The phenomenon of the tax underground or shadow economy has been a central concern for economists and policymakers since the early 1980s. This study aims to map the conceptual and methodological developments in tax underground research from 1980 to 2025 using a systematic literature review approach. By analyzing 35 academic publications and institutional reports, this study highlights the paradigm shift from early currency demand estimation methods (Tanzi, 1980) to the Multiple Indicators Multiple Causes (MIMIC) model (Schneider & Enste, 2000; Schneider & Buehn, 2018). Classical studies emphasize the role of tax rates and law enforcement in expanding the shadow economy (Allingham & Sandmo, 1972; Jung et al., 1994), while modern research extends the focus to institutional quality, tax morale, and governance (Torgler & Schneider, 2007; Alm & Torgler, 2006). Recent studies indicate that financial technology and digitalization create new challenges in detecting hidden tax activities (Lin et al., 2020; Zhang et al., 2020), and agent-based modeling as well as machine learning are increasingly applied to simulate tax evasion dynamics in the digital era (Lazebnik & Shami, 2025). Overall, this review identifies three major trends: the diversification of methodological approaches from macroeconomic to behavioral perspectives, the growing integration of social factors such as education and morality in tax compliance models (Ciucci, 2024), and a policy shift toward digital detection and fiscal transparency. The study concludes by proposing a future research agenda emphasizing the need for multidisciplinary approaches and the use of big data to better understand the tax underground phenomenon in developing countries, including Indonesia.