Measurements of stratospheric NO2 by ground-based visible spectrometers rely on laboratory measurements of absorption cross-sections. We review low-temperature laboratory measurements, which disagree by amounts claimed to be significant. Our recalculation of their errors shows that in general disagreements are not significant and that errors in the ratios of cross-sections at low to room temperature are between ±3% and ±8.8%. Of these errors, up to ±3.5% was contributed by errors in the equilibrium constant,Kp, in those measurements where the pressure was above 0.1 mbar. We review measurements and calculations ofKp, which were accurate to ±5% from 300 to 233 K. Each method was potentially flawed. For example, infrared measurements of the partial pressure of NO2 ignored the dependence of absorption on total pressure. From thermodynamic theory, formulae forKpcan be derived from expressions for the variation of heat capacity with temperature. Contrary to common belief, coefficients in the formulae used by spectroscopists were not derived from the thermodynamic quantities. Rather, they were fitted to measurements or to calculations. Hence, they are empirical and it is dangerous to extrapolate below 233 K, the lowest temperature of the measurements. There are no measurements of NO2 cross-sections below 230 K. Extrapolation of these cross-sections to analysis of measurements of NO2 at the low temperatures of the Arctic and Antarctic stratosphere is also dangerous. For satisfactory analysis of polar spectra, the NO2 cross-sections should be measured at temperatures down to 190 K with a relative accuracy of ±1%. This difficult experiment would need a cell of minimum length 32 m whose length can be adjusted. Because their effects are circular, many errors cannot be removed simply. Although circular errors also arise in the measurements ofKpand of the infrared spectrum, their weights differ from those in the visible spectrum. The optimum experiment might therefore simultaneously measure the visible and infrared spectra andKp.