Amsterdam Business School
MSc Business Economics: Finance Track
Master Thesis
How do the Chinese Stock Market react to
Changes in the Demographic Pattern?
Student Name: Guoxiao Xia
Student Number:10825746
Supervisor:Rafael P.Ribas
State of originality
The work presented in this thesis is, to the best of my knowledge and belief, original, except as acknowledged in the text, and the material has not been submitted, either in whole or in part, for a degree like this or any other university.1 The Faculty of
Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Guoxiao Xia July 2015
1. I gratefully acknowledge my supervisor Rafael P. Ribas who sheds light on the topic of this thesis, provides abundant wonderful papers to read or study and gives many helpful instructions and thoughts to me.
Catalog
Abstract 3
1.Introduction 4
2.Literature Review 8
3.Methodology 14
4.Data and descriptive statistics 20
5.Results 27
6.Robustness Checks 39
7.Conclusion 40
8.References 43
Abstract
Although the shifts of the stock price, sales, and investment cannot be predicted every trading day, we can still find some trends which will help us to predict the stock prices, and the other two aspects in the future. Investors respond to the current good or bad news right on time, but how do they react to predictable shifts in a relatively long time horizon, for example, 1 year? We here are very interested in how demographic shifts will affect the stock return, sales growth and investment growth in certain industries in the Chinese stock market.We consider five specific cohorts and industries to test whether the change in different cohort size will affect the stock return in different industries. Our regression models are based on the panel regression model with time trend.We find that the log difference in some cohorts do affect the stock return, sales growth and investment growth in certain industries. For instance, the stock return of cars industries will escalate nearly 7.89% more when 1% increase occurs in the cohort which is consisted of people who are 20 to 29 years old compared to the cohort which is consisted of people who are 0 to 19 years old. Simultaneously, the sales growth proliferates 2.50% more and there is no significant influence on the investment growth of the industry. However, like pharmaceutical industry, investment growth is affected prominently by the shifts in demographics.We notice that the investment growth of the companies which operate drugs and electronics are influenced by demographic changes visibly. Finally, a trading strategy based on the result is recommended and we discuss some possible explanations for the results.
I. Introduction
Do the shifts in demographic patterns have impacts on stock returns, sales growth and investment growth across industries? There is little literature considering the effect of demographics on stock returns and the other two aspects aforementioned while abundant research delve into the impact of demographic fluctuations on aggregate stock returns (Andrew B. Abel 2003; Andrew Ang and Angela Maddaloni 2005; James M.Poterba 2001). In our work, we investigate the effect of demographics on panel stock returns, sales growth and investment growth. We consider the impact of changes in cohort sizes on stock returns, sales growth and investment growth in the previous decade and predict the trend of certain industries in the next decade. If the demographic shifts do provide evidences on the future investment opportunities or sales booming strategies, certain firms can envision increase in their sales and the investors can anticipate considerable abnormal portfolio return if they exploit this information. Moreover, the companies can also switch their investing and developing decisions. The contribution of this thesis is digging out the influence of predetermined demographic changes on the stock returns, sales growth and investment growth as well as illuminating investors and business decision makers to manage their money more efficiently in the future.
Specifically, we use panel regression with time trend and controls to test the effects of the log difference of five cohort sizes on firms’ stock returns, sales growth and investment growth if they are quoted in the stock market. There has been some literature which shed light on how to divide the cohorts (Stefano DellaVigna and Joshua M. Pollet 2007), however, they just have three groups: age 0-18, age 65+ and age 0-99. This kind of method cannot be applied here because people have different consuming patterns nowadays from the last decade and the method above is too broad. Therefore, we define the cohorts ourselves. This may not be the most convincing one but we guess it is reasonable in this thesis. Cohort 1 is consisted of people who are 0
to 19 years old and we consider them as youth since most of them are children or students who do not have their own income and depend their lives mainly on their parents or relatives. So what they want most are just foods, electronics and products that are needed in daily life. Cohort 2 is consisted of people who are 20 to 29 years old and we consider them as young workers who have worked only a short time and do not accumulate much money and some of them are still studying for their bachelor, master or doctor degree. Thus, they are not able to buy expensive items such as cars or houses themselves but they can get money from their parents to buy these expensive items. However, electronics are most popular here as you can see in most schools or universities. The apple mania in 2011 elucidates that electronics zeal are boundless among teenagers and young workers. Therefore, it is reasonable to categorize them into a cohort which demands electronics the most. People who are 30-39 years old (mature workers) constitute the cohort 3. The differentiation of demands among these people are very common. They have worked for a relatively long time and may save some money for the expensive items. Some of them have intense demand for electronics like the teenagers, some of them are planned to marry and willing to buy cars or houses and some of them do not have strong signs that reveal what they demand most. Additionally, some of these people have children to raise so they have great demand of foods. Hence, they cannot be considered to be the most consumers for any specified goods by intuition. Cohort 4 is consisted of people who are 40 to 59 years old (steady workers). Their lives are tend to be more steady and their children are going to be mature. So cohort 4 is similar to cohort 3. These people do not have that intense need of houses or cars but may have more and more demand for drugs. Therefore, they cannot be considered to be the most consumers for a specified good but they have more needs for drugs and less needs for expensive items than the people who are categorized into cohort 3. Finally, elderly people who are above 60 years old are categorized into cohort 5. Obviously, these people have the most needs for drugs.
that they can be predicted in advance based on the birth rate, mortality rate, and previous demographic structure. Current cohort sizes enable us to have accurate forecasts of future cohort sizes even at long horizons assuming that the fertility and mortality rate do not have a significant change.
Different goods have distinctive age profiles of consumption based mainly on the decision of household. Therefore, predictable changes in the cohort sizes can be used to get forecastable demand shifts for various goods. Digging out this information may lead us to get a better sketch of the operating decisions of certain industries. The market is not efficient if this information cannot be found by every individual. Hence, how the stock market or managers react to aforementioned predictable demand shifts produces evidence about how investors respond to such additional information on the stock market. Simultaneously, how the managers of companies interpret this information may affect the companies sales and investment decisions.
Why we choose China? To begin with, according to the research done by the United States Census Bureau, the total population in the world is 7.359 billion (1st Mar 2015). China has the most population in the world (1.368 billion on 1st Mar 2015). So, the demographic changes in China is considered to be more significant compared to other countries. Subsequently, the research done by Bloomberg said that the estimated Chinese GDP growth rate is 7.0% in 2015 ranking the first in G20 countries while the estimated GDP growth rate of United States and Europe is 3.1% and 1.2%, respectively. Consequently, this prediction elucidates the importance of the Chinese economy in the future and implied that there are still plenty of investment opportunities of high return in China. Finally, the Shanghai Stock Exchange and the Shenzhen Stock Exchange are founded after 1991, and this fact illustrates that the Chinese stock markets are immature or in other words incomplete compared to the American ones. Thus, the Chinese stock markets may react inconsistently with American ones since the markets are less effective and the investors who get the information about the demographic shifts may have a more significant abnormal
return in China. As for the sales growth and investment growth, we will see how they are affected. But we guess the most important factors may be the macroeconomic indicators like GDP growth in real estate and cars industry since they are affected by the government policies and the whole economy.
In section II, we will review the existing literature and the theories they have founded and derive our hypotheses. Meanwhile, we will explain how our research are related to the previous research. In section III, we explain the type of data we need, the econometric model we are going to test and the causal relationship among variables. In section IV, we present summary statistics of the data. And in section V, we present our main results, interpret the economic meaning of our results.The robustness checks are presented in the sixth section. And finally, in section VII, we draw our conclusions and develop some discussions.
II. Literature Review
Della Vigna and Pollet(2007, 2013) sheds light on our idea. They use demographic information and consumption patterns to build the demand forecasts. Then, they claim that cohort size fluctuations would provide predictable demand changes for age-sensitive goods or services, such as bikes, toys, and nursing homes. And these demand changes can be predicted when a specific cohort is born. They present the main idea of the paper by an example. Assuming that a large cohort is born in 2004, the demand for school buses in 2010 will increase visibly since these babies are 6 years old and need to go to kindergarden. Meanwhile, if the school bus industry is not perfectly competitive, that is to say some companies involved will have an abnormal increase in their sales due to the baby boom in 2004. The investors are aware of the positive demand shift induced by that baby boom and they predict that the quoted company which operates the school bus will enjoy an increase in the near future. Thus, the price of school bus shares will proliferate in 2004 until this so-called additional information is known by other investors. In other words, the opportunities to gain abnormal portfolio returns on school bus shares in the future will diminish as time goes by. The most important thing is that how fast the investors respond to such a predictable trend of cohort size. Investors may have a myopic attitude toward stocks so they only care about information that will affect the tendency of stock in the short run. Or, investors may be inattentive to information about future trend that will occur in a long time, for example, five years which is the longest horizon that financial analysts make predictions about the companies’ future profitability in the previous literature. A third scenario occurs when investors overreact to this information. That is to say, abnormal portfolio returns would be extremely high in 2004 and becomes lower and lower in subsequent years since realized profits are not able to meet the anticipations from investors. Based on the reasoning above, it is necessary to take the short term and the long term effects into account. Then, they apply lagged consumption and demographic data to forecast future consumption demand growth
and find that demand forecasts can predict profitability by industry. To estimate age-consumption profiles, they use four historical surveys (1935-1936, 1960-1961, 1972-1973, and 1983-1984 Consumer Expenditure Survey) on consumer expenditure and find three major characters of consuming pattern. First of all, consumption of most goods is decided mainly by the age of the household. Subsequently, the age profile of consumption differs significantly across goods. Finally, the age profile is quite steady among the four surveys for a given good. All these findings serve as a solid support for the use of cohort size as a predictor of consuming pattern.
What’s more? They find that demand changes derived from demographics predict abnormal stock returns by investigating the relationship between ROE ( Return on Equity) and the demand changes. Next, they employ panel regressions with year fixed effects or industry fixed effects and use Fama-MacBeth regressions to test return predictability.
This paper sheds light on the original idea of our work. That is, forecastable change in demographic patterns can give us some information about the future financial aspects of certain industries and may predict stock returns in the near future. For instance, the more demand by the cohort, the more sales and profits can envision in the future. Although we do not find the consumption data like the four aforementioned surveys in US, we get the main causal relationships from this paper. Our research consider the causal effects of the change in five specified cohorts on stock return of certain industries.
Cunha and Pollet(2014) use predictable demand shifts from DellaVigna and Pollet (2007, 2013) to investigate why firms hold cash and provide an prominent assumption about cash flows. They mention that predicted demand shifts ought not to affect firm’s current cash flow. To be specific, these shifts only happen farther into the future and, as a result, would not have considerable influence on firm’s current operating decisions. This is vital indeed since it guarantees that demographic shifts are not
related to current circumstances such as macroeconomic indicators and firm fundamentals. In other words, this assumption rules out omitted variable bias that may cause endogenous problems in the regression model.
They also provide some specifications to measure the effects of consumption shifts on firms’ cash holdings and categorize the effect of demand shifts into two main types: long-term (basically 5 to 10 years after the shift) and short term (less than 5 years).
The major findings of this paper are as follow. To begin with, they show that managers would adjust cash holding of the firm based on a predictable demand shift in the long run. And this result is more significant for financially constrained firms which should handle their current dire straits urgently. Secondly, when the managers have high leverage and investment chances in the industry become worse and worse (due to growing competitions), they also raise their cash holdings. Finally, their research illustrate that firms’ operation decisions would balance the costs and benefits of holding cash by targeting an optimal cash holding level. And their results elucidate that changes in investment opportunities have causal effects on firms’ cash holdings and are not related to endogenous factors such as firms specific characteristics.
We are illuminated by their work since demand shifts affect the firms’ cash holdings and may have impacts on their future development decisions. Therefore, if investors dig out such information before the stock price fluctuate, they may be able to have a clear estimate of the trend of the stock price in a short horizon. And this fact might lead a distinctive trading behavior by such investors from others who do not know this information. Consequently, the cash holdings of the firm which is partly determined by demographic shifts may have influence on the stock price.
Abel (2003) employs an overlapping generations closed economy in which consumers live for two periods to estimate the influence of baby booms on capital price. They base their research on the overlapping model theory in macroeconomics. By
illustrating a Cobb-Douglas function of consumption and technology, the author shows some insight about the relationship between consumption goods and capital goods which enlighten me about how to measure the influence of demographic shifts on the consumption pattern of people.
What the author has found is that a baby boom will raise the price of capital. When baby boom generations go to work and earn wages, the national saving and investment are relatively high. Based on the aforementioned two-sector economy that produces consumption goods and capital goods, a high rate of investment would only appear when the supply price of capital inflates. Therefore, a baby boom drives up the price of capital.
This paper shows that a baby boom would affect the price of capital and forms the foundation of our research, scilicet, it reveals that a big change in demographic pattern do affect the price of capital which we can understand as the stock price.
In accordance with what Abel (2003) has found, James M. Poterba (2001) also shows that the baby boom generation who born two decades after the World War II has had and will continue to have eminent impacts on the U.S. economy. The author claims that when the baby boom generation are young, they exhibit high demands on infrastructure for education. Then, when they go to work and earn wages, they might be associated with an escalation in some macroeconomic indicators such as aggregate unemployment rate. When they are about to retire, they will depend their lives on government pension funds or program like Social Security and Medicare.
Also, previous scholars suggest that the growing demand of baby boomers partly attributed to the rise in U.S. stock prices during the 1990s. There is a fierce debate among scholars about how the aging of the baby boom will affect the financial markets. A very limited number of studies have inspected the historical relationship among asset prices, demographic patterns, and asset returns even though the potential
relationship between demographic patterns and asset returns is widely discussed and studied before. Thus, to study and investigate how the demographic shifts may affect the stock returns and other financial aspects is urgently needed and may draw a impressive and appropriate remark for the controversial issue in the future.
Vuolteenaho (2002) uses a vector autoregressive model (VAR) to divide a company’s stock return into two components. One is the changes in cash flow anticipations, for instance, cash flow news, headlines or financial reports while the other is the changes in discount rates. The author derives three main results from the VAR model. Firstly, firm level stock returns are predominantly driven by cash flow news. And for market adjusted log returns, the variance of expected return news is one-fifth of the cash flow news variance. Secondly, the author finds that cash flow news is positively correlated with impacts to expected returns. This correlation is prominent for smaller stocks and approximately zero for large stocks. Thirdly, cash flow news is more easily to dissipate in portfolios than expected return news. For an equal weight portfolio, the cash flow news variance is nearly 75% of the expected return news variance. The result implies that cash flow information is mainly based on firm specific characteristics while return anticipation is determined by systematic and market wide components.
This paper reveals that cash flows play an important role in stock markets and may influence the stock returns. This fact strengthens the theoretical foundation of our research.
Here, we purpose our main hypotheses.
From the real side and the financial side, we present hypothesis 1---market has less information than the managers. Hypothesis 2---different cohort has distinctive consumption pattern. Hypothesis 3---demographic shifts can be used to predict the stock returns, sales growth and investment growth to some extent which means we
can predict the demand for certain goods and forecast the stock returns, sales growth and investment growth by analyzing the demographic shifts.
III. Methodology
The ultimate goal of our research is to find out the impacts of predominated demographic shifts on the stock returns, sales growth and investment growth. The time horizon we choose is 2004-2013. And the five industries in our sample are quoted companies which operate cars, real estate, foods, electronics and drugs. Foods are daily needed for every cohort especially for the people who are 0 to 19 years old. They do not have much income and depend their lives mainly on their parents or relatives. Therefore, what they need most is food. Recently, more and more people have bought electronic products since the 21st century. And the apple mania in the world shows the intense need for electronics among teenagers and young workers. So, electronics can represent what the young adults want most. As for people who are 30 to 39 years old, they have less enthusiasm to buy electronics but want to purchase more cars or houses than the younger generation. Meanwhile, the people who are 40 to 59 years old need less cars, foods and houses than people who are 30 to 39 years old and these people begin to consume drugs. We consider the time from 30 years old to 59 years old to be the process of transition in one’s life since the demand of individual varies markedly across this period. When the people turn to their retirement age, here we say above 60 years old, they do not have much need for cars or electronics but have very strong need for drugs. Hence, the five industries we choose can represent the age profile of consumption during one’s life.
As for the data, firstly, we need the data for the structure of the population from 2003 to 2013. The data can be found in the Statistic Yearbook of China. We use the linear interception method to manipulate the missing data. The data are used to get a overall sketch of the shifts in each cohort (with 5 years gap, for instance, 0-4, 5-9, ... and 94-99) and make a simple prediction based on the tendency of the cohort sizes. Secondly, we need the sales data and total investment data from 2003 to 2013. The data can be found in the GTA database. For the total investment data, we should note
that the total investment is calculated as the sum of other three items of the fiscal report of the company. They are net increase of fixed assets, net increase of projects under construction and long term investment.Thirdly, we need the monthly data for stock return for the quoted companies which operate electronics, cars, real estate, foods and drugs in Shanghai Stock Exchange and Shenzhen Stock Exchange. Fourthly, the monthly market return of the Shanghai and Shenzhen Stock Exchange. Here we should note that the stocks whose code begin with “60” are the companies which list in Shanghai Stock Exchange, so the market return for them are calculated by the log difference of the monthly market index of Shanghai Stock Exchange.Meanwhile, for the stocks whose code begin with “00” are the companies which list in Shenzhen Stock Exchange and the market return for them are calculated by the log difference of the monthly market index of Shenzhen Stock Exchange.
We should include a bunch of control variables in the model, too. Specifically, we choose Fama French three factors for the market, including SMB, HML and (Rm-Rf). Here we use (Rm-Rf) which is defined as the excess return on the market, value-weight return of all quoted firms in the Asia Pacific including Australia, Hong Kong, New Zealand, and Singapore, namely, the excess return on foreign stock market in the robustness check part since we have market return in domestic stock market in the regression. Monthly SMB (Small Minus Big) is the average return on the three small portfolios minus the average return on the three big portfolios, that is “SMB”=1/3 (Small Value + Small Neutral + Small Growth)- 1/3 (Big Value + Big Neutral + Big Growth). Monthly HML (High Minus Low) is the average return on the two value portfolios minus the average return on the two growth portfolios, that is HML=1/2 ( Small Value + Big Value) - 1/2 (Small Growth + Big Growth). All these data can be downloaded from Kenneth R.French Data Library. Additionally, we need the Chinese GDP data and the Chinese population data from 2003 to 2013 which can be found in the world bank.
2. See Fama and French, 1993, “ Common Risk Factors in the Returns on Stocks and Bonds,” Journal of Financial Economics, for a complete description of the factor returns.
The specifications are: Controls TimeTrend MR Shift Shift Shift Shift return t t t t t t i 5 4 3 2 1 0 , 60 4059 3039 2029 (1) Controls TimeTrend MR Shift Shift Shift Shift h salesgrowt t t t t t t i 5 4 3 2 1 0 , 60 4059 3039 2029 (2) Controls TimeTrend MR Shift Shift Shift Shift growth investment t t t t t t i 5 4 3 2 1 0 , 60 4059 3039 2029 (3)
The three regressions here use panel data with time trend and control variables to analyze the influence of the change in five cohorts on company’s stock return, sales growth and investment growth in the current year. The data we use in the regressions above have been altered from the original ones. The following is the variable description.
Table 1: Definition Of Different Variables
The table gives definition of different variables and the method of transferring the original data into the data we need for the regression.Cohort 1 contains people who are 0 to 19 years old.Cohort 2 contains people who are 20 to 29 years old.Cohort 3 contains people who are 30 to 39 years old and cohort 4 contains people who are 40 to 59 years old while cohort 5 contains elderly people who are more than 60 years old.
Dependent Variable Definition Method of calculation
return (i,t) yearly stock return for entity i in year t sum up log monthly stock return in 12 months for each entity i in year t
sales growth (i,t) sales growth for entity i in year t log difference of sales for entity i in year t
investment growth (i,t) investment growth for entity i in year t log difference of investment for entity i in year t
Independent Variable
Shift 019(t) * shift for cohort 1 in year t
log difference of the percentage of people who are 0-19 years old in year t
Add 0-4, 5-9, 10-14 and 14-19 years old group up
Shift 2029(t) shift for cohort 2 in year t
log difference of the percentage of people who are 20-29 years old in year t
Add up the cohort with 5 year gap per group
Shift 3039(t) shift for cohort 3 in year t
log difference of the percentage of people who are 30-39 years old in year t
Add up the cohort data with 5 year gap per group
Shift 4059(t) shift for cohort 4 in year t
log difference of the percentage of people who are 40-59 years old in year t
Add up the cohort with 5 year gap per group.
Shift 60(t) shift for cohort 5 in year t
log difference of the percentage of people who are above 60 years old in year t
Add up the cohort with 5 year gap per group
MR (t) market return in year t sum up log monthly market return in 12 months
in year t
Rm-Rf (t) ** excess return in year t (Fama French Factors 3) logarithm the original divided by 100 in year t
Time Trend Time trend of the panel regression Minus year by 2004. So the time varies from 0 to 9
Controls
SMB Small Minus Big in year t(Fama French Factors 1) logarithm the original divided by 100 in year t
HML High Minus Low in year t(Fama French Factors 2) logarithm the original divided by 100 in year t
Log GDP growth Chinese GDP growth in year t log difference of Chinese GDP in year t
When you read the table above, you should notice that the Shift 019 (t) is marked with a “*” which denotes that we will not use this variable in the regression since the shifts in five cohorts add up to 0. In every single year because these five cohorts constitute the whole population and the percentage of the five components add up to 100%. Therefore, in order to avoid the collinearity problem we omit one of the five cohorts (here we choose to omit cohort 1) in the regression. Nevertheless, the coefficients of other four cohorts turn to be the relative change compared to the cohort 1. And we mark Rm-Rf (t) with “**” which denotes that we would use this factor in robustness check but not in the regression part since the return on domestic stock market is involved as an independent variable in the specification. So we use excess return in Asia foreign markets excluding Japan as an independent variable in the robustness check part.
For the hypothesis 1---market has less information than the managers, we expect the coefficients of Shift 2029(t), Shift 3039(t), Shift 4059(t) and Shift 60(t) are significantly different from 0 in regression (1). This fact illustrates that managers can exploit additional information from demographic shifts and can use them to decide for a better trading strategy. That is to say, the market has less information than the managers since individual investors cannot know this information. For the hypothesis 2---different cohort has distinctive consumption pattern, we expect the coefficients of Shift 2029(t), Shift 3039(t), Shift 4059(t) and Shift 60(t) are significantly different from 0 in regression (2). So we can find the difference of consumption pattern for different cohort group. To test the hypothesis 3---demographic shifts can be used to predict the stock returns, sales growth and investment growth to some extent. We expect that for certain industries the coefficients of the change in four cohorts are significantly different from 0 for two of the specifications or even three of them. If it is the case, we can conclude that the demographic shifts do provide meaningful and helpful instructions for companies to invest and operate wisely in the future. Meanwhile, we expect the coefficient of some cohorts are positive or negative. For instance, for the pharmaceuticals industry, we expect the coefficients of Shift 4059(t)
and Shift 60(t) are positive while the coefficients of the other two cohorts are negative because the elderly people obviously need more drugs than the younger generation according to our basic knowledge. Additionally, for the investment growth part, we expect the coefficients of the four cohorts would vary visibly when we add controls and time trend to the regression since the investment strategy of the industry, such as real estate, is mainly depended on the overall macroeconomic circumstances and the government policies.
IV. Data and descriptive statistics
(i) Dependent Variables
Our stock return, market return, sales and investment data can be found in GTA database. And we can get the cohort size, population and GDP from Statistic Yearbook of China while we can find the Fama French three factors from Kenneth R.French Data Library. Here we present the summary statistics for the dependent variables of three specifications.
Table 2: Descriptive statistics of three dependent variables in the five industries
The table shows the smallest, largest, mean and median value for the five selected industries---real estate, foods, drugs, electronics and cars. Also, the table presents the standard deviation of each variable in the last column.
Industry (1) (2) (3) (4) (5) Real Estate Smallest Largest Mean Median Std.Dev.
return -0.6469 0.7241 0.0171 -0.0427 0.2862
Sales growth -0.0898 2.1312 0.2597 0.0732 0.5662
Investment growth -2.4395 2.3490 0.0140 0.0535 0.9423
Foods Smallest Largest Mean Median Std.Dev.
return -0.4850 0.5232 0.0360 0.0363 0.2544
Sales growth 0.0573 0.1198 0.0906 0.0956 0.0183
Investment growth -1.2301 1.2386 0.0415 0.0869 0.5334
Drugs Smallest Largest Mean Median Std.Dev.
return -0.3568 0.5758 0.0599 0.0955 0.2852
Investment growth -1.2073 1.2537 0.1232 0.1716 0.5255
Electronics Smallest Largest Mean Median Std.Dev.
return -0.5675 0.6442 0.0274 0.0200 0.2555
Sales growth 0.0103 0.1611 0.0728 0.0720 0.0385
Investment growth -1.3565 1.4392 -0.0077 0.0434 0.5269
Cars Smallest Largest Mean Median Std.Dev.
return -0.6028 0.6758 0.0277 -0.0008 0.2918
Sales growth 0.0374 0.1636 0.0908 0.0862 0.0391
Investment growth -1.8258 2.2667 0.0282 0.0348 0.8017
1. Real Estate Industry
As depicted in the table above, the logarithm yearly stock return of the real estate industry, varies from -0.6469 to 0.7241 for different firms, and there is also quite some variation in stock returns, as can be seen from the standard deviation, reported in column (5), which is around 0.2862 for each quoted company in the Chinese stock market. Also, the mean value and the median value of the logarithm yearly stock return of the real estate industry are 0.0171 and -0,0427 which are reported in column (3) and column (4), respectively. As for the yearly sales growth of the real estate industry, it has the smallest value of -0.0898 and the largest value of 2.1312, reported in column (1) and column (2), respectively. At the same time, the mean value, reported in column (3), is 0.2597 which is greater than the median value, reported in column (4) and had a value of 0.0732. Additionally, the standard deviation of the sales growth of the real estate industry is 0.5662 which is reported in column (5). Then, we find that the yearly investment growth of the real estate industry ranges from -2.4395 to 2.3490 which can be seen in column (1) and column (2). Meanwhile, the column
(3), column (4) and column (5) illustrate that the mean value, median value and the standard deviation of the yearly investment growth of real estate industry are 0.0140, 0.0535 and 0.9423, respectively. Overall, the yearly stock return, sales growth and investment growth increase over time on an average level since the mean value of these three variables are all positive.
2. Foods industry
The table shows that the logarithm yearly stock return of the foods industry, varies from -0.4850 to 0.5232 for different firms, and there also exists some variation in stock returns, as can be seen from the standard deviation, reported in column (5), which is around 0.2544 for each quoted company in the Chinese stock market. Meanwhile, the mean value and the median value of the logarithm yearly stock return of the foods industry are 0.0360 and 0.0363 which are almost the same and are reported in column (3) and column (4), respectively. As for the yearly sales growth of the foods industry, the smallest value is 0.0573 and the largest value is 0.1198, reported in column (1) and column (2), respectively. Also, the mean value, reported in column (3), is 0.0906 which does not have much difference from the median value, reported in column (4) and had a value of 0.0956. In addition, the standard deviation of the sales growth of the foods industry is 0.0183 which is reported in column (5). Then, we find that the yearly investment growth of the foods industry ranges from -1.2301 to 1.2386 which can be seen in column (1) and column (2). Simultaneously, the column (3), column (4) and column (5) illustrate that the mean value, median value and the standard deviation of the yearly investment growth of foods industry are 0.0415, 0.0869 and 0.5334, respectively. All in all, the yearly stock return, sales growth and investment growth rise over time on an average level because the mean value of these three variables are all positive.
3. Drugs industry
The table depicts that the logarithm yearly stock return of the drugs industry, varies from -0.3568 to 0.5758 for different firms, and there is some variation in stock returns,
which can be known from the standard deviation, reported in column (5) and had a value around 0.2852 for each quoted company in the Chinese stock market. Then, the mean value and the median value of the logarithm yearly stock return of the drugs industry are 0.0599 and 0.0955, reported in column (3) and column (4), respectively. As for the yearly sales growth of the drugs industry, the smallest value is 0.0424 and the largest value is 0.1223, reported in column (1) and column (2), respectively. Furthermore, the mean value, reported in column (3), is 0.0878 which is almost the same as the median value, reported in column (4) and had a value of 0.0891. Finally, the standard deviation of the sales growth of the drugs industry is 0.0192 which is reported in column (5). Subsequently, we find that the yearly investment growth of the drugs industry varies from -1.2073 to 1.2537 which is shown in column (1) and column (2). The column (3), column (4) and column (5) present that the mean value, median value and the standard deviation of the yearly investment growth of drugs industry are 0.1232, 0.1716 and 0.5255, respectively. Overall, the yearly stock return, sales growth and investment growth proliferate over time on an average level which can be known from the positive values of mean statistics in column (3).
4. Electronics Industry
The logarithm yearly stock return of the electronics industry, ranges from -0.5675 to 0.6442 for different firms, and there is also some variation in stock returns, which can be represented by the standard deviation, reported in column (5) and had a value around 0.2555 for each quoted company in the Chinese stock market. Also, the mean value and the median value of the logarithm yearly stock return of the electronics industry are 0.0274 and 0.0200, reported in column (3) and column (4), respectively. For the yearly sales growth of the electronics industry, it has the smallest value of 0.0103 and the largest value of 0.1611, reported in column (1) and column (2), respectively. Then, the mean value, reported in column (3), is 0.0728 which is almost the same as the median value, reported in column (4) and had a value of 0.0720. And the standard deviation of the sales growth of the electronics industry is 0.0385 which is reported in column (5). Subsequently, we find that the yearly investment growth of
the electronics industry varies from -1.3565 to 1.4392 which is shown in column (1) and column (2). The column (3), column (4) and column (5) show that the mean value, median value and the standard deviation of the yearly investment growth of electronics industry are -0.0077, 0.0434and 0.5269, respectively. The unique feature of the electronics industry is that it has the only negative value of the mean statistic among the five selected industries. In a word, based on the average scale, though it is less than one hundredth, the investment growth of electronics industry has reduced over the last decade.
5. Cars Industry
As shown in the table 2, the logarithm yearly stock return of the cars industry, varies from -0.6028 to 0.6758 for different firms, and there is also quite some variation in stock returns, as can be seen from the standard deviation, reported in column (5), which is around 0.2918 for each quoted company in the Chinese stock market. Meanwhile, the mean value and the median value of the logarithm yearly stock return of the cars industry are 0.0277 and -0,0008 which are reported in column (3) and column (4), respectively. The yearly sales growth of the cars industry has the smallest value of 0.0374 and the largest value of 0.1636, reported in column (1) and column (2), respectively. Also, the mean value, reported in column (3), is 0.0908 which is greater than the median value, reported in column (4) and had a value of 0.0862. In addition, the standard deviation of the sales growth of the cars industry is 0.0391 which is reported in column (5). Afterwards, we note that the yearly investment growth of the cars industry varies from -1.8258 to 2.2667 which can be known from column (1) and column (2). At the same time, the column (3), column (4) and column (5) indicate that the mean value, median value and the standard deviation of the yearly investment growth of cars industry are 0.0282, 0.0348 and 0.8017, respectively. Obviously, the yearly stock return, sales growth and investment growth grow over time on an average level because of the three positive values of the mean statistic of these three dependent variables for cars industry.
(ii) Independent and Control Variables
While the dependent variables vary among industries, the independent variables are the same, scilicet, they have identical characteristics.
Table 3: Descriptive statistics of independent and control variables
The table below shows the smallest value, largest value , mean value, median value and the standard deviation of the independent and control variables we use in the regressions. The five indicators are presented in column (1), (2), (3), (4) and (5), respectively.
Independent Variable Smallest Largest Mean Median Std.Dev.
(1) (2) (3) (4) (5) Shift 019 -0.0203 0.0006 -0.0087 -0.0091 0.0073 Shift 2029 -0.0257 0.0912 0.0090 -0.0014 0.0294 Shift 3039 -0.0223 0.0039 -0.0100 -0.0095 0.0076 Shift 4059 -0.0243 0.0207 0.0050 0.0049 0.0114 Shift 60 -0.0231 0.0212 0.0080 0.0104 0.0112 MR -0.4608 0.4254 0.0246 -0.0303 0.2276 Control Variable SMB -0.127 0.327 0.181 -0.009 0.141 HML -0.096 0.204 0.051 0.049 0.080 Rm-Rf -0.527 0.773 0.186 0.246 0.325
log population growth 0.002081 0.002609 0.002264 0.002207 0.000190
log GDP growth 0.04277 0.11201 0.07305 0.07090 0.02300
According to table 3, we notice that the smallest value of the shifts of the five cohorts varies from -0.0257 to -0.0203. For the largest value, cohort 2 which contains people
who are 20 to 29 years old has the most positive largest value (0.0912) among the five cohorts while cohort 1 has the least positive one (0.0006). The mean and median value of cohort 1 do not have much difference and they are -0.0087 and -0.0091, respectively. Identically, the mean and median value of cohort 3 and cohort 4 are almost the same which is around minus one hundredth for cohort 3 and five thousandths for cohort 4. Then the standard deviation of the five cohorts ranges from 0.0073 to 0.0294. The market return of the domestic stock markets has a smallest value of -0.4608, a largest value of 0.4254, a mean value of 0.0246, a median value of -0.0303 and a standard deviation of 0.2276.
As for control variables, SMB which denotes small minus big, serving as one of the three Fama French factors, has a smallest value of -0.127, a largest value of 0.327, a mean value of 0.181, a median value of -0.009 and a standard deviation of 0.141 while the five indicators of HML which is defined as high minus low are smaller than SMB’s except the median value which is reported in column (4) and has a value of 0.049. Foreign market excess return (Rm-Rf) fluctuates more conspicuously than the domestic market return. All of the absolute value of the five indicators are greater than the ones of MR. Additionally, log population growth’s five indicators except standard deviation are all on a thousandth scale, with a smallest value of 0.002081, a largest value of 0.002609, a mean value of 0.002264 and a median value of 0.002207. The standard deviation of the log population growth is 0.000190. Like the indicators for log population growth, the log GDP growth’s indicators except the largest value are all on a hundredth scale. The largest value of log GDP growth is 0.11201 which is on a tenth scale.
V. Results
In this section, we will talk about our results and examine the three hypotheses by industry separately.
1. Real Estate Industry
Table 4a: The influences of demographic shifts on return, sales growth and investment growth in real estate industry
This table looks at the determinants of logarithms of stock return (column 1-3), logarithms of sales growth (column 4-6) and logarithms of investment growth (column 7-9) for different quoted companies. In column (1), (4) and (7), they show the impacts of cohorts 2-5 on different dependent variables compared to cohort 1; in column (2), (5) and (8), we introduce time trend to control for the causal effects; in column (3), (6) and (9), we import additional macroeconomic indicators including log population growth and log GDP growth and two of Fama French factors, namely, SMB and HML as control variables to see whether the circumstances of the whole economy diversify our results? All column give the impact of domestic market return on the dependent variables. The table also reports the standard errors in parentheses and all the standard errors are clustered. The regressions use annual data after some manipulation from 2004 to 2013, including 1396 observations. And for the significance level, *, ** and *** denote significance at 10%, 5%, and 1%, respectively.
Dependent
Variable: Return Sales growth
Investment growth (1) (2) (3) (4) (5) (6) (7) (8) (9) Shift 2029 4.70*** 3.82*** 11.20*** 14.73*** 16.01*** 8.20*** 5.16* 5.48* 27.88 (0.48) (0.73) (2.89) (0.80) (0.54) (1.19) (2.77) (3.24) (24.47) Shift 3039 10.13*** 9.43*** 19.10*** 40.74*** 64.89*** 42.20*** 11.60 11.58 44.36 (0.92) (0.97) (4.64) (2.22) (1.00) (1.89) (7.71) (7,74) (39.26) Shift 4059 4.78*** 4.17*** 20.26*** 18.64*** 18.94*** 9.60*** 18.25** 7.26 57.39 (0.96) (1.26) (7.44) (1.59) (1.26) (0.82) (7.57) (9.27) (62.80) Shift 60 10.99*** 9.60*** 11.35*** -43.72*** -55.14*** -68.19*** -2.12 -8.06 -12.00 (0.91) (1.22) (1.30) (2.14) (1.20) (2.85) (6.72) (8.28) (9.04) MR 0.92*** 0.94*** 1.03*** 0.17*** -0.55*** -0.20*** -0.096 -0.015 -0.51 (0.03) (0.03) (0.05) (0.02) (0.01) (0.02) (0.14) (0.14) (0.41)
Time trend X X X X X X
Controls X X X
constant -0.043*** -0.080*** 0.1567 1.10*** 2.52*** -7.61*** -0.035 -0.123* 0.92 (-0.00) (0.02) (0.13) (0.04) (0.01) (0.07) (0.06) (0.07) (0.90) Observations 1396 1396 1396 1396 1396 1396 1396 1396 1396
In column (1), (2) and (3), we can see that all of the coefficients of the demographic shifts are significantly different from 0. We can know from column (1) that cohort 1 which contain people who are 0 to 19 years old (youth) has the least impact on the return of real estate industry, though the coefficient of cohort 1 is not shown in the table. People who are 20 to 29 years old (young workers) and 40 to 59 years old (steady workers) have about 4.7% more influence on the stock return than cohort 1 (youth). In other word, 1% increase in young workers and steady workers can raise the annual stock return of real estate industry around 4.7% more than 1% increase in youth. As for people who are 30 to 39 years old (mature workers) and more than 60 years old (elderly), their impacts are more than twice of the young workers and steady workers. Specifically, 1% increase in mature workers and elderly expands the stock return for real estate industry more than 10% than 1% increase in youth.
Then, when we put time trend into the regression, we find that all of the coefficients of the four cohorts decline to some extent, though still significant. The influence of elderly people is still the biggest among cohorts, however, it has only 9.6% more influence than youth. Finally, when we introduce controls into the model, the coefficients are still significant but have a really big change from the former results. Here we notice that the cohort 4 (steady works) has the biggest impact with cohort 3 (mature workers) ranking the second, cohort 5 (elderly) ranking the third and cohort 2 (young workers) ranking the last. The results make sense in reality world. The youth do not have that much money to buy houses. People demand more houses when they begin to work and accumulate their wealth. Therefore, the more time you work the
more chances for you to buy a house, caeteris paribus. For the elderly people they reach their retirement age so they do not demand as many houses as workers but still buy more houses than youth since the wealth they have enable them to do so.
Judging from the above result for stock return in real estate industry, hypothesis 1 is proved since all of the coefficients of the cohort group are significantly different from 0. This fact elucidates that the stock market is incomplete and has less information than managers who exploit this information.
From the sales growth aspect, the coefficients vary markedly when we import time trend and control variables in the regression. It indicates that the sales growth of real estate industry depends mainly on the macroeconomic circumstances and the year we analyze. We note that, as reported in column (6), the four coefficients of cohort group are all significant and self explanatory. That is, the mature workers contribute the most in the sales growth of real estate industry since 1% increase in this cohort can lead a 42% more increase in sales growth than the effect of youth. Young workers and steady workers demand more houses than youth but their influences are nearly one fifth of the mature workers. Meanwhile, 1% increase in elderly induces 68.19% less change in the sales growth than the youth. All these features provide an overall sketch of the current housing markets, you have more and more impacts on housing market before you grow up to be a mature workers and your impacts diminish over time very fast after 40 years old, especially when you reach your retirement age. Additionally, based on the results in the table, we claim that different cohort has distinctive consumption pattern in the real estate industry.
The investment growth of the real estate industry seems to be not sensitive to the shifts in demographics, especially when the time trend and controls are included in the model. We guess that the investment decision of the companies operating real estate is predominately determined by the government policies and the macroeconomic situations. Thus, hypothesis 3 is denied in this industry.
2. Foods Industry
Table 4b: The influences of demographic shifts on return, sales growth and investment growth in foods industry
This table looks at the determinants of logarithms of stock return (column 1-3), logarithms of sales growth (column 4-6) and logarithms of investment growth (column 7-9) for different quoted companies. In column (1), (4) and (7), they show the impacts of cohorts 2-5 on different dependent variables compared to cohort 1; in column (2), (5) and (8), we introduce time trend to control for the causal effects; in column (3), (6) and (9), we import additional macroeconomic indicators including log population growth and log GDP growth and two of Fama French factors, namely, SMB and HML as control variables to see whether the circumstances of the whole economy diversify our results? All column give the impact of domestic market return on the dependent variables. The table also reports the standard errors in parentheses and all the standard errors are clustered. The regressions use annual data after some manipulation from 2004 to 2013, including 152 observations. And for the significance level, *, ** and *** denote significance at 10%, 5%, and 1%, respectively.
Dependent Return Sales growth Investment growth Variable: (1) (2) (3) (4) (5) (6) (7) (8) (9) Shift 2029 4.69*** 4.84*** 5.27*** 0.19*** 0.21*** 7.87*** 3.52 4.33 12.44 (1.44) (1.53) (1.57) (0.02) (0.02) (0.22) (3.20) (3.60) (46.38) Shift 3039 1.38 1.51 1.58 0.34** 0.36** 12.44*** 11.61 12.24 25.14 (3.62) (3.70) (3.86) (0.17) (0.17) (0.31) (10.14) (10.58) (71.50) Shift 4059 -6.46* -7.33** -10.51*** -2.26*** -2.15*** -21.45*** -13.48 -8.96 -23.92 (3.52) (3.47) (3.50) (0.19) (0.19) (0.43) (9.73) (9.54) (114.32) Shift 60 -15.34*** -16.45*** -32.67*** -1.02*** -0.87*** -0.91*** -1.75 -7.63 -11.55 (2.24) (2.90) (4.37) (0.16) (0.15) (0.10) (7.07) (10.83) (12.58) MR 0.91*** 0.90*** 0.89*** 0.012*** 0.011*** 0.15*** 0.04 0.08 -0.36 (0.05) (0.05) (0.10) (0.00) (0.00) (0.00) (0.29) (0.31) (0.86) Time trend X X X X X X Controls X X X constant -0.08*** -0.07** -0.99*** 0.09*** 0.09*** 0.35*** 0.02 -0.05 0.36 (0.02) (0.03) (0.30) (0.00) (0.00) (0.00) (0.09) (0.15) (1.59) Observations 152 152 152 152 152 152 152 152 152
The stock return of the foods industry is affected by the demographic shifts visibly, except that the mature workers do not have a significant influence. When we introduce time trend to the model, the results do not vary a lot, while the results differ a lot from the original ones with controls and time trend in the regression. Seen from column (3), we conclude that the young workers have larger impact on the stock return of foods industry than the youth by 5.27% while the steady workers and elderly have smaller impacts by 10.51% and 32.67%, respectively. We assert that market has less information than managers in foods industry despite the fact that the mature workers do not much influence on the stock return.
The coefficients of each cohort are significantly different from 0 in regression (2), reported in column (4), (5) and (6). Identically, the sales growth of foods industry does not fluctuate a lot when we add only time trend into the model. Nevertheless, the macroeconomic features of the country and other markets do have a prominent impact on the sales growth. Three of the four coefficients of the cohort change a lot except cohort 5 which is constituted by elderly people. We know from column (6) that, the influences of young workers and mature workers on the sales growth of foods industry are 7.87% and 12.44% larger than the youth. While the steady workers have the least influence among cohorts and the elderly have approximately 1% less influence than the youth. We declare that different cohort has different consumption preference with the mature workers consuming the most and the steady workers consuming the least. These facts stay in accordance with the aforementioned analyses in the methodology part, that is to say, the people who are above 40 years old have less demand than the younger generations and the elderly people need more foods than steady workers.
At the same time, we find that demographic shifts do not much influences on the investment growth of foods industry. The reason might be the same as the real estate industry, namely, the investment decision depend mainly on the macroeconomic characteristics. That foods are needed by all the people and the total needs of food
may not fluctuate across time would also be a reason. Hence, hypothesis 3 is also denied in foods industry.
3. Drugs Industry
Table 4c: The influences of demographic shifts on return, sales growth and investment growth in drugs industry
This table looks at the determinants of logarithms of stock return (column 1-3), logarithms of sales growth (column 4-6) and logarithms of investment growth (column 7-9) for different quoted companies. In column (1), (4) and (7), they show the impacts of cohorts 2-5 on different dependent variables compared to cohort 1; in column (2), (5) and (8), we introduce time trend to control for the causal effects; in column (3), (6) and (9), we import additional macroeconomic indicators including log population growth and log GDP growth and two of Fama French factors, namely, SMB and HML as control variables to see whether the circumstances of the whole economy diversify our results? All column give the impact of domestic market return on the dependent variables. The table also reports the standard errors in parentheses and all the standard errors are clustered. The regressions use annual data after some manipulation from 2004 to 2013, including 909 observations. And for the significance level, *, ** and *** denote significance at 10%, 5%, and 1%, respectively.
Dependent Return Sales growth Investment growth Variable: (1) (2) (3) (4) (5) (6) (7) (8) (9) Shift 2029 -8.78*** -18.75*** -17.77*** -0.54** -0.76*** -0.31*** -3.39 -5.77 -10.72*** (1.32) (2.50) (5.65) (0.22) (0.02) (0.01) (5.78) (7.02) (2.17) Shift 3039 -8.80*** -10.44*** -35.58*** -1.33** -1.64*** -2.06*** -21.67 -3.12 -17.96*** (3.22) (2.58) (8.24) (0.58) (0.02) (0.02) (13.34) (6.35) (3.10) Shift 4059 5.65* 11.95*** 22.52*** 1.44*** 2.11*** 3.78*** 3.35 8.90 10.34* (3.41) (3.96) (3.45) (0.23) (0.01) (0.004) (17.32) (13.27) (6.11) Shift 60 27.93*** 34.95*** 66.78*** 2.82*** 7.67*** 5.64*** 6.00 14.62** 25.96*** (2.27) (3.34) (16.03) (0.64) (0.05) (0.024) (15.77) (6.61) (5.15) MR 0.75*** 0.73*** 0.23 -0.01 0.03*** -0.01*** -0.44*** -0.70*** -1.67*** (0.02) (0.04) (0.15) (0.01) (0.00) (0.00) (0.09) (0.08) (0.41) Time trend X X X X X X Controls X X X
constant -0.15*** 0.44** -1.42*** 0.10*** 0.21*** 0.45*** -0.03 -1.19** -1.49
(0.03) (0.20) (0.20) (0.04) (0.00) (0.00) (0.19) (0.49) (0.98)
Observations 909 909 909 909 909 909 909 909 909
The stock return of the drugs industry is influenced by the demographic shifts conspicuously. When we introduce time trend and controls to the model, the results vary a lot. This fact indicates that time trend and the macroeconomic circumstances play an important role in determining the stock return of the drugs industry. We conclude that the elderly people have the largest impact on the stock return of drugs industry as you can see in column (3). We assert that market has less information than managers in drugs industry since we can expect that the elderly people have enormous needs for drugs and ought to have a great impact to the stock market.
.
The coefficients of each cohort are significantly different from 0 in regression (2), reported in column (4), (5) and (6). The coefficients of the four cohorts vary a lot when we put time trend alone or together with controls into the model. This fact shows that the sales growth of drugs industry is affected not only by the time trend of every year but also affected notably by the whole economy. We know from column (6) that, the influences of steady workers and elderly on the sales growth of drugs industry are 3.78% and 5.64% larger than the youth. While the mature workers have the least influence among cohorts and the young workers have approximately 0.31% less influence than the youth. We allege that different cohort has different consumption pattern because we can imagine that the older you are , the more drugs you need according to our basic knowledge. Also, when you are not old , let us say, before 40 years old, the older you are, the stronger your body is. Therefore, you need less drugs when you are working than when you are a baby or teenager.
It is reasonable that as the aging problem becomes worse and worse, companies operating drugs foresee or keep in pace with this trend and invest more and more money in the research or development of new drugs or therapies. Furthermore, the companies will reduce the expenditures if more and more young people are born. But
basically, the most important indicator is the change in the cohort constituted by elderly people. You can see column (9) to get a direct feeling and more details about why the elderly matter a lot for the drugs industry. By the way, we should say that hypothesis 3 is accepted here since the demographic shifts affect stock return, sales growth as well as investment growth of the drugs industry and can provide useful information about the future trading strategy, sales and investment decisions.
4. Electronics Industry
Table 4d: The influences of demographic shifts on return, sales growth and investment growth in electronics industry
This table looks at the determinants of logarithms of stock return (column 1-3), logarithms of sales growth (column 4-6) and logarithms of investment growth (column 7-9) for different quoted companies. In column (1), (4) and (7), they show the impacts of cohorts 2-5 on different dependent variables compared to cohort 1; in column (2), (5) and (8), we introduce time trend to control for the causal effects; in column (3), (6) and (9), we import additional macroeconomic indicators including log population growth and log GDP growth and two of Fama French factors, namely, SMB and HML as control variables to see whether the circumstances of the whole economy diversify our results? All column give the impact of domestic market return on the dependent variables. The table also reports the standard errors in parentheses and all the standard errors are clustered. The regressions use annual data after some manipulation from 2004 to 2013, including 1165 observations. And for the significance level, *, ** and *** denote significance at 10%, 5%, and 1%, respectively.
Dependent Return Sales growth Investment growth Variable: (1) (2) (3) (4) (5) (6) (7) (8) (9) Shift 2029 3.88*** 11.87*** 12.34*** 1.89*** 3.42*** 4.80*** 8.73*** 8.12*** 13.36* (0.55) (1.05) (1.72) (0.03) (0.01) (0.02) (1.15) (2.46) (7.06) Shift 3039 -1.76 -6.16*** -5.63*** -0.34*** -0.18*** -0.10*** 6.68 5.94* 6.69*** (1.48) (1.56) (0.84) (0.06) (0.01) (0.02) (4.09) (3.56) (1.81) Shift 4059 -8.43*** -7.07*** -9.47*** -1.09*** -0.96*** -1.36*** -1.50 -1.55 -2.15*** (1.35) (2.40) (1.07) (0.09) (0.01) (0.02) (3.09) (10.37) (0.55) Shift 60 -14.02*** -18.80*** -31.69*** -3.91*** -6.70*** -4.92*** -14.88*** -26.36*** -31.05***
(1.11) (1.44) (4.71) (0.06) (0.03) (0.03) (3.68) (4.85) (10.32) MR 0.85*** 0.76*** 0.16 0.01*** -0.03*** 0.05*** 0.72*** 1.12*** 1.38*** (0.03) (0.03) (0.15) (0.00) (0.00) (0.00) (0.06) (0.11) (0.17) Time trend X X X X X X Controls X X X constant -0.11*** 0.31*** -1.47*** 0.12*** 0.01*** -0.24*** -0.04 2.08*** 3.25*** (0.01) (0.08) (0.12) (0.00) (0.00) (0.00) (0.03) (0.30) (0.43) Observations 1165 1165 1165 1165 1165 1165 1165 1165 1165
The stock return of the electronics industry is affected by the demographic shifts notably. When we import time trend and controls to the model, the results vary a lot. This fact illustrates that the stock return of electronics industry has a clear tendency that related to the time horizon and is influenced markedly by the outside factors like GDP growth rate. We notice that the elderly people have the largest negative impact on the stock return of electronics industry as you can see in column (3). The stock return of electronics industry will drop a lot if elderly people are more than before. We claim that market has less information than managers in electronics industry because four cohorts have significant influences on the stock return. As you can see in table 4d, 1% increase in cohort 3 to 5 has a 5.63%, 9.47% and 31.69% less impact than 1% increase in cohort 1 while cohort 2 has more impact than cohort 1 by 12.34%.
The coefficients of each cohort are significantly different from 0 in regression (2), reported in column (4), (5) and (6). The coefficients of the four cohorts vary a lot when we put time trend alone or together with controls into the model. This fact indicates that the sales growth of electronics industry is affected not only by the time trend of every year but affected visibly by the macroeconomic factors as well. We know from column (6) that, the influences of mature workers, steady workers and elderly on the sales growth of electronics industry are 0.10%, 1.36% and 4.92% smaller than the youth, respectively. The young workers have a greater influence than
the youth. We conclude that different cohort has different consumption pattern because we can imagine that the older you are, the less electronics you need, especially when you have retired, like my grandparents who are still not able to use smart phone. On the contrary, when you are young, you would like to keep in pace with the popular trend or the fashion, like the apple mania in 2011. Thus, you could imagine why the youth and the young workers have intense needs for electronics.
As for the investment growth in electronics industry, we can conclude from the results in column (7) to (9) that the investment growth are determined by the demographic shifts to some extent. Companies will invest more if there are more young workers and mature workers and less steady workers and elder in the society. Specifically, 1% increase in the young workers and mature workers will raise the investment growth of electronics industry 13.36% and 6.69% more than 1% increase in youth. And we can see that young workers have the greatest positive impact while the elderly have the greatest negative impact. By the way, we note that the coefficients vary a lot if we add time trend and controls into the regression. This fact shows that the investment decision of the electronics industry is influenced by the time and macroeconomic circumstances to a large extent. Additionally, hypothesis 3 is accepted here.
5. Cars Industry
Table 4e: The influences of demographic shifts on return, sales growth and investment growth in cars industry
This table looks at the determinants of logarithms of stock return (column 1-3), logarithms of sales growth (column 4-6) and logarithms of investment growth (column 7-9) for different quoted companies. In column (1), (4) and (7), they show the impacts of cohorts 2-5 on different dependent variables compared to cohort 1; in column (2), (5) and (8), we introduce time trend to control for the causal effects; in column (3), (6) and (9), we import additional macroeconomic indicators including log population growth and log GDP growth and two of Fama French factors, namely, SMB and HML as control variables to see whether the circumstances of the whole economy diversify our results? All
column give the impact of domestic market return on the dependent variables. The table also reports the standard errors in parentheses and all the standard errors are clustered. The regressions use annual data after some manipulation from 2004 to 2013, including 567 observations. And for the significance level, *, ** and *** denote significance at 10%, 5%, and 1%, respectively.
Dependent Return Sales growth Investment growth Variable: (1) (2) (3) (4) (5) (6) (7) (8) (9) Shift 2029 7.89*** 9.76*** 11.93*** 2.50*** 2.38*** 2.65*** 0.93 0.93 17.37 (0.83) (1.34) (3.72) (0.18) (0.10) (0.15) (2.59) (2.51) (15.43) Shift 3039 11.60*** 17.77*** 48.11*** 4.54*** 5.74*** 8.43*** 3.48 6.20 42.63 (1.98) (1.92) (8.83) (0.28) (0.24) (0.37) (8.80) (9.29) (40.27) Shift 4059 11.06*** 5.30** 27.00*** 4.29*** 4.21*** 7.39*** 3.24 5.59 24.95 (1.97) (2.17) (5.77) (0.30) (0.10) (0.25) (8.96) (25.95) (22.46) Shift 60 -3.29* -4.2 -1.61*** -0.56*** -0.83*** -1.52*** -1.63 -3.86 -1.03 (1.97) (3.60) (0.25) (0.11) (0.10) (0.07) (8.08) (11.60) (12.11) MR 0.65*** 0.95*** 0.63*** -0.07*** -0.11*** 0.07*** -0.14 -0.10 0.09 (0.03) (0.04) (0.07) (0.00) (0.00) (0.00) (0.17) (0.21) (0.44) Time trend X X X X X X Controls X X X constant -0.26*** 0.001 -1.00*** 0.13*** -0.07*** 0.18*** -23.37*** 0.52 0.29 (0.02) (0.12) (0.18) (0.00) (0.01) (0.06) (5.48) (0.76) (1.08) Observations 567 567 567 567 567 567 567 567 567
The stock return of the cars industry is affected by the demographic shifts visibly. When we introduce time trend to the model, the results vary a lot and the results differ a lot from when we add controls and time trend together into the regression. Known from column (3), we conclude that the mature workers have larger impact on the stock return of cars industry than the youth by 48.11% while the steady workers and young workers have larger impacts by 27.00% and 11.93%, respectively. Meanwhile, the elderly has the smallest effect. We conclude that market has less information than managers in cars industry since all the coefficients of the cohort are significant different from 0.
As for sales growth, the coefficients of each cohort are significantly different from 0 in regression (2), reported in column (4), (5) and (6). The sales growth of cars industry does not fluctuate a lot when we add only time trend into the model. However, the macroeconomic factors have a prominent impact on the sales growth. Three of the four coefficients of the cohort vary a lot except cohort 5 which is constituted by elderly people. We know from column (6) that, the influences of young workers, mature workers and steady workers on the sales growth of cars industry are 2.65%, 8.44% and 7.39% larger than the youth, respectively. While the elderly have the least influence among cohorts and they have approximately 1.52% less influence than the youth. We allege that different cohort has different consumption pattern with the mature workers consuming the most and the elderly consuming the least. These facts stay in accordance with the aforementioned analyses in the methodology part, that is to say, the people who are above 60 years old have less demand than the younger generations and the majority of the cars are purchased by people who are working.
We find that demographic shifts do not much influences on the investment growth of cars industry. The reason might be the same as the real estate industry, that is, the investment decision depend mainly on the macroeconomic characteristics. Therefore, hypothesis 3 is also denied in cars industry.
VI. Robustness Check
As stated before, here we present the results of robustness check to see whether the regression models work or not. We use Fama French factor (Rm-Rf) instead of MR to analyze how foreign market excess return influence our results.
We know from Appendix Table 5 that when we replace MR by (Rm-Rf) in the regression, the robustness of the coefficients do not change. However, almost all of the standard errors in the parentheses proliferate conspicuously since (Rm-Rf) fluctuates more visibly than MR. The economic meaning of each coefficient do not change since the shift in the coefficient is not significantly. We can test by the following formula: ) , .( 2 1 2 2 2 2 1 2 1 coeff coeff Cov se se Coeff Coeff Diff
If the absolute value of the “Diff” above is larger than 1.64, namely, the t-statistic value which indicates that the difference between the two estimators is significantly different from 0 at confidence level of 10%. After this test, we find that all of the “Diff” is not significant. Therefore, we can conclude that when we replace MR in the regression by Fama French factor (Rm-Rf), the coefficients of the four selected cohorts do not change.
VII. Conclusions
We have conducted a research about how do the stock return, sales growth and investment growth of the quoted companies operating real estate, foods, drugs, electronics and cars in the Shanghai Stock Exchange and Shenzhen Stock Exchange react to the demographic shifts in the last decade. We divide the people into five cohorts and defined them as youth, young workers, mature workers, steady workers and elderly. Then, we test the impacts of the demographic shifts of four selected cohorts (we omit youth to avoid the collinearity problem) on the stock return, sales growth and investment growth by panel regressions with time trend and control variables. Subsequently, we inspect the results if they are the same as we expect, give economic meanings to our findings, and analyze the possible reasons for some insignificant outcomes. Finally, we do a robustness check to verify our results by replace one of the variables in our regressions.
We have found that for all of the five selected industries, hypothesis 1 and hypothesis 2 are accepted. We prove that market has less information than managers since demographic shifts have significant impacts on the stock return. Also, we notice that demographic shifts affect the sales growth of the selected industries to a large extent which verify that different cohort has distinctive consumption pattern. For example, the more elderly people in the society, the less is the sales growth of real estate, electronics and cars industry. However, the sales growth of drugs industry would proliferate markedly if there is more steady workers and elderly people and less young people who are under 40 years old in the society. All these findings are in accordance with what Della Vigna and Pollet(2007, 2013) have found in their research where they claim the age profile of consumption differs significantly across goods. Further more, the demographic shifts can be used to predict the stock return, foresee the sales growth and formulate the investment decision of drugs and electronics industry. Demographic shifts cannot provide reliable information for real estate, foods
