An accurate prediction of the crack width is important for assessing the serviceability limit state of reinforced concrete structures. Often, experiments are conducted to measure the crack width in reinforced concrete elements. The measurements can then be used to develop formulas to predict the crack widths. The effect of shrinkage, prior to testing, is often not considered when formulas are developed. Consequently, there is a difference between crack widths predicted taking shrinkage into account and without shrinkage when comparing them to measured widths. This contribution aims to quantify this difference due to the inclusion of the effect of shrinkage prior to testing. This difference is dependent on how the effect is included in the crack width formula. A database was created based on experiments reported in the literature. The restrained shrinkage strain was calculated based on this data to represent the axial shortening and curvature. The crack widths were then analysed using some frequently used formulas based on curve-fitting, with and without considering shrinkage prior to testing. Finally, the predicted widths were compared with crack width measurements. Results show that the crack widths are generally more underestimated when the effect of shrinkage prior to testing is neglected, and the degree of underestimation is dependent on the formula. Hence, in the development of new crack width formulas, where an accurate comparison with predicted widths and measured widths is needed, the effect of restrained shrinkage cannot be ignored. The presented methodology can be incorporated for the development of new formulas.