How Could Age Dependency Ration Influence The Savings And Following Capital Outflow.
Introduction.
The aim of my work it to analyze how the age dependency ration influences the saving. According to the definition of the World Bank, Age dependency ratio is the ratio of dependents-people younger than 15 or older than 64--to the working-age population--those ages 15-64. So the questions of my research is to find out how the quantity of pre-working-aged children and retired influences the saving per capita. Would people save more in their working age in order to save money for retirement period or to invest -or they will earn and spend. Another issue is to see if countries with huge level of GDP( China, the USA) have the different extent of ADR influence on savings and how it is differ from Russia and ,for instance, Brazil. This problem is actual and important because nowadays countries face the problem of capital outflow . As we can see from the current statistics databases, there were such demographical crisis since the 1950 , especially in Asian countries and because they are mostly overpopulated, it has an impact on the world situation in general. According to the theoretical economic models, Investment=Savings, so if I define what impact have demographical structure of population on saving per capita I could predict the future saving and investment in countries using the obtained results of influence.
The model.
To define, what could also has the impact on savings and what I could add to my regression mode, I analyzed some economical articles and papers. In that case, I started by reading the research work of F. Modigliani, famous Italian economist and Nobel-awarded, about the conception of the life-cycle model. Working out this conception, he focused on the fact that the Income level is fluctuating during the whole life of every person and that the saving allows them to distribute their income by period of high and low income. It means that people try to save some money to spend it after they get retired. The main aim of his model was to explain how the saving and consumption are formed. But the weak part of his research could be the inflation and the expectance age of retired. A.Tylecote, the British economist, in his research showed that these effects of inflation and some another economical indicators could remarkably influence the demand function and could provoke economics cyclic oscillations taking the changing in demographical structure of the society into account. Then I read the book , written by Obstfeld M. and K. Rogoff. In their intertemporal account balance models of opening economies and closed economies and two period economies they gained that indicators such as real interest rate, substitution effect, income effect and welfare effect influence savings in the most extent.
In this way , thanks to the empirical analyses I want to see whether the influence of ADR on saving is significant or not. Also I want to see the individual effect of every country on savings and is it significant or not. In addition I will define whether ADR in general is necessary or we could use only retired people or only under labour age children in regression model.
Firstly , I will start by considering the regression with not fixed effect . I defined the panel structure of variables and use the command : Xtreg.

The regression with no fixed effect differ from fixed effect that in equation , ui- normal distributed variable, not correlated with εit . We can see the estimation of ADR_GEN and another variables, but not exactly to each country. According to that regression, the coefficients by real_ir and gdp_pc are significant. Sigma u- it is the standard deviation of coefficients by dammi- variables by countries. Dispersion coefficients is higher than dispersion residues. It means that influence of individual country is high . When the real interest rate will rise by 1 that will lead to the fall in savings by -0,14 and if ADR_GEN rised - fall of savings by -0,09, and meanwhile the GDP increase provokes slight rise. The correlation between regressors and between regressor and residues could give us the proposal that the estimates are biased and because of that it is more efficient to use fixed effect model. In the fixed effect model corr(ui, Xi) might be not equal to 0, because here ui is not unbiased error .here it is included in the regressive function. where Zi - not observed variables , which didn't change for i-country. Every country will have each own meaning- will not change in dynamic(no t-index)
So we can estimate this model : . In this case, every country has the same β1, but they each of them has its own constant variable.
I make the fixed-effect regression , where I fix effect for each country

2-93 - it is a country code, where reference country is Afghanistan by alphabet.
Moreover, it should be noted that the dependence from the year could be nonlinear and that for every country the savings could vary in 2008, when there was a financial crisis, from 2012m when there was a recovering period. So I also should create a regression function taking fixe year effect into account.
There is a model:
Yit=β0+β1Xit+β2Zi+εit+Vt
Each country has its own meaning that will not change in the dynamic.

So, we can compare 2 models with fixed effects- where we fixed country ans where we took country and year into account. When we fixed effect of the country, the coefficient by ADR starts be more significant and became positive. According to the table, the growth of ADR_GEN by one provokes the increase of ADJSAV by 0,33. In the model where both countries and code are fixed the coefficient by Adr_Gen is not significant.( it equals 0, 272).
We also can compare it with not-fixed model. As we could noted , in the not-fixed model the coefficient by adj_sav stops being significant. It means that probably our expectancy of correlation between ui and regressor is correct. Thanks to that we should prefer the fixed-effect model in order to get more plausible result. We also can conduct the Hausman test to see whether the regression model is consistent or not.

Because the Ho is not accepted, the fixed effect model is consistent by alternative hypothesis, and the not-fixed regression model- not. Due to that fact we choose fixed-effect model.
Moving on to the next point , I would like to find out if there is no difference between the influence of ADR on savings in Russia and in other countries. In order to do that , I generate new fixed variable ADERUS=ADR_GEN* rus, where rus=1 if the c.code is Russia and ) in other cases. Then will have a look, how the coefficient by AGERUS will change.

The coefficient by arerus is not significant, so I can make a conclusiob the ADR has the same influence in russian as in the world in general. Then I want to compare it with a country with big economy, for example The USA. I genereated the same variable as for Russia, but using the effect of the USA.

The coefficient by ageuse is significant so I can conclude that the influence of ADR in the USA significantly differ from its influence in other countries. In the general model the coefficient is positive and makes saving for 0,33 while the rise of ADR, whereas in the USA the increase of AGEUSA by one leads to the decrease of savings by 0,263.
Then I made the same estimation for Brazil- the country with big economy but weak, and for China- the biggest economy nowadays. The coefficient by agebra isn't significant so I can conclude that in Brazil, as in Russia, the influence of ADR on savings doesn't differ from general model. Then I did the same for China. The coefficient by AGECHI is high- while the increase of afechi by one ,the savings will rise for 0,75.
On the one hand , the effect is not seemed- if to fix the country effect, we have the same coefficient for every country. On the other hand, the analyse the influence of ADR in choosed country showed, that in countries with big economies like the USA and Chine, and Brazil m where economy is falling down during the last 5 years, and Russia ,the effect could both be differ or not. It means that influence of that coefficient in some countries is more remarkable than in others.
Then I want to find out if there is a significant difference in coefficient by ADR-OLD and ADR_Young. So now I generate fixed effect model but instead of ADR_GEN I used divided variables.

(* Here I published only the first part of the table). Coefficients by old and young are different- one is positive and other is negative one. Because of that I need to gen 2 different regressive model- in one I will take only ADR_Young and in other only ADR_OLD to see, whether the impact on saving significant or not.

In this way , deviding one regression model into two showed that ADR_OlD influences significantly on saving, whereas the coefficient by ADR_Young is not significant. Regressors ARD_OLD and ADR_young are giving the ADR_GEN- it means they are multicolleniar and we need to take only one into account. Therefore I choose only one model with significant coefficient- the model with ADR_OLD. The Data. For empirical analyses I used data from the site World Development Indicators. I took date for every country but there was no data for some indicators for some countries or in some periods. As a result, I obtained the sample of 93 countries. I analyzed periods from 2008 to 2012. I used such variables: The adjustment savings per capita(ADGSAV), general Age dependency ratio(ADR_SAV), then I used the ADR_OLD and ADR_YOUNG- Age dependency ratio only of retirement and only of young. Then I also used the indexes of the real interest rate(REAL_IR) and GDP per capita (GDP_PC). Empirical results. My aim was to find out how the demographical structure influences the saving in the world and how it differs for different countries and also to see how it differs for countries with big economies and weak economies. The empirical analyses showed that firstly, the proposal that in countries with huge economies, that finished demographical cycle , where there is demography crisis(China) and there is a tendency of demographical crisis(The USA) , the coefficient ADR differs from the influence of the coefficient in general- is correct. Secondly, ADR influences the saving, therefore the changes in demographical structure influences the saving. Thirdly, coefficients by Old and Young are different that means that they will have different extent of impact- the influence of the coefficient by ADR_OLD is more significant. It proved the proposal that different structure groups influence in different ways . Fourthly, the influence of ADR in Russia doesn't differ from the general impact in the world, whereas in China there is a tendency to the further demographic bias and crisis that will lead to the economical crises. Conclusion. To sum up I should say that according to the analyses , the increase of ADR influences saving , so savings per capita will decrease. To implicate the theoretical and practical part it seems that people consume in different ways, depends on age and life expectancy and the real interest rate and the GDP per capita. The bigger part of retirement and children makes the saving less - less income , less saving, less investment. By this we can say , that older people have more savings if there is no demographical crises in the country. References 1. Campbell ,John Y and N. Gregory Mankiw. The Response of Consumption of Income- A Cross-country Investigation//European Economic Review.1991.№ 35. P.723-767 2. Modigliani F. Life Cycle, Individual Thrift and the Wealth of Nations //American Economic Review. 1986. № 76. P. 297-313 3. Obstfeld M. and Kenneth Rogoff. The Intertemporal Approach to the Current Account.//Handbook of International Economics. 1995. vol. III, Elsevier 4. Tylecote A. The Long Wave in the World Economy. L .: Longmans , 1992 .