Roman Semko

The rise of robo-advisor wealth management services, which constitute a key element of fintech revolution, unveils the question whether they can dominate human-based advice, namely how to address the client’s behavioral biases in an automated way. One approach to it would be the application of machine learning tools during client profiling. However, trained neural network is often considered as a black box, which may raise concerns from the customers and regulators in terms of model validity, transparency, and related risks. In order to address these issues and shed more light on how neurons work, especially to figure out how they perform computation at intermediate layers, this paper visualizes and estimates the neurons’ sensitivity to different input parameters. Before it, the comprehensive review of the most popular optimization algorithms is presented and based on them respective data set is generated to train convolutional neural network. It was found that selected hidden units to some extent are not only specializing in the reaction to such features as, for example, risk, return or risk-aversion level but also they are learning more complex concepts like Sharpe ratio. These findings should help to understand robo-advisor mechanics deeper, which finally will provide more room to improve and significantly innovate the automated wealth management process and make it more transparent.

Type of the article: Theoretical Article

Abstract
A common assumption in corporate finance is that firms maximize profits. However, in countries with weak tax administration and limited contract enforcement, firms may understate reported pre-tax earnings to reduce tax liabilities through evasion. This paper revisits the canonical threshold-based earnings management framework and extends it by incorporating an additional (often illegal) incentive to reduce corporate income payments. Simulation results indicate that manipulation does not significantly change when latent earnings are negative. In contrast, when latent earnings are moderately positive, firms combine legal earnings management with illegal underreporting to reduce reported earnings for two purposes: (i) to shift earnings forward to increase the likelihood of meeting next-period benchmark and (ii) to lower current tax payments. At higher earnings levels, both channels plateau as manipulation costs, marginal legal costs, and detection risk increase. Using distributional (histogram-based) diagnostics, discontinuity tests, and a Probit regression model, we find that Ukrainian companies in 2024 were more focused on reducing taxable income than on beating the zero-earnings benchmark. Excess mass is concentrated in the first positive bin, and it is associated with a higher effective tax rate and lower discretionary accruals. Overall, the results suggest that in weak-enforcement settings, tax-motivated earnings suppression can dominate classic threshold-beating incentives.