A proposal of a suspicion of tax fraud indicator based on Google trends to foresee Spanish tax revenues

Abstract

This article contributes to the relationship between fiscal fraud and tax collection in the Spanish economy, creating a composite suspicion tax fraud indicator (STFI) based on Google Trends searches to study the dynamics and foresee tax revenues evolution in Spain. Also, we expand knowledge in the field of fraud tax indicators, following the UNODC (2020) and OECD (2016) recommendations. To this purpose, we apply factor analysis to create the composite indicator and next we utilize techniques centred on fractional integration (ARFIMA) and fractional cointegration VAR (FCVAR) to assess the STFI behavior against tax collection and GDP. The outcomes indicate that the d is less than 1 in all the time series analyzed. As we can see, the tax collection and the leading indicator have a similar statistical behavior (d=0.49 and d=0.40, respectively), which implies mean reversion. On the other hand, GDP will behave similarly to the other two time series, with d = 0.05, which implies that the shocks will have a temporary effect on the GDP behavior, and these effects will disappear by themselves in the short term and in less time than the other two time series. FCVAR results indicate a short-lived duration of the shock due to the error correction term and their short-run stationary behaviour. In the end, applying wavelet analysis, we determine that the composite suspicion tax fraud indicator maintains a negative association with tax collection, except in 2017 and 2018, when the high economic growth offsets the fiscal fraud.

Financial Fraud Detection Using Value-at-Risk With Machine Learning in Skewed Data

“…To handle [12] large-scale non-uniform transactions more quickly, the authors employ the AR model, which makes it appropriate for detecting money laundering operations. The study [11] uses factor analysis to generate the composite indicator, fractional integration (ARFIMA), and fractional cointegration VAR (FCVAR) approaches to evaluate the behavior of the composite suspicion tax fraud indicator about GDP and tax collection. The study [13] employs the AR model, which is appropriate for studying networks with such topologies and applying it to the detection of financial transaction fraud since it considers the block-wise structure of networks.…”

“…Even tree is employed in risk assessment and analysis to pinpoint different event sequences for both fraud and nonfraud that may result in a particular outcome. An event tree also known as an incidence response tree (IRT) [49] contains four possible outcomes as given (9)(10)(11)(12) that are based on prevention, detection, and response. The event tree significantly tracks fraudulent activities and estimates the return outcomes of monitoring decisions, hence improving prediction and risk models.…”

“…Previous efforts in tackling fraud involve developing rule-based expert systems, statistical methods, machine learning, and riskbased methods [9], [10]. Due to the cost of maintenance and the inefficiency of rule-based methods [10], decision-makers decide to utilize statistical methods such as autoregressive models to handle financial fraud [11], [12], [13]. The complex patterns and high dimensional nature of frauds make the statistical methods less effective, as such machine learning models were deployed [10], [14].…”

Evolving Tax Compliance in the Digital Era: A Comparative Analysis of Ai-Driven Models and Blockchain Technology in U.S. Tax Administration

“…Their study demonstrates a positive correlation between forensic accounting skills and the detection of tax evasion, indicating that enhancing these skills can lead to more effective tax fraud detection. Monge, Poza and Borgia (2022) propose a big data analytics-based approach to tax evasion detection. By applying algorithms like K-means clustering and decision trees, their system can characterize and detect probable tax evaders, showcasing the potential of big data analytics in enhancing tax fraud detection.…”

Consumer Sentiment and Luxury Behavior in the United States before and after COVID-19: Time Trends and Persistence Analysis

“…Finally, the research carried out by [56][57][58][59] suggests that misleading results will be found if we apply a typical cross-correlation to study statistical relationships between two multifractal time series.…”