This paper presents a method for consistent partial least squares (PLS) path modeling. The authors discuss the variance-based structural equation modeling (SEM) approach, where path coefficients are derived from the construct score correlation matrix R. They show that under the assumption of standardized construct scores, the regression equation can be used to estimate the path coefficients. The authors also derive an expression for the estimated path coefficient in variance-based SEM, which is shown to be proportional to the true loading. The paper then discusses the basic design of PLS, which is essentially a factor model. It assumes that the observed variables are i.i.d. and standardized. The latent variables are assumed to have zero mean and unit variance, and the correlation between latent variables is denoted by ρ_ij. The covariance matrix of the observed variables can be written as the sum of the product of the loadings and the correlation matrix. The authors describe the Mode A algorithm, which is a fixed-point algorithm used in PLS. This algorithm is numerically stable and converges quickly from arbitrary starting vectors. The algorithm produces estimated weight vectors that are proportional to the true loadings. The authors show that the PLS estimates are consistent and asymptotically normal. The paper also discusses the estimation of the proportionality factor between the estimated weight vectors and the true loadings. The authors propose a method for estimating this factor, which involves minimizing the Euclidean distance between the sample covariance matrix and a matrix constructed from the estimated weight vectors. The authors show that this method produces consistent estimates of the proportionality factor. Finally, the authors discuss the implications of their findings for PLS modeling. They show that PLS proxies tend to underestimate the squared correlations between latent variables. However, they also show that the PLS estimates of the correlations between latent variables are consistent. The authors conclude that PLS is a consistent method for estimating the relationships between latent variables.This paper presents a method for consistent partial least squares (PLS) path modeling. The authors discuss the variance-based structural equation modeling (SEM) approach, where path coefficients are derived from the construct score correlation matrix R. They show that under the assumption of standardized construct scores, the regression equation can be used to estimate the path coefficients. The authors also derive an expression for the estimated path coefficient in variance-based SEM, which is shown to be proportional to the true loading. The paper then discusses the basic design of PLS, which is essentially a factor model. It assumes that the observed variables are i.i.d. and standardized. The latent variables are assumed to have zero mean and unit variance, and the correlation between latent variables is denoted by ρ_ij. The covariance matrix of the observed variables can be written as the sum of the product of the loadings and the correlation matrix. The authors describe the Mode A algorithm, which is a fixed-point algorithm used in PLS. This algorithm is numerically stable and converges quickly from arbitrary starting vectors. The algorithm produces estimated weight vectors that are proportional to the true loadings. The authors show that the PLS estimates are consistent and asymptotically normal. The paper also discusses the estimation of the proportionality factor between the estimated weight vectors and the true loadings. The authors propose a method for estimating this factor, which involves minimizing the Euclidean distance between the sample covariance matrix and a matrix constructed from the estimated weight vectors. The authors show that this method produces consistent estimates of the proportionality factor. Finally, the authors discuss the implications of their findings for PLS modeling. They show that PLS proxies tend to underestimate the squared correlations between latent variables. However, they also show that the PLS estimates of the correlations between latent variables are consistent. The authors conclude that PLS is a consistent method for estimating the relationships between latent variables.