Through the years, researchers investigated whether stock returns and macroeconomic variables are correlated. Chen, Roll and Ross (1986) provide evidence that macroeconomic variables influence stock prices. The goal of their research is to model stock returns as a function of macroeconomic variables. Since their theory suggests that stock prices are responding to exogenous shocks.

Meaning that stock prices are only driven by macroeconomic variables. Because by the diversification argument, risk of individual stocks can be avoided by diversifying the portfolio. They test this for stocks listed in the New York Stock Exchange (NYSE) and test whether the stocks are systematically affected by the inflation rate measured as the consumer price index, the risk premium of low graded bonds against long-term government bonds, the term structure of long-term government bonds and the treasury bill of one month, and industrial production. Their results show that inflation, industrial production, the term structure and the risk premium systematically affect stock returns for stocks listed in the NYSE. And conclude that this set of macroeconomic variables are significantly priced.

In more recent studies they examine different macroeconomic variables than Chen et al.(1986) did. For instance, Flannery and Protopapadakis (2002) identify macroeconomic risk factors by using a GARCH model for the US stock exchange. They examine the impact of macroeconomic announcements on the daily stock returns and conditional volatility of returns.

A macroeconomic announcement is considered a risk factor if either the stock return or the conditional volatility of the returns change. They considered seventeen macroeconomic announcement to have an impact on either stock returns or conditional volatility or even both. They find significant results for six macroeconomic variables. The consumer and producer price index affect only the stock returns. The balance of trade, employment report and housing starts affect only the conditional volatility of the returns.

And monetary aggregate (M1) affects both the returns and the conditional volatility. Flannery and Protopapadakis conclude that macroeconomic variables do have an impact on either stock returns or the volatility in returns. Bhargava (2014) is investigating whether quarterly stock prices can be explained by firm characteristics and macroeconomic variables. Bhargava is using a simple dynamic random effects model and a comprehensive dynamic model to model quarterly stock prices for over 3000 US firms. Bhargava is using a autoregressive random effect model to test the null hypothesis that stock prices follow a random walk. The comprehensive dynamic model consists firm characteristics and macroeconomic variables to explain the stock prices.

The explanatory variables in the comprehensive dynamic model are total assets, long-term debt, earnings and dividend per share, unemployment rate, consumer price index and interest rate on treasury bills. The main findings are that there exists persistence in quarterly stock prices, and therefore the null hypothesis that stock prices follow a random walk can be rejected. The comprehensive dynamic model shows a negative effect of macroeconomic variables on quarterly stock prices. The interest rate on treasury bills and the unemployment rate have significant negative influence. The model also displays that total assets, long-term debt and earnings and dividends per share are significant variables to predict stock prices. Other researchers have conducted more detailed research in the sense whether macroeconomic variables not just influence stock returns, but whether macroeconomic variables give direction to stock returns, i.

e. are they cointegrated. Such analysis has been done by Ratanapakorn and Sharma (2007) and Humpe and Macmillan (2007). Ratanapakorn and Sharma (2007) are investigating the relationship between stock returns and macroeconomic variables by means of a Vector Error Correction Model (VECM) coupled with Granger causality test.

They are examining whether stock prices and macroeconomic variables have a long-run equilibrium and test whether the variables have a long- and short-term causal relationship. The macroeconomic variables they consider are the money supply, industrial production, inflation, the exchange rate, and the long- and short-term interest rate. They test this for stocks listed in the S&P 500. Their results indicate a negative relation for long-term interest rates, and a positive relation for money supply, industrial production, inflation, the exchange rate and the short-term interest rate. Additionally, the six variables are Granger caused by the stock prices in the long-run, but not in the short-run.

Humpe and Macmillan (2007) also find a cointegrated relation between US stocks and macroeconomic variables. They are modelling stock prices with a discounted value model (DVM) to test for cointegration effects for a number of macroeconomic variables. The stock price is determined by discounting the cash flows. The advantage of discounting the cash flows is that it can be used on the long-run relationship between stock market and macroeconomic variables. Because many long-term investors base their investment decision on the assumption that the cash flow should grow in line with the economy. They examine whether there exists a cointegrated relation between stock prices and industrial production, inflation, money supply, and the long-term interest rate. Their results indicate that US stocks are positively influenced by industrial production and negatively by inflation and the long-term interest rate. Unfortunately, they were not able to find significant results for the money supply.

Chen (2008) is examining whether macroeconomic variables can predict economic recessions, i.e. bear markets. Chen uses the Markov-switching model and the Bry-Boschan dating method to distinguish cyclical variations in stock prices from recessions.

After identifying recession periods Chen is investigating whether these recession periods can be predicted by macroeconomic variables. The various macroeconomic variables Chen considers are the interest rate spread, inflation rates, money stocks, aggregate output, unemployment rates, federal funds rate, federal government debt, and nominal effective exchange rates. Chen is using the S&P 500 index for his research. The results suggests that only the spread in interest rates and inflation rates were significant, consistent and useful in predicting bear markets. However, they did not find evidence that one was better over the other and conclude that the term spread and the inflation rate have equal forecasting accuracy.

Also, Chen found that macroeconomic variables are better able to predict bear markets than market returns. Researchers in Asia also started investigating cointegrated relationships between stock prices and macroeconomic variables, especially the countries in the growth engine of Asia (e.g. Japan, Singapore, Malaysia, and Korea).

Mukherjee and Naka (1995), for example, enlarge the findings of Chen et al. (1986) for the Japanese stock market. They try to find a cointegrated relation between six macroeconomic variables and the Tokyo Stock Exchange (TSE). The variables they use were industrial production, the exchange rate, long-term government bond rate, money supply, inflation, and call money rate. By applying a VECM model they try to determine the relationship between the six macroeconomic variables and the returns of the TSE. Mukherjee and Naka find a positive relation for industrial production, call money rates, and money supply, and a negative relation for inflation and long-term government bond rates. A possible reason why the long and short term interest rate have mixed results is that the long-term government bond rates are a better proxy for the nominal risk free rate than the short-term rate (call money rate) as discount rate for the discounted value model (DVM). Likewise, Kwon and Shin (1999) investigate whether macroeconomic fluctuations can explain stock returns for the Korean Stock Exchange (KSE).

They also use Granger causality test to find a cointegrated relation between the KSE and four macroeconomic variables. Kwon and Shin also use a VECM model to determine whether a cointegrated relation exists. The variables they consider are foreign exchange rates, trade balance, production level, and money supply. Their results show that there does not exists a cointegrated relation with the KSE and a single macroeconomic variable. However, there exists a cointegrated relation between the KSE and a combination of the four macroeconomic variables. They conclude that there exists a long-run equilibrium, though, they argue that KSE is a lagging indicator, contradicting the findings that the stock market rationally reflects changes in the economy.

They suggest that the movements in the KSE are rather due to international trading activities than to, for instance, inflation or interest rate. According to Kwon and Shin, a possible explanation could be that the KSE is more sensitive to speculative activities, manipulations and government interventions than a more developed market, e.g. US market.

Maysami, Howe and Hamzah (2004) are investigating the cointregrated relationship between the Singapore stock index, the Stock Exchange of Singapore (SES) All-S Equities Finance Index, the SES All-S Equities Property Index, and the SES All-S Equities Hotel Index and various macroeconomic variables. They employ a VECM model to examine the long-term equilibrium relationship between the stocks and macroeconomic variables. The variables they consider are the long- and short-term interest rate, industrial production, price levels, exchange rate and money supply. The results of the VECM model indicates that the Singapore stock exchange and the SES All-S Equities Property index both have significant cointegrated relationships with all the variables. While the SES All-S Equities Finance Index is only affected by inflation rates, exchange rates, and long- and short-term interest rates.

And the SES All-S Equities Hotel Index only the exchange and inflation rate were significantly priced. They conclude that there exists inefficiencies in the Singapore stock exchange and stock picking could lead to superior returns. Vejzagic and Zarafat (2013) test for cointegrated relation between the FTSE Bursa Malaysia Hijrah Shariah Index (FBMHS) and four macroeconomic variables. The FBMHS index is a response of increasing interest in Shariah compliant investments. The constituents of the index are complying the principles of the Koran. The variables that Vejzagic and Zarafat consider are the interest rate, money supply, consumer price index, and exchange rate.

They use a VECM model to determine the cointegrated relation between the index and the macroeconomic variables. Their results show that the FBMHS is influencing and leading macroeconomic variables. The FBMHS is significantly related to the money supply, consumer price index, and exchange rate. If the FBMHS is deviating from its equilibrium, it is positively affecting the money supply, and negatively the interest- and exchange rate. Unfortunately, they did not find any significant results for the consumer price index variable. The survey of professional forecasters (SPF) is since its introduction in 1968 widely examined by researchers. In 1997, Diebold, Tay and Wallis (1997) investigate the adequacy of the inflation forecast of the SPF.

They evaluate the density forecast of inflation, where they are more interested in assessing the adequacy of the forecast rather than the construction of the forecast. Because there is little known about the construction of the forecast. Their results indicate that the SPF forecast for inflation is not the best. In the deep recession of 1973-1974, inflation increased sharply but was underestimated by the forecasters. The same occurrence happened, although more weakly, during the period from 1976 till 1980. While during the recovery of 1975-1976, inflation was overestimated. The same holds for the period from 1981 till 1985.

Finally, from 1986 till 1989, forecasters predicted the inflation rate in general quite well. They conclude that forecasters are overestimating large negative shocks in inflation. Meaning that a negative shock in inflation occurs less often than the forecasters expect. They also even find evidence that surprises in large positive shocks in inflation are overestimated. They also find that the forecasts are autocorrelated and that lower inflation coincide with lower uncertainty. Campbell (2004) investigate the predictability and uncertainty in the real output growth during the Great Moderation.

The Great Moderation is seen as the period from 1980 till late 2000s. This period is known for its reduction in the volatility of macroeconomic variables. In his research, Campbell is comparing the point forecast of the SPF with an autoregressive model.

More specifically, Campbell is comparing the probability of a decline of the SPF with a autoregressive model. Since the SPF contain information on the probability of a rise/fall in the macroeconomic variable. The results show that before the Great Moderation the forecast of the SPF and the autoregressive model, rise before and fall after a recession. However, the SPF shows more variation in its forecasts than the autoregressive model does, i.e. the SPF rises and falls more after recessions than the autoregressive model. After the Great Moderation, the SPF still has more variation than the autoregressive model, though the difference between the two is smaller.

They conclude that the SPF forecast outperform the autoregressive model before the Great Moderation. After the great moderation the outperformance of the SPF is roughly the same as the autoregressive model. Campbell’s findings are in line with Diebold et al. (1997) that shocks in macroeconomic variables are overestimated in the SPF forecasts.

Forecasters have different beliefs about the future, and therefore different predictions. Engelberg, Manski and Williams (2006) try to find dispersion between individuals with the same beliefs. Their research is comparing this subjective probability distribution with the point forecast using the SPF data for inflation and GDP growth. Thus, the point forecast from each individual is compared with the mean, median and modes of their subjective distribution. The results show that the predictions differ significantly among forecasters. Many forecasters are giving a more favorable point forecast relative to their subjective distribution. For GDP growth, if the prediction of the forecaster is deviating from its mean, median or modes, then the forecaster is more likely to report a prediction above the upper bound than below the lower bound. While the opposite holds for inflation.

Meaning that forecasters are optimistically predict the GDP growth and pessimistically predict the inflation rate. They also notice that forecasters whom report favorable forecasts tend to do so in the periods following up. This is in line with Diebold et al.

(1997) that the forecasts are serially correlated. They conclude that the SPF should not be used to point forecast macroeconomic variables. Lamont (2001) tries to find a reason why forecasters are overestimating their predictions. Lamont suggests that the forecast depends on the age and the reputation of the forecaster. In case of age, the older the forecaster becomes the more thorough forecast he will make.

Lamont finds evidence that older people are making more daring and risky forecasts compared to their forecasts when they were younger. In case of reputation, the forecaster is rewarded for their forecast. The forecaster is trying to acquire a reputation by giving a favorable prediction. Therefore the forecaster has an incentive to manipulate their own beliefs and forecast relative to other forecasters, because they want to maximize the market value of their reputation.