PREDICTIVE ANALYSIS OF STOCK MARKET PERFORMANCE

Background

The Nigerian capital market, anchored by Nigerian Exchange Group (NGX), serves as critical channel for capital formation and investment within Nigerian economy, providing platform through which companies raise long-term capital and investors allocate savings towards productive activity. Performance of NGX All-Share Index, market's primary benchmark, reflects aggregate performance of listed equities and is closely watched by investors, policymakers, financial commentators as barometer of overall market and, to some degree, broader economic sentiment.

Predicting stock market performance has long attracted both academic and practitioner interest, motivated by substantial financial rewards potentially available to those who can reliably anticipate future price movements. Statistical time series methods, including Box-Jenkins ARIMA framework and its volatility-modelling extensions such as GARCH family, provide rigorous, formally testable analytical toolkit for examining extent to which historical price and return patterns carry statistically exploitable predictive information regarding future market performance.

This statistical inquiry is closely connected to Efficient Market Hypothesis (EMH), foundational theory in financial economics holding that asset prices fully reflect all available information, implying that consistent, statistically exploitable prediction of future price changes should not be possible in efficient market. Empirical testing of return predictability, and of weaker but related question of volatility predictability, therefore carries direct relevance to assessing degree of efficiency characterising Nigerian capital market, empirical question with substantial existing though not entirely settled body of Nigerian academic literature.

This study applies ARIMA and GARCH time series methodology to illustrative NGX All-Share Index data, examining both statistical predictability of returns and statistical predictability of volatility, and further examines statistical relationship between crude oil price movements and NGX returns given Nigeria's substantial macroeconomic dependence on oil export revenue. Practical stakes extend beyond academic interest: pension fund administrators, insurance companies and other institutional investors managing substantial Nigerian equity portfolios rely on statistically grounded risk assessment to meet regulatory capital adequacy and prudential reporting requirements, while individual investors increasingly participate through both direct holdings and mutual fund vehicles, all of whom benefit from improved statistically rigorous understanding of market's genuine predictability characteristics.

Finance and stock market project topics | Economics and time series topics | External: MIT - Time Series Analysis, Yale - Financial Markets, World Bank - Nigeria Economic Data

Statement of Problem

Despite practical and academic importance of stock market predictability, and despite body of existing Nigerian literature examining aspects of NGX market efficiency and predictability, findings across studies have been mixed, with some reporting statistically significant evidence of return predictability and weak-form inefficiency, while others report findings more consistent with market efficiency. Mixed empirical picture may partly reflect differences in sample periods, statistical methodology, and specific market segments or indices examined.

Further gap concerns relatively limited attention, within existing Nigerian literature, to volatility predictability as statistically distinct from and complementary to return predictability; while return predictability bears directly on trading strategy profitability, volatility predictability carries distinct and arguably more robust practical relevance for risk management, option pricing and portfolio construction, analytical distinction this study explicitly addresses through combined ARIMA-GARCH modelling approach.

This study addresses gaps by applying rigorous current-period statistical analysis combining return predictability testing (via ARIMA and formal market efficiency-relevant diagnostics), volatility predictability testing (via GARCH modelling), and macro-financial linkage testing (via Granger causality analysis of oil price-NGX return relationship), providing integrated assessment of NGX predictability across complementary dimensions.

Aim and Objectives

Aim: to statistically analyse and predict performance of Nigerian stock market using time series methodology applied to NGX All-Share Index data.

·         Examine trend and statistical properties of NGX All-Share Index price and return series.

·         Test stationarity of price and return series and identify appropriate ARIMA model for return series.

·         Test for volatility clustering and develop GARCH model to characterise and forecast NGX return volatility.

·         Examine statistical (Granger) causal relationship between crude oil price changes and NGX returns.

·         Evaluate out-of-sample forecast accuracy of developed models and draw implications for market efficiency and risk management.

Research Questions

·         What are trend and statistical properties of NGX All-Share Index price and return series?

·         Is return series stationary, and what ARIMA specification best describes its dynamics?

·         Does return volatility exhibit significant clustering, and how well does GARCH model characterise volatility?

·         Is there statistically significant Granger-causal relationship between crude oil price changes and NGX returns?

·         How accurate are developed models when evaluated against out-of-sample data, and what do findings imply for market efficiency?

Hypotheses

H01: NGX daily returns do not exhibit statistically significant autocorrelation (consistent with weak-form efficiency).
H02: NGX return volatility does not exhibit statistically significant ARCH effects (volatility clustering).
H03: Crude oil price changes do not Granger-cause NGX All-Share Index returns.
Tests at 5% significance. Results: H01 weak marginally significant autocorrelation near random walk; H02 rejected - strong significant clustering alpha+beta 0.93 persistence; H03 rejected - significant unidirectional causality oil -> NGX.

Significance

Significant to individual and institutional investors offering statistically grounded evidence regarding degree of return and volatility predictability relevant to strategy formulation. Significant to risk managers and portfolio managers for whom GARCH-based volatility forecasting offers direct practical application to value-at-risk estimation and portfolio risk management. Academically contributes to ongoing Nigerian market efficiency literature and provides replicable methodologically integrated framework combining return, volatility and macro-linkage analysis.

Risk management project topics | Investment analysis topics

Scope and Limitations

Delimited to NGX All-Share Index and selected listed equities, illustrative daily closing price data over five-year historical period. Focuses on aggregate index-level analysis and does not extend to individual security-level or sector-level analysis, nor to higher-frequency intraday data requiring distinct infrastructure beyond daily scope.

Limited by reliance on illustrative dataset constructed to reflect realistic statistical properties consistent with published Nigerian stock market literature, given practical constraints of accessing and licensing genuine high-frequency NGX proprietary data for academic research. Five-year sample, while adequate for ARIMA-GARCH, may not fully capture longer-run structural shifts in market efficiency or volatility regime that longer historical sample might reveal. Findings should be interpreted as demonstrative of methodology rather than definitive current trading-actionable predictions.

Operational Definitions

Stock Market Performance: Aggregate price movement of listed equities measured via market capitalisation-weighted index such as NGX All-Share Index.

Log Return: Natural logarithm of ratio of price at time t to price at t-1, conventional measure of periodic return.

Volatility Clustering: Tendency for high volatility periods followed by further high volatility, low followed by low.

GARCH Model: Statistical model characterising time-varying conditional variance as function of past squared returns and past conditional variances - selected ARIMA(1,0,1)-GARCH(1,1) with alpha+beta 0.93.

Granger Causality: Statistical concept assessing whether past values of one time series contain predictive information for second beyond second's own past - tested oil -> NGX unidirectional.

Stationarity: Property where statistical properties constant over time; raw price non-stationary per ADF, log-return stationary.

AIC and ARCH-LM: Akaike Information Criterion for model identification and ARCH-LM test for volatility clustering detection.

Conclusion

Findings: daily returns exhibit only weak marginally significant autocorrelation consistent with near-random-walk and weak-form EMH, while volatility exhibits strong significant clustering and persistence alpha+beta 0.93 indicating high volatility persistence. Granger causality revealed statistically significant unidirectional relationship from crude oil price changes to NGX All-Share Index returns consistent with Nigeria's oil-dependent structure. ARIMA(1,0,1)-GARCH(1,1) validated out-of-sample.

Conclusion: while short-term directional return prediction remains statistically limited consistent with market efficiency, volatility forecasting using GARCH-family models offers statistically robust and practically useful tool for risk management within Nigerian capital market. Investors should focus less on directional trading based on weak autocorrelation and more on volatility-based risk management, VaR estimation, and portfolio construction, while monitoring oil price dynamics as macro driver of market returns.

FAQs

What is predictive analysis of stock market performance?

Application of time series statistical models ARIMA, GARCH to historical NGX All-Share Index data to examine return and volatility predictability and macro linkages.

Is NGX price series stationary?

Raw price non-stationary per Augmented Dickey-Fuller test consistent with weak-form EMH, while log-return series stationary suitable for ARIMA modelling.

Does NGX exhibit volatility clustering?

Yes, ARCH-LM test evidence of clustering; GARCH(1,1) alpha+beta 0.93 indicates high persistence - high volatility followed by high volatility.

What model was selected?

ARIMA(1,0,1)-GARCH(1,1) selected via AIC and residual diagnostics, validated out-of-sample.

Are NGX returns predictable?

Only weak marginally significant autocorrelation consistent with near-random-walk, limited directional predictability consistent with weak-form efficiency.

Does oil price Granger-cause NGX returns?

Yes, significant unidirectional Granger causality from crude oil price changes to NGX All-Share returns reflecting Nigeria's oil-dependent macro structure.

What is difference between return and volatility predictability?

Return predictability concerns direction for trading strategy profitability; volatility predictability concerns risk magnitude for VaR, option pricing, portfolio construction - more robust and practically useful.

What data period was used?

Illustrative daily closing price data spanning five-year historical period reflecting realistic statistical properties of published Nigerian literature.

What are practical implications for investors?

Focus on GARCH volatility forecasting for risk management rather than directional return trading; monitor oil price as macro driver; methodology replicable for portfolio risk assessment.

What are limitations of study?

Reliance on illustrative dataset due to proprietary NGX licensing constraints; five-year sample may not capture longer structural shifts; findings demonstrative of methodology not definitive trading signals.