A molecular-based second-order perturbation theory is applied to calculate the relative permittivity of polar liquids. Our basic model is the dipolar hard sphere fluid. The main purpose of this work is to propose various approaches to take into account the molecular polarizability. In the continuum approach, we apply the Kirkwood-Fröhlich equation and use the high-frequency relative permittivity. The Kirkwood g-factor representing molecular correlations is calculated by a perturbation theory. In the molecular approach, the molecular polarizability is built into the model on the molecular level (the polarizable dipolar hard sphere fluid). To calculate the relative permittivity of this system, an equation obtained from a renormalization procedure is used. In both approaches, we apply a series expansion for the relative permittivity and show that these series expansions give results in better agreement with simulation data than the original equations. After testing our theoretical equations against our own Monte Carlo simulation results, we compare the results obtained from our theoretical equations and simulations to experimental data for amines, ethers, and halogenated, sulfur, and hydroxy compounds. We propose a procedure to calculate potential parameters (hard sphere diameter, reduced polarizability, and reduced dipole moment) from experimental data such as the permanent dipole moment, refractive index, density, and temperature. We show that for compounds of low relative permittivity the polarizable dipolar hard sphere (PDHS) model and the continuum approach give reasonable results. For nonassociative liquids of higher relative permittivity, the PDHS model overestimates experimental data due to unsatisfactory representation of the shape of the molecules. In the case of associative liquids, the PDHS model works well, and in some cases it underestimates the experimental values due to the unsatisfactory treatment of electrostatic interactions.