RAS Technique

Definition

RAS Technique or RAS Procedure or RAS algorithm or biproportional matrix balancing technique is an algorithm originally introduced to update IO-tables in situations in which only limited survey data are available for the projection year.

The method is used in the preparation of updated I-O accounts that are based on partial survey information. The technique applies row and column balancing factors iteratively until the adjusted matrix (the transactions table) satisfies the row and column totals (commodity and industry output). The basic idea is to adjust in an iterative procedure the matrix for intermediate inputs column and row wise with appropriate multipliers until the given totals for intermediate input requirements are met. The technique converges to a solution resulting in a balanced I-O matrix.,

Procedure

The general form of the procedure is as follows:

$A^{'} = R A S$

where

• $A^{'}$ is a matrix of corrected input coefficients
• $A$ is the matrix of input coefficients in base year
• $R$ is a diagonal matrix of multipliers for rows
• $S$ is a diagonal matrix of multipliers for columns

Assume that an input-output direct coefficients table A for an n-sector economy for a given year in the past (designate this as base year “0”) and that we like to update those coefficients to a more recent year (which we will designate year “1”).

Issues and Challenges

• The signs of coefficients are preserved (No positive input coefficient will be changed to a negative coefficient).
• Zero elements remain zero (New inputs or new products are neglected).
• Enforcement of consistency may cause implausible change of some coefficients.
• The basic RAS procedure will normally fail to produce an acceptable projection of an input-output table if structural change, change in relative prices and change in technology is substantial. However, the incorporation of other exogenous data in the modified RAS procedure will tend to improve the quality of the projection. Many variations of the RAS technique exist.
  • TRAS
  • GRAS
  • KRAS