The relational database relation normalization algorithm assumes sequential projection of the source table to meet the specific requirements of each normal form. The procedure for fetching data from a set of tables is based on their cascade connection, which, depending on the implementation (manufacturer and version of the database management system – DBMS), creates the need to perform recursive calculations. This means increased delays in the database when performing multi-tabular queries. The interpretation of the essence of the "normalization paradox" is that the need to ensure data integrity by normalizing relationships leads to a decrease in database performance. Testing the hypothesis that the query execution time with joining tables in a factographic relational database is correlated with the required depth of algorithm recursion due to the number of tables involved in query execution. The "normalization paradox" in a factographic relational database managed by the PostgreSQL database management system is investigated. The strength of the relationship between the degree of normalization of database tables and the time of issuing a query with subqueries depending on the depth of recursion has been established. The dependence of the issue time on the depth of recursion and on the volume of the factual database is studied separately. Based on the results of the correlation analysis, it was concluded that in databases based on the PostgreSQL v.12 database management system, the relationship between the query execution time and the depth of required recursion to generate output is weak, and the relationship between the output generation time and the database volume, determined by the number of registered instances of accounting objects, is lower than average. The conclusion is formulated about the inconsistency of the "normalization paradox" in formalized relational databases managed by PostgreSQL v.12, therefore, the unreasonableness of database denormalization as a way to increase its performance. The scientifically substantiated grounds for the necessity of bringing the relations of a factual relational database to the highest normal forms are formulated, regardless of the scope and method of its application. The results obtained for the PostgreSQL v.12 platform were not transposed to databases running other popular databases.