Accounting Information and Stock Price Dynamics in Saudi Arabia: Evidence from a Markov-Switching Regression Model

1. INTRODUCTION

Stock prices are shaped by many factors, one of which is the accounting information reported in financial statements¹ and that recent evidence highlights that stock markets in the gulf cooperation council (GCC) region exhibit dynamic and nonlinear behaviour influenced by macroeconomic and financial variables.² Financial reports, commonly referred to as financial statements, are official documents that summarise the financial transactions and activities of an organisation, individual, or other entity. The financial statements of a company enterprise contain all important financial information provided in a systematic and easily understood style. Their primary objective is to supply stakeholders with relevant information regarding a firm’s financial condition, operating results, and variations in its financial standing over time. However, financial institutions evaluate financial statements to assess whether to provide a business with working capital or security (such long-term bank loans or debentures). The quality of accounting information plays a critical role in enhancing transparency and decision-making, which ultimately affects financial markets.³ Accounting systems and cost measurement approaches significantly influence financial decision-making and pricing efficiency.⁴ However, while for government, it uses financial statements to verify the legitimacy and accuracy of taxes and other fees a business pays. Whereas, owners and management of a corporation, use financial statements to make crucial decisions in their operations. At the same time, employees use financial statements to debate their pay, bonuses, promotions, and rankings. More so, a prospective investor could use financial statements to evaluate the possibility of investing in a business, including purchasing shares, debentures, etc. When an investor is faced with deciding whether or not to buy a particular stock, they must analyse the stock accounting information included on the financial statements of the company’s business over time. This will allow them to determine whether or not profits are increasing or decreasing, as well as what this means for the company’s potential future performance. However, because it can be difficult to obtain complete and accurate information, any available financing may be allocated to the incorrect businesses or industries. Despite reporting enormous amounts of revenue on their annual financial statements, it is perplexing that several companies’ stock values do not continue to increase. The question here is that, is there a connection between the accounting information on financial statements and stock prices?

Researchers^5-10 concluded that accounting information affects stock prices. However, according to the findings of a large number of researches, the effect of accounting information on stock prices varies significantly from country to country^11-15 and despite extensive global evidence, there is limited research examining nonlinear dynamics between accounting information and stock prices in Saudi Arabia using advanced econometric techniques. Again, according to the existing literature, only a handful of studies^16,17 have looked into the influence that accounting information on financial statements can have on the values of stocks traded on the Saudi Arabian stock market. Also, many existing studies, not only in Saudi Arabia, lack the application of modern or advanced modelling techniques that could provide comprehensive information. Therefore, even though the connection between the accounting information on financial statements as well as stock prices has been the subject of extensive critical analysis in previous researches, this study provides a fresh perspective on the topic while also making reference to the stock market prices in Saudi Arabia considering that the country’s Capital Market Authority and several other financial institutions, such as the Saudi Arabian Monetary Authority, took various actions to further develop the financial sector in the recent years; while making a number of innovative contributions. Additionally, this study utilised a cutting-edge and an advanced technique called Markov-Switching Dynamic Regression (MSDR) model due to the nature of the study’s objectives which to the best of my knowledge, this is the pioneering study to use such a sophisticated technique on the Saudi Arabian stock prices that is capable of answering various questions that are relevant to financial/investment decisions.

Specifically, this study aims to examine whether the relationship between accounting information on financial statements and Saudi Arabian stock prices is linear or nonlinear; to investigate the impact of accounting information on the financial statements of Saudi Arabian stock prices; to analyse the regime in which the stock prices are more stable or rather the regime that the market spends more time on; to determine the probability of the Saudi Arabian stock prices switching from a low to a high regime and from a high to a low regime; and to historically explore how long the stock prices stayed in both low and high regimes.

The study’s findings revealed that the link between accounting information on financial statements as well as Saudi Arabian stock prices is non-linear. Moreover, accounting information on financial statements, commonly referred to stock market return, value of shares traded, and market capitalisation are influencing Saudi Arabian stock prices, but volume of shares traded retards it while stock market turnover has no impact. Furthermore, the stock prices are more stable when the stock prices are low, than when the stock prices are high. In addition, when the market is in the low regime, there is a 99% likelihood of transitioning to the high-state regime, but the probability of shifting from the high regime to the low regime stands at 90%. More so, 19 years or 67.86% of the years studied, the stock prices are in the low regime, and 9 years or 32.14% of the years studied, the stock prices are in the high regime. The remaining parts of this study are split up into four sections: section two, which is devoted to a literature review; section three, which discusses the methodology that will be used to achieve the study’s objectives; section four, reports and interprets the empirical results; and section five, which provides a summary of the study as well as some suggestions derived from the findings.

2. MATERIAL AND METHODS

2.1. Data and Its sources

This study made use of the Saudi Arabian stock prices accounting information on financial statements, which are usually reported annually, namely stock market return (%, year-on-year) (SMR), stock market turnover ratio (%) (SMT), the value of shares traded (billion SAR) (VST), number of shares traded (million) (NST), the market capitalisation of issued shares (billion SAR) (MCAP), and share price index (SPI) to investigate the influence of accounting information on financial statements of the Saudi Arabian stock prices. However, the data range from 1994 to 2021 and this range is based on the data availability. The measurement of variables follows prior studies: stock market return (Filip & Raffournier, 2010), turnover ratio (World Bank, 2023), market capitalisation (Habib, 2004), and trading variables (Gul & Javed, 2009). The data sourced from World Bank Development Indicators (2021) statistical bulletin and Saudi Central Bank annual report on the Saudi Capital Market (Tadawul) (2021) statistical bulletin. However, Figure 1 reports the study’s conceptual framework.

Close

Figure 1: Conceptual framework.

Export to PPT

2.2. Estimation techniques

The estimation techniques stated with the graphical representation of the variables for understanding the trend of the variables; descriptive statistics for unveiling the statistical characteristics of the variables; unit root tests, namely Augmented Dickey-Fuller (ADF), Phillips Perron (PP), and Dickey-Fuller Generalised Least Squares (DF-GLS) unit root tests for verifying the stochastic properties of the variables and the essence for employing up to three different unit root tests is to ensure that the acceptance or rejection of the null hypothesis of unit root is not ambiguous; linearity test using the Ramsey Reset test (RESET) for linearity, Brock-Dechert-Scheinkman-LeBaron (BDSL) test for linearity, and Incremental F-test for linearity where the essence for applying these tests is to ascertain whether nonlinear model could be used to estimate the relationship and the essence for employing up to three different linearity tests is to ensure that the acceptance or rejection of the null hypothesis of linearity is not ambiguous; estimation of information criteria to test for the number of regime to use in MSDR model estimation; estimation of MSDR model; post-estimation tests to diagnose for the statistical healthiness of estimated model including serial correlation test, heteroscedasticty test, normality test, and Ramsey Reset test to check whether the estimate model requires any more nonlinear combination; and lastly is the estimation of the direction of causality of the variables involved in the analysis. The reason for choosing MSDR model is strictly based on the need of our objectives. In fact, according to the nature of this study’s objectives, no other estimation techniques can answer all the research questions/objectives except the MSDR model; hence, its choice over other options. The MSDR model is an advanced cutting-edge technique. In addition, according to the author’s understanding, this may be the initial research to use such a technique on the Saudi Arabian stock prices analysis despite its ability to provide answers to the questions whom their answers are crucial or garment to investors and policymakers. Moreover, the MSDR model is particularly suitable for identifying regime-dependent dynamics, including differences in market stability and duration across regimes. Endogeneity concerns are mitigated by the use of the MSDR model, which captures regime-dependent dynamics and reduces bias arising from structural changes over time.

The MSDR model is a nonlinear model that captures more intricate dynamic patterns by enabling switching between each regime. Both observable and unobservable state variables control the Markov Switching Model’s switching mechanism. The benefit of the MSDR model is that it can give statistical approaches by systematically pulling past regime shift information of the data. Additionally, this model can estimate model parameters with consistency and efficiency, and identify recent changes.³¹ Let’s consider a general MSDR model specification comprising a dependent variable indexed by time y, a matrix of regression variables X, a fitted vector of coefficients β_cap, and residual errors. Then, let y represent a n $\times$ 1 matrix of ‘n’ time-indexed observations y_t:

$\left[\begin{array}{c}{y}_1\\ y_2\\ .\\ \begin{array}{c}.\\ .\\ y_n\end{array}\end{array}\right]\begin{array}{c}t=1\\ t=2\\ .\\ \begin{array}{c}.\\ .\\ t=n\end{array}\end{array}$

Now for the regression variables matrix, let X be a [n $\times$ (m + 1)] matrix of regression variables in which the first column contains the intercept value x_t is a column in this matrix. $\begin{array}{c}t=1\\ t=2\\ .\\ \begin{array}{c}.\\ .\\ t=n\end{array}\end{array}\left[\begin{array}{c}1\\ 1\\ .\\ \begin{array}{c}.\\ .\\ 1\end{array}\end{array}\begin{array}{c}{\Psi }_{11}\\ {\Psi }_{21}\\ .\\ \begin{array}{c}.\\ .\\ {\Psi }_{n_1}\end{array}\end{array}\begin{array}{c}{\Psi }_{12}\\ {\Psi }_{22}\\ .\\ \begin{array}{c}.\\ .\\ {\Psi }_{n_2}\end{array}\end{array}\begin{array}{c}{\Psi }_{13}\\ {\Psi }_{23}\\ .\\ \begin{array}{c}.\\ .\\ {\Psi }_{n_3}\end{array}\end{array}\begin{array}{c}{\Psi }_{1m}\\ {\Psi }_{2m}\\ .\\ \begin{array}{c}.\\ .\\ {\Psi }_{mn}\end{array}\end{array}\right]$

Then, for the fitted regression coefficients β_cap, let β_cap be an [(m+1) X 1] vector of regression coefficients. The fitted value of the coefficient obtained is indicated by the ‘cap’ on β. $\left[\begin{array}{c}{\stackrel{⌢}{\beta }}_0\\ {\stackrel{⌢}{\beta }}_1\\ .\\ \begin{array}{c}.\\ .\\ {\stackrel{⌢}{\beta }}_m\end{array}\end{array}\right]$

Now, let’s have a look at the following general specification for a time series model that has an additive error component:

(1)

$y_t={\stackrel{⌢}{\mu }}_t+e_t$

The observed value y_t in the model above results from adding the anticipated value μ_cap_t and the residual error ε_t. This is how the model predicts the value that will be observed in the future. Let’s assume that ε_t is a random variable with a normal distribution that is homoskedastic (has a constant variance) and has a mean of zero, denoted by the expression N(0, σ²). The value of μ_cap_t is the result of applying the regression function η(.) in such a way that μ_cap_t = η(x_t, β_cap). Taking into account the Markov model, let us now explore how the k-state Markov process might influence the regression model that was discussed earlier in the following manner:

(2)

$y_t={\stackrel{⌢}{\mu }}_t+e_t\text{when}{S}_t=j$

where, as was the case before, the expected value is a function of the two variables x_t and β_cap: μ_cap_t_j = η(x_t, β_cap_j). However, this time, note that the vector containing the regression coefficients has been given the name β_cap_j, which corresponds to the j^th Markov state. In general, the regression coefficients are a matrix with the size [(m+1) X k] if the Markov model is operating over ‘k’ states [1, 2, …, j,…, k]. The following is an example of how the matrix is structured: ${\stackrel{⌢}{B}}_s=\stackrel{\begin{array}{c}CoefficientCoefficient\\ ForRegime1ForRegime\text{k}\end{array}}{\left[\stackrel{}{\begin{array}{cc}{\stackrel{⌢}{\beta }}_{01}\mathrm{...}& {\stackrel{⌢}{\beta }}_{0k}\\ \begin{array}{c}.\\ .\\ .\\ {\stackrel{⌢}{\beta }}_{m1}\mathrm{...}\end{array}& \begin{array}{c}.\\ .\\ .\\ {\stackrel{⌢}{\beta }}_{mk}\end{array}\end{array}}\right]}$

The reasoning behind this is that the coefficients of the regression model will transition to the adequate regime-specific vector β_cap_j from β_cap_s based on the current Markov state, also known as a “regime,” which is j in [1, 2,…,k]. The idea behind this is that the coefficients of the regression model will shift. Markov Switching Dynamic Regression model was conceived as a consequence of this. In addition to this, the k-state Markov process is itself controlled by the following variable for the state transition: $P\left(s_t=1\left|s_{t-1}\right.=1=P^{11}\right)$, $P\left(s_t=2\left|s_{t-1}\right.=1=P^{12}\right)$, $P\left(s_t=1\left|s_{t-1}\right.=2=P^{21}\right)$, and $P(s_t=1\left|s_{t-1}\right.=2=P^{22})$. However, it is often presented as: $P_{ij}=\left(\begin{array}{cc}{P}_{11}& P_{12}\\ P_{21}& P_{22}\end{array}\right)$. The restrictions on these transition probabilities ensure P₁₁ + P₁₂ = 1, and P₂₁ + P₂₂ = 1. See Figure 2: Flowchart of the analysis

Close

Figure 2: Flowchart of the analysis.

Export to PPT

3. RESULTS

Figure 3 is the graphical representation of namely SMR (billion SAR), SMT (billion SAR), VST (billion SAR), NST (million), MCAP (billion SAR), SPI of the Saudi stock exchange market from the period of 1994 to 2021. From the figure, all the variables show a fluctuating trend over the time horizon; however, they all increased in recent years except for the number of shares traded, which decreased in 2021. Though the stock market return fluctuated throughout, it experienced a boom in 2005 and the worst decline from 2007 to 2009. The stock market turnover remains low from 1994 up to 2002, which is its worst decline for the whole period, then rises from 2003 to 2006, which is its peak throughout the period. After that, it rushes to decline up to the year 2019, where it gradually continues to pick up. The value of shares traded is very low right from 1994 up to 2002, which is its worst decline over the period, then rises to 2006, which is its peak over the period, after that it rushes down up to the year 2010 after then it moderately fluctuates up to the year 2019 and then rapidly continues to rise. The amount of shares traded relatively shares the same trend as the value of shares traded except that the former has decreased recently. The market capitalisation, just like the value of shares traded, stock market turnover, as well as the number of shares traded, remains low right from 1994 up to 2002, which is its worst decline over time under consideration, then rises a little to 2005 and fluctuated moderately up to the year 2008, then remains partially constant up to the year 2018 after then it rapidly continues to increase. The share price index, just like the value of shares traded, stock market turnover, number of shares traded, as well as market capitalisation; remains low right from 1994 up to 2002, which is its worst decline for the whole time horizon, then rises rapidly to 2005 which is its peak over the time period, then it ruches down to 2006 and fluctuates up to the year 2008, then considerably continues to rise.

Close

Figure 3: Graphical representations of the variables. Source: Author’s computation note: SMR: Stock market return (%, year-on-year), SMT: Stock market turnover ratio (%), VST: Value of shares traded (billion SAR), NST: Number of shares traded (million), MCAP: Market capitalisation of issued shares (billion SAR), and SPI: Share price index.

Export to PPT

Table 2 displays the statistical attributes of the variables under investigation. Form the table, the means and standard deviations of SMR, SMT, VST, NST, MCAP, and SPI in the Saudi stock exchange market are 11.1, 83.6, 1252.3, 35824.4, 1923.9, 6081.6, and 29.7, 86.8, 1277.1, 31463.1, 2720, 3697.9, respectively. Therefore, for VST, NST, MCAP, and SPI, the mean values exceed the standard deviations, indicating that the data are relatively concentrated within the countries. In contrast, for SMR and SMT, the standard deviations are larger, suggesting greater dispersion in the data. The countries’ maximum and minimum statistics of SMR, SMT, VST, NST, MCAP, and SPI are 111.2, 429.2, 5261.9, 96095.9, 10009.2, 16712.6 and -37.0, 9.8, 23.2, 117, 145.0, 1282.9, respectively. The skewness indicates that all variables exhibit positive skew. However, the kurtosis of all the variables is low. Thus, there is less variation among the elements of each variable. Furthermore, the p-value of the Jarque-Bera statistic is substantial at a 1% level for all the variables except that of NST and SPI, which is insignificant, which means that the distribution of all the elements of each of the variables is normal except that of NST and SPI which seems to be volatile.

The findings of the unit root tests of the evaluated variables are depicted in Tables 3a and b, respectively, and are broken down by level and initial difference. According to Table 3a, both the PP and ADF tests reject the null hypothesis of a unit root in the case of SMR and SMT. For the former, the ADF and PP tests reject at 1% levels, whereas for the latter, the ADF test rejects at 10% levels, and the PP test rejects at 1% levels. The ADF and the DF-GLS fail to support the null hypothesis of a unit root for the VST at the 10% and 5% significance levels, respectively. In addition, the ADF concludes that the null hypothesis of a unit root on SPI cannot be supported at the 1% level. However, the results of all three-unit root tests indicate a unit root existing in the series of NST and MCAP. This means that the unit root hypothesis has been accepted. According to Table 3b, the series are all considered stationary after the first difference. Because of this, SMR, SMT, VST, and SPI are all considered to be stationary at level, however, NST and MCAP are considered to be stationary at first difference.

The findings of the Ramsey Specification Error Test (RESET), which was used to determine whether or not a nonlinear component was present in the relationship, are presented in Table 4. The hypothesis that was examined using this test was as follows: H0: ϵ ∼ N (0, σ² I) (the relationship is specified correctly, i.e., it does not require nonlinear combinations to estimate the response variable against that it does require a nonlinear combination to estimate the response variables) (the relationship is specified correctly, i.e. it does not require nonlinear combinations to estimate the response variable against that it does require a nonlinear combination to estimate the response variables). According to the table, all three tests have a significant p-value at 5% level or better, which is evidence of rejecting the null hypothesis that the relationship is specified correctly, i.e., linear relationship. Hence, the relationship requires a nonlinear combination to estimate the response variable, thus paving the way to employ a nonlinear model to estimate the relationship.

The results of the BDSL test for linearity are illustrated in Table 5. This test contrasts the null hypothesis, which states that the dependent variable is linearly dependent on the independent variables, to the alternative, which states that the dependent variable is not linearly dependent on the independent variables. Table 5 presents the findings of this test. The null hypothesis that the dependent variable is linearly dependent on the independent variables was rejected, and this indicates that the relationship is nonlinear, as shown by the table; when considering the significance of the bootstrap p-values, regardless of the dimension, it is possible to observe that the dependent variable is not linearly dependent on the independent variables, which indicates that the relationship is nonlinear.

Table 6 displays the results of the Incremental F-test for linearity, which tests the null hypothesis that there is no need for polynomial terms in the relationship against the alternative hypothesis that at least one polynomial term should be in the relationship. Considering the significance of the p-value in most blocks, the null hypothesis is rejected, showing that at least one polynomial term should be in the relationship.

Therefore, the three linearity tests employed, namely the Ramsey Reset test for linearity, Brock-Dechert-Scheinkman-LeBaron test for linearity, and Incremental F-test for linearity, suggest that the connection between accounting information on financial statements considered and the Saudi Arabian stock prices is nonlinear.

Table 7 reports the linearity test of the N-Regime Markov Switching Dynamic Regression model, where first, the linear likelihood ratio (LR) test is conducted in order to test whether two or more regime-switching models can fit the relationship. The Chi-square statistic of the LR test has a significant p-value at a 1% level for two regime switching. This propose that the null hypothesis of no regime switching is rejected in favour of the presence of two regimes. Therefore, two-state regime Markov Switching Dynamic Regression model is supported by the connection between accounting information on financial statements considered and the Saudi Arabian stock prices.

Table 8 shows the estimate of the Markov Switching Dynamic Regression model for the influence of accounting information on financial statements, namely SMR (billion SAR), SMT (billion SAR), VST (billion SAR), NST (million), MCAP (billion SAR) of the Saudi Arabian stock prices. From the table, SMR is positively and substantially pertinent to the share price index, where a 1% increase in SMR will induce the share price index to rise by 4.1. VST is positively and substantially pertinent to the share price index, where a 1 billion SAR increase in VST will increase the share price index by 2.7. NST is negatively and significantly related to the share price index, where a 1 million increase in NST will lead the share price index to increase by 0.2, MCAP is positively and substantially pertinent to the share price index, where a 1 billion SAR increase in MCAP will lead the share price index to increase by 0.4; however, SMT has no substantial influence on the share price index.

This suggests that accounting information on financial statements, particularly the stock market return, value of shares traded, and market capitalisation of issued shares, influence Saudi Arabian stock prices. On the other hand, the number of shares traded drag it down, while the stock market turnover ratio has no influence on it. In addition, the computed coefficients of the regime-switching models show that the projected annual increments in stock prices are greater in the low regime (regime 1) (16.6) than they are in the high regime (regime 2) of the stock market (9.5). According to these findings, regime 1, also known as the low or tranquil regime, is more reliable, and the market spends a greater proportion of its time in this regime than in regime 2 (high regime).

Table 9 shows the transition probabilities, which determine the probability that the Saudi Arabian stock prices will switch from low to high regime and from high to low regime. Following the table, the probability of the market switching from the low to the high regime is 99% and 0.01% from the high to low regime. However, when the market is in the high regime, the probability of switching to a low regime is 92% and 8% from a low to a high regime.

Table 10 illustrates regime classification relies on smoothed probabilities which show that over the period considered in the study, the Saudi Arabian stock prices enjoyed stability for a period of 19 years or 67.77% of the time in the low regime, while in the high regime it instability for a period of 9 years or 32.54% of the in the high regime.

Figure 4 shows smoothed probabilities of the Saudi Arabian stock prices. From the figure, a 1-sep prediction revealed that the model fitted the stock prices (SPI); that is to say, it accurately predicted it using the accounting information on financial statements pertinent to the stock prices. Moreover, the distribution of the scaled residuals of the 1-step prediction possesses normality. Furthermore, in the low regime (regime 0 according to the graph), the filtering process is high from the being up to around the year 2002 and low afterwards. In regime high regime (regime 1 according to the graph), filtered is in the opposite direction from the being and low afterwards.

Close

Figure 4: Smoothed probabilities of the Saudi Arabian stock prices. Source: Author’s computation, SPI: Share price index.

Export to PPT

Table 11 displays the diagnostic tests of the estimated Markov Switching Dynamic Regression model for testing the model’s statistical healthiness, including autocorrelation, heteroscedasticity, normality, and Ramsey Reset tests. From the table, the test statistics for all the tests show their p-values to be statistically insignificant; thus, each of the three tests respectively rejected the null hypothesis of no serial connection, no heteroscedasticity, normality of the residuals of the estimated model, and the model specified correctly that it does not require any more nonlinear combination to estimate the response variables.

Table 12 shows the causality test result to identify the direction of the causality across the variables in the analysis. From the table, according to the p-values of SMR, Granger cause SPI and vice versa; SMR is an important determinant of SPI but not otherwise; thus, unidirectional causality goes from SMR to SPI. For the p-values of SMT, Granger cause SPI and vice versa; both the two are Granger causing each other; thus, bidirectional causality. For the p-values of VST, Granger cause SPI and vice versa; both are Granger causing each other; hence, bidirectional causality. For the p-values of NST, Granger cause SPI and vice versa; both are Granger causing each other, which means bidirectional causality. For the p-values of MCAP, Granger cause SPI and vice versa; both are Granger causing each other; therefore, bidirectional causality. However, for the overall causality, the p-values of all the variables are Granger-causing SPI is true, so SMR, SMT, VST, NST, and MCAP are all Granger cause SPI. This shows that SMR is an important determinant of SPI, SMT and SPI are important determinants of each other, VST and SPI are important determinants of each other, NST and SPI are important determinants of each other, MCAP and SPI are important determinants of each other, and jointly the variables SMR, SMT, VST, NST, and MCAP are an important determinant of SPI.

4. DISCUSSION

The finding of this study is consistent with that of¹⁶ on the stock market prices of the Saudi Stock Exchange that, like any other stock exchange, the Saudi Arabian stock market reacts to changes in market information, which are then directly reflected in stock prices;¹⁰ on the Saudi Stock Market from the viewpoint of investors in Najran City that accounting information occurred to be the most impactful aspect on Saudi Stock Market stock prices;³² with respect to the market capitalisation that investors can use it to determine the relative size of some companies. Market capitalisation represents the price investors are willing to pay for a company’s stock and serves as an indicator of the company’s overall market value and its perceived future potential.³³ The volume of shares traded significantly affects this valuation. When a company issues additional shares, the total number of shares available in the market increases, which can dilute existing shareholders’ stakes by reducing the value of their current holdings. In the Saudi stock market, trading activity has been found to impact the volatility of stock prices, illustrating the influence of market dynamics on share value. Furthermore, shares are subject to the same fundamental economic principles as other commodities: supply and demand determine their price, and consequently, the value of each share is inversely related to the total shares outstanding.²³

5. CONCLUSION

This study has examined the impact of Saudi stock market accounting information on financial statements by using updated data and apprehending some of the uncaptured or inadequately captured parameters in order to obtain new insight; in fact, it includes stock market return, the value of shares traded, stock market turnover, the market capitalisation of issued shares, and the number of shares traded by employing the technique of the Markov-Switching Dynamic Regression (MSDR) model. The findings revealed that the connection between accounting information presented in financial statements and stock prices in Saudi Arabia is nonlinear. Furthermore, accounting information on financial statements, namely the value of shares traded, stock market return, as well as market capitalisation of issued shares, influence Saudi Arabian stock prices. However, the volume of shares traded retard it, while the stock market turnover ratio has no impact. Also, regime 1 (the low regime) is the period in which the stock prices are more stable or the regime that the market spends more time on than in regime 2 (the high regime). However, when the market is in a low regime, the probability of switching to a high regime is 99%, but when the market is in a high regime, the probability of switching to a low regime is 90%. Moreover, in 19 years or 67.86% of the years studied, the stock prices are in the low regime, and 9 years or 32.14% of the years studied, the stock prices are in the high regime. Additionally, this shows that SMR is an important determinant of SPI, SMT and SPI are important determinants of each other, VST and SPI are important determinants of each other; NST and SPI are important determinants of each other, MCAP and SPI are important determinants of each other, and jointly the variables SMR, SMT, VST, NST, and MCAP are an important determinant of SPI. It is anticipated that the findings of this study will be helpful to policymakers in reshaping the market benefits, owners and managers in making important business decisions, employees in negotiating collective bargaining agreements with the management, prospective investors in assessing, financial institutions (banks and other lending companies) in deciding whether to issue a company and government entities (tax authorities) in ascertaining the propriety and accuracy of taxes and omissions. The following recommendations were offered:

Given the nonlinear relationship identified, policymakers should adopt flexible and adaptive strategies rather than linear forecasting approaches.

Authorities or investors in this market can rely on accounting information on financial statements, including the stock market return, the value of shares traded, the market capitalisation of issued shares for the market return maximisation and market loss minimisation.

Investors may consider allocating funds to Saudi Arabia–listed equities and initial public offerings (IPOs). The Saudi Equity Strategy aims to achieve long-term capital growth by investing in these equities and IPOs. To accomplish this goal, the strategy employs a combination of top-down sector allocation informed by macroeconomic analysis and bottom-up stock selection, supported by rigorous risk management practices.

Developing future expected returns for the Saudi Arabian equity markets requires a substantial amount of clean, reliable data, along with the effective application of quantitative techniques.

Investors should stay informed about the investment outlook for Saudi Arabia by downloading the guide to investing in Saudi Arabia 2023.

However, as a suggestion, subsequent studies could be conducted across global stock markets to obtain a clearer insight of the accounting information on financial statements of the stock market. Moreover, incoming studies should use advanced/sophisticated techniques that could provide expedient information for policymakers and investment decisions.

This study is limited by the use of annual data; future studies may use other scales of data once available and extend the dataset, as well as apply alternative nonlinear techniques for robustness. Additionally, this study does not include control variables (e.g., firm size, leverage, inflation, interest rates, or industry effects), which may influence stock prices. Future research should incorporate relevant control variables to reduce omitted variable bias and strengthen the validity of the findings.