Error-component analysists argue that residuals generated by subtracting predicted values of test scores from the unconditional class-level mean and then aggregated at the class level can be regarded as teacher effectiveness. The paper tests this argument by using self-collected teacher-student matched data from 22 counties or districts in Hubei province and Guangdong province. Correlated random effect model (CRE) is employed to predict students’ test scores residuals, main findings of this paper are listed as below: (1) there are huge differences among teachers in terms of teaching performance, if a teacher of average quality is replaced by a colleague at 69 percentile whom Hanushek called “a good teacher”, the whole class will witness a net achievement gain of 0.2-0.3 standard deviations depending on which subject is under concern, the effect of which is equivalent to one-year acquisition of knowledge and skills. If he or she is replaced by a “great teacher” standing on the top 16% of the teaching performance distribution, test scores of the class will boost by 0.3-0.5 standard deviations. (2) The economic value of quality teachers is enormous, if a class of 49 students taught by a good teacher is assigned to an average teacher, the present value of earnings losses suffered by the whole cohort of students is between 110 thousand RMB and 240 thousand RMB. By our conservative estimation, the monetary value of a good instructor’s teaching is worth of at least 4.25 million more than a mediocre one across his or her 38-year career life. The article shades light on how to evaluate and value teaching performance.