ACCURACY LEVEL OF CAPM AND APT MODELS IN DETERMINING THE EXPECTED RETURN OF STOCK LISTED ON LQ45 INDEX

This study determines the accuracy level of CAPM and APT in determining the expected return of LQ45 and comparing the expected return from CAPM and APT models. This study uses descriptive and comparative research approaches. The population is all stocks listed in the LQ45 index while the sampling method used is purposive sampling with stock criteria that have complete data for the period November 2015 November 2019. This study uses an independent sample t-test in testing the expected return differences between the CAPM and APT models. The result showed that the CAPM Model was more accurate in determining the expected return of LQ45 stocks compared to the CAPM method. The result also showed that there was a significant difference in expected return between CAPM and APT models. Based on this result, investors can use the CAPM model in predicting the returns of the stock listed on the LQ45. For further research, can use another index in the capital market as a research object, used a longer period to get a more accurate result, and add some more macro variables.

Stocks are one of the capital market instruments that are increasingly in demand in Indonesia. Figure 1 shows the increase in the number of stock investors in the Indonesia Stock Exchange in 2015 -2019. In investing in shares, there are several types of indices that can be used as benchmarks for the performance of company shares found in the Indonesian capital market, one of which is the LQ45 index. According to Tandelilin (2017), the index can be used as an indicator to observe price movements of existing securities in the capital market. The LQ45 index consists of 45 types of shares that have stock criteria that are easily traded (liquid). Figure 2 shows a graph of the growth of the monthly LQ45 index from 2015 to 2019. The increase of the LQ 45 Index shows an increase in company performance on the Indonesia Stock Exchange. It also shows an increase in stock investor demand in Bursa Efek Indonesia along the periods.
When investing, investors will consider the level of return and risk. In the stock market, the ability to estimate the expected rate of return is needed to be able to determine which stocks are suitable to buy. The Capital Asset Pricing Model (CAPM) and the Arbitrage Pricing Model (APT) can be used to estimate the expected rate of returns.
According to Zubir (2011), CAPM was introduced by William Sharpe, John Litner, and Jan Mossin in 1964. CAPM calculates the expected return based on the stock beta. The Capital Asset Pricing Model (CAPM) was developed by William Sharpe, John Lintner, and Jan Mossin twelve years after Harry Markowitz put forward modern portfo-lio theory in 1952 (Zubir, 2011). Bodie, Kane, and Marcus in Zubir (2011) stated that the model has two main functions, as a benchmark in evaluating the rate of return of investment and assist in predicting the expected return of an asset that is not or has not yet been traded on the market.
To minimize risk at a certain level of return, investors will focus on undiversifiable risk Ahmad (2004). Models that determine the relationship between risk (covariance) and rate of return is Securities Market Line (SML) (Sharpe et al., 2005 andJogiyanto, 2016) Stephen A. Ross introduces the Arbitrage Pricing Theory in 1976 (Zubir, 2011). APT method calculated expected return based on economic factors besides market factor. According to Zubir (2011), the expected return that is calculating in APT is not only determined by mean and variance of the market, but also by various macroeconomic factors and stock beta. According to Zubir (2011), the APT model can be more accurate in predicting expected return than CAPM because includes many factors besides than market.
Mean Absolute Deviation (MAD) is used in determining the accuracy of the expected return value of shares using the CAPM and APT models. Herjanto in Yunita (2018), stated that MAD is the sum of forecast errors regardless of the algebraic sign divided by the amount of data observed. Mean Absolute Deviation is a measure of overall forecasting error for a model (Heizer and Reinder, 2005). There are several results of research using the MAD in calculating the accuracy level of APT and CAPM models in predicting stocks' expected returns. Previous research by Aqli (2015), CAPM is more accurate in predicting expected return because of MAD value smaller than MAD of the APT model. Meanwhile, Yunita (2018), found that the APT model is more accurate in predicting stock returns compared to the CAPM model. Other results indicate that there is no significant difference between MAD of APT and MAD of CAPM.
Therefore, the purpose of this study is to calculate the value of expected returns using CAPM and APT models and determine which LQ45 stocks to choose, determine the level of accuracy of the

METHOD
This study uses a descriptive and comparative research approach. Descriptive research is research that seeks to describe a phenomenon, event, event that is happening now (Noor, 2011). Sugiyono (2017), described that descriptive research as research conducted to find out the existence of an independent variable, either only on one or more variables without making a comparison or correlation with other variables (the independent variable is a stand-alone variable, instead of an independent variable, because if the independent variable, it is always paired with the dependent variable). In this research, a descriptive approach is used to identify and explain the expected rate of return of stocks using the CAPM and APT models. Comparative research is research that compares the state of one or more variables in two or more different samples, or two different times (Sugiyono, 2017). This research will conduct a comparative test of the accuracy level of the APT and CAPM models in determining the rate of return of stocks in the LQ45 index. This research using 3 macro variables that are inflation rate, exchange rate, and market index.
Operational variables are used to describe research variables into variables, indicators, and scales. Operational variables can be seen in Table 1.

Variable Indicator Scale
Expected

Return APT Model
+  n (R n -R f ) R1-n = expected return of inflation's rate, rupiah to the dollar exchange rate, and market index (IHSG) Ratio

Data Analysis Techniques
Silalahi (2012), stated that data analysis is performed to answer research questions or to test hypotheses that have been stated previously. Data analysis aims to group data based on variables and types of respondents, tabulate data based on variables from all respondents, present data from each variable studied, make calculations to answer the problem formulation, and do calculations to test the hypotheses that have been proposed. Meanwhile, the stages of data analysis are as follows: Determine the MAD value using the following formula: 9. Perform a different test using the Independent sample t-test method in hypothesis (3) with the following decision-making criteria:  If probability > 0.05, H 0 is received  If probability < 0.05, H 0 is rejected

Descriptive Statistic
The sample in this study was 44 stocks contained in the LQ45 index for the period August 2019 to November 2019 which had complete data from November 2015 -November 2019. The description of this study included the average return and beta of a stock, the average market return, risk-free, expected return of stock of CAPM and APT models from November 2015 to November 2019. Table 2 shows the average actual return of stock at 0.00978 (9.8%) up to -0.022657. During the study period, the highest rate of return was BPRT (Barito Pacific, Tbk.), while the lowest rate of return was LPKR (Lippo Karawaci, Tbk). The beta value shows the sensitivity of the company's stock price to the price of the Composite Stock Price Index (CSPI). The highest company beta value was PTBA (Bukit Asam Tbk.), while the lowest was ELSA (Elnusa Tbk.). Table 3 shows the monthly expected return of the Composite Stock Price Index (CSPI) during the study period which was 0.8%, while the average monthly inflation rate was 0.5%. The Expected Return of Stocks using the CAPM method was at 0.4% -1.5% per month. Stock beta is between -0.090 (ELSA) up to 2.807 (PTBA). Stock beta shows the sensitivity of stock prices to market prices where the greater the beta the riskier a stock is.
Based on the CAPM method, the highest expected return value was PTBA (Bukit Asam Tbk.), while the lowest expected return value was ELSA (Elnusa Tbk.). If the return value is greater than the expected return value of the CAPM method, the stock is feasible to buy because of the average value of actual returns > expected return value.     Table 4. shows the average change in the monthly value of the inflation rate, the exchange rate of the rupiah, and the return of the composite Stock Price Index (CSPI) during the study period, which were -1.07%, -0.06%, and 0.81%, respectively. If the return value is greater than the expected return value of the APT method, the stock is feasible to buy because of the average value of actual returns > expected return value. Examples of stocks with the AKRA code are 0.052 > 0.011; thus, it is feasible to buy. Besides, there were 17 feasible stocks to buy based on the APT model including  AKRA, ADRO, ANTAM, BBCA, BBRI, ERAA,  BPRT, EXCL, ICBP, INDY, INKP, ITMG, CPIN,  MEDC, TKIM, TLKM, TPIA. Based on data from Table 4.3 and Table 4.4, 17 stocks can be an option to invest because the return value was greater than the expected return using the CAPM and APT methods including AKRA, ADRO, ANTAM, BBCA, BBRI, ERAA, BPRT, EXCL, ICBP, INDY, INKP, ITMG, CPIN, MEDC, TKIM, TLKM, TPIA.     Table 6. shows the results of the significance level, sig (2 tailed) which was 0.00. Therefore, the null hypothesis (H0) is accepted and it can be concluded that there are no differences in the MAD of the CAPM method and the APT method in determining the expected return value of stocks listed on the LQ45 Index from November 2015 to November 2019.

DISCUSSION
This study aims to determine the accuracy of the CAPM and APT models in predicting stock returns of companies listed on the LQ45 index by using Mean Absolute Deviation (MAD), where the smaller the MAD value, the more accurate a model is in predicting stock returns. The results in Table 5 show that the average expected stock return generated from the CAPM model approaches the actual expected return value so that investors can use the CAPM model as a tool to predict stock returns and as a consideration in making investment decisions. APT can be used by investors who want to know in detail what macro factors affect changes in stock prices, but the use of this model is far more complicated than CAPM.
In terms of assumptions, CAPM has more assumptions than APT, so the formation of the APT model is more flexible than the CAPM, this can be seen from the use of independent variables from each model. Besides, APT explains in more detail the factors that influence stock returns compared to CAPM, but the factors included in the APT model are very broad and according to Tandelilin (2017), there is no agreement on the relevant risk factors and the number of factors included so more research is needed on these risk factors and this model will continue to evolve according to the existing situation and condition.
From the perspective of APT, the macro factors used in this study are different from the factors used in previous studies. Indra (2018) used variable market prices, risk-free assets, inflation, exchange rates, and the money supply. While Aqli (2015), used monthly SBI, the money supply, the exchange rate of Rp / USD, and monthly inflation. Lemiyana (2015), used Gross Domestic Product (GDP), SBI, and inflation. The difference in macro variables is what causes the expected return value differences obtained from the APT model.
The results of this study are consistent with the results of previous research, Lemiyana (2015), Aqli (2015), and Indra (2018), which stated that CAPM is more accurate in predicting stock returns. And not in line with research conducted by Yunita (2018), which states that the APT model is more accurate in predicting stock returns. The difference is due to differences in the object of research used as well as differences in macro variables used in research.
The results in Table 6 show that there are no significant differences in the level of accuracy between the CAPM and APT models. It means that the two models have different levels of in predict-ing stock returns. This result is consistent with previous research, Aqli (2015), Lemiyana (2015), and Ibrahim et al. (2017).

CONCLUSIONS AND RECOMMENDA-TIONS Conclusions
Based on the result, it can be concluded that the CAPM model is more accurate in determining the expected return of stocks listed on the LQ45 compared to the APT method. The result also shows that there was a significant difference in the CAPM method and the APT method in determining the expected return value of stocks listed on the LQ45 Index from November 2015 to November 2019.

Recommendations
Investors can use the CAPM method in predicting the returns of the stock listed on the LQ45 because this method has the smallest MAD. This research only uses monthly time series data from November 2015 -November 2019. It is better to use a longer period to get more accurate results in calculating expected stock returns. This research only uses 3 variables. For further research, it is better to use more significant macro variables in calculating expected stock returns using the APT method.