement (Total Income, Total Revenues, Total Expenses, Net Income), Cash flow (Other Changes in Operating Activities, Other Changes in Investing Activities, Other Changes in Financing Activities, Cash at End of Period). - DiD specification: The model includes pre/post period indicators and their interactions with the listed financial metrics to capture differential changes after the policy. The paper presents a detailed linear specification including main and interaction terms for each indicator. - Time-series approach: To mitigate endogeneity and handle time dependence, the study estimated ARIMA(p,d,q) models on first-differenced logged series: • Stationarity tests: Augmented Dickey-Fuller (ADF) test indicates stationarity after first differencing (Z(t) ≈ −7.003, p < 0.001), supporting d = 1. • Model selection: Based on autocorrelation/partial autocorrelation diagnostics and AIC/BIC, ARIMA(1,1,0) was selected as the preferred model. Estimation via maximum likelihood with semi-robust standard errors; missing data handled via Kalman methods. • Diagnostics: Autocorrelation and partial autocorrelation functions suggest limited residual autocorrelation; a general C–H test indicates acceptable serial correlation properties across lags (e.g., at lag 1, χ2 ≈ 2.439, p ≈ 0.1183). - Sectoral comparative analysis: Means and standard deviations of key indicators compared pre/post across sectors (e.g., Energy, Materials, Capital Goods, Commercial & Professional Services, Transportation), with percentage change summaries. - Notes: Financial sector excluded due to distinct transactions; the design isolates the VAT change coinciding with COVID-19 but acknowledges potential confounding.

- Overall short-run impact (2019 vs. 2020): • Broad declines in firm liquidity and activity: Cash at end of period (CEP) average fell from ≈ 2,092,480 (2019) to ≈ 698,409 (2020); Total Income (TI) declined from ≈ 1,813,028 to ≈ 331,292; Total Revenues (TR) also decreased substantially post-increase. • Expenses: Total Expenses (TE) generally increased in 2020, consistent with higher VAT and crisis-related costs. • Equity: Shareholders’ Equity declined overall, reflecting stress from COVID-19 and the VAT hike. • Profitability: A reported average profitability decrease of about −2.16% associated with the 10 percentage-point VAT increase (from 5% to 15%), alongside more inactive companies and higher bankruptcy risk. - Sector-specific highlights: • Energy: Declines in CEP (≈ 939,278 to ≈ 528,862), TI (≈ 1,000,091 to ≈ 643,458), and TR (≈ 1,082,161 to ≈ 661,921); slight reductions in TA and TLSE; increased TE and operational activity changes. • Materials: Short-term drops in cash and TLSE; TR increased slightly on average, but NI fell markedly, indicating demand contraction effects. • Capital Goods: TA and TLSE roughly stable (−1%); exceptionally large increase in NI (reported ≈ +4,865%), with increases in COA, CIA, and CFA (≈ +84%, +129%, +39%), suggesting sector-specific demand/resilience in 2020. • Commercial & Professional Services: Most indicators slightly down in 2020; CEP notably weaker; firms faced staffing and operational challenges. • Transportation: Increases in TA and TLSE but decreases in NI and CEP; sector heavily impacted by global travel collapse (e.g., passenger load factor down to ≈ 65.1% in 2020 from 82.5% in 2019; Middle East passenger revenue per km down ≈ 71.6%). - Time-series/ARIMA-DiD results: • ADF: Strong stationarity after first differencing (p < 0.001). • ARIMA(1,1,0): Preferred by AIC/BIC; autoregressive term significant (e.g., ar ≈ −0.559, p < 0.001); Wald tests indicate overall model significance. • Many pre/post coefficients on differenced logged indicators show significant opposite-signed effects pre vs. post, consistent with a structural break after VAT increase/COVID-19. - Interpretation: The VAT rate increase, amid COVID-19, significantly affected firms’ financials with heterogeneous sectoral responses. Short-term negative effects on profitability and liquidity dominate, alongside increased inter-industry volatility.

The findings indicate that raising VAT in Saudi Arabia during COVID-19 led to significant short-run declines in liquidity, revenues, and profitability for many non-financial firms, with substantial inter-industry heterogeneity. This aligns with literature showing that while VAT can improve efficiency in theory, real-world design elements—exemptions, multiple rates, and incomplete input crediting—can distort production decisions and impair growth (Crawford et al., 2010; Keen, 2013; Ormaechea & Morozumi, 2019). Short-run negative output effects of VAT rate increases are consistent with macro evidence (Dabla-Norris & Lima, 2018). The study’s sectoral patterns suggest that demand shocks, supply-chain disruptions, and sector-specific policy contexts (e.g., transport restrictions) interacted with the VAT change to shape outcomes. While some sectors (Capital Goods) showed resilience or gains in specific metrics, others (Energy, Materials, Transportation) experienced notable stress. The results support careful VAT design (broader bases, fewer exemptions, simpler rate structures) and complementary policies to buffer vulnerable sectors and households. Over the long run, increased government revenues from VAT could support public investment that partially offsets short-run firm-level losses; however, without thoughtful design and enforcement, VAT can exacerbate regressivity and hinder firm performance, especially in developing contexts with sizable informal sectors and limited administrative capacity.

The study provides evidence that Saudi Arabia’s 2020 VAT rate increase from 5% to 15% had significant short-run impacts on non-financial listed firms, including declines in cash, revenues, and profitability, and increased volatility across sectors. The average profitability decrease (≈ −2.16%) and indications of higher inactivity and bankruptcy risk underscore near-term corporate stress, compounded by COVID-19. Policy implications include: refine VAT design to minimize distortions (reduce exemptions, simplify rates), consider targeted mitigation for vulnerable sectors, and sequence reforms to support growth and employment objectives. In the longer term, VAT revenues may finance investments that bolster productivity and equity, but careful implementation is critical. Future research should expand to multiple taxes, broader samples, and longer horizons, explore governance-taxation linkages in developing contexts, and apply complementary methods (e.g., meta-analysis, probability plots) to strengthen causal inference.

- The VAT rate increase coincided with COVID-19, making full causal separation challenging; unobserved confounders (“abandoned reasons”) may bias estimates. - Short observation window (2019 vs. 2020) limits long-run inference. - Focus on non-financial listed firms in Saudi Arabia; results may not generalize to financial firms, SMEs, or other countries. - Only one tax instrument (VAT) is examined; broader fiscal interactions are not modeled. - Aggregation and data limitations (e.g., sectoral heterogeneity, potential measurement inconsistencies) may affect precision. - Model specification choices (ARIMA(1,1,0), DiD structure) and dropped variables due to collinearity constrain interpretability of some coefficients.