We examine a unique Korean data set, the Kookmin Bank apartment price index, which is based on broker appraisals, in order to shed light on how brokers determine valuation over real estate cycles. We build a repeat sales apartment price index as well as a hedonic apartment price index by using the transaction data, which have become available in the public domain since 2006, and compare them with the Kookmin Bank apartment price index. By examining the volatility in the broker appraisals as well as the partial adjustment models of broker appraisals, we find that the appraisals are smoothed. Furthermore, the smoothing is asymmetrical and greater during down markets than up markets. These findings are consistent with the hypothesis that brokers impound new transaction prices by using the quality of information as weight. The extent of smoothing and asymmetry in smoothing broker appraisals persist over two subsequent cycles contrary to the expectation that they would become more sensitive to transaction prices as transaction prices become more widely spread.