THE LATENCY EFFECT IN LIFE INSURANCE

The Latency Effect in the Demand for Life Insurance in the United States Chengran Ouyang University of Western Ontario

THE LATENCY EFFECT IN LIFE INSURANCE

Abstract In this paper, we compare the effect of latency of some factors, which affect households life insurance purchasing decision, of the Ordinary Least Square (OLS) model. The data used in this study all take from the statistical abstract of the United States. The thinking behind this experiment was to figure out whether number of births and number of marriages have lag effects on the sales of life insurance and how many years these lag variables are usually taken. Thus, we did some experiment with various models and found out these two variables (birth and marriage) do have different level of hysteresis quality on the sales of life insurance. The lag effect for number of births turns out to be two to three years and the lag effect for number of marriages is three years. By taking the effect of latency of these two variables in to account, the new demand function for life insurance would be more concise and accurate. Life insurers could more precisely anticipate the demand of life insurance in the future.

THE LATENCY EFFECT IN LIFE INSURANCE

I.

Introduction

Life insurance is a type of pooling and hedging of risks, which are the universal principles of risk management. By pooling the risks, the benefits of insurance are that nobody will suffer alone. Whatever happens to someone, it is spread out over all the applicants in this insurance company. If the firm is big enough, it will end up unobservable. It sounds like an efficient financial instrument to reduce the fiscal effect of someones misfortune of losing his or her relatives; however, life insurance Industry is suffering sluggish development because American insurers do not have a really practice and effective demand function for life insurance due to latency effect of some variables have not been taken into consideration. In order to solve these troubles, insurance companies need to have a more concise and accurate demand function to forecast the sales of life insurance by deeply understanding the lag effect of some factors. Facing the current situation, the percent of U.S. households life insurance has continually dropped over the past two decades until recent years. Ownership of individual life insurance, which covers a single life, declined more than ten percent by the end of 2004. (Li, 2008) According to Life Insurance Marketing and Research Associations report (2006), the percentage of insured households owning permanent life insurance decreased during the 1992 to 2004 period. Eighty percent of households with individual life insurance carried some permanent life insurance as part of their portfolio of life insurance in 1992, while only 64 percent of insured households carried permanent life insurance in 2004. In the mean time, the percentage of group life insurance, which covers many lives that are members of the group, the ownership by household has barely remained at the same level. (Li, 2008) The purchase of life insurance is one of the most important purchasing decisions for individuals and families (Anderson & Nevin, 1995) and it is a critical component of a long-term

THE LATENCY EFFECT IN LIFE INSURANCE

financial plan (Devaney & Keaton, 1994). Even though approximate 75 percent of Americans believe that life insurance is one the best way to protect against the premature death of a chief income earner, the report in 2006 prepared by Life Insurance Marketing and Research Association (LIMRA) reveals that consumers consider the purchase of life insurance to be a complex process and eight in ten find it difficult to purchase life. Therefore, it is even harder for life insurance companies to anticipate the demand for life insurance. The primary goal of this study is to find out whether number of births and number of marriages have lag effects on the sales of life insurance and how many years these lag variables are usually took. Since life insurers need to have much more precise estimation of the demand of life insurance to understand household life insurance purchase decision, the results in this study will enable life insurers to better understand customers life insurance behavior and have a new demand function for life insurance, which would be more concise and accurate. Hence, life insurers could more precisely anticipate the demand of life insurance in the future. The remainder of this paper is organized as follows. Section II is the literature review that mainly discusses the previous researches for this study or related to this study. Section III shows the data selecting process and the feature of these data. Section IV describes the ordinary least square (OLS) model and explains the reason why we choose this model and how to set up the models. Section V presents the results and the interpretation of the results. Section VI offers a summary of our findings.

THE LATENCY EFFECT IN LIFE INSURANCE

II.

Literature Review

This section reviews prior researches related to the demand for life insurance. This part includes 1) Demand for life insurance model for forecasting was made in previous study and discuss the factors that affect household insurance purchasing decision. 2) By comparing the differences between Mexico and the United States and analysing international demand for life insurance, we demonstrate the purchasing feature of the United States. In order to help insurance companies anticipate the demand for the sales of life insurance in the future, they need to have a cheap, quick and logical method of estimating demand in advance. According to the research conducted by Headen, R. S., and Lee, J. F.(1968), they created a model - Q=f(X1, X2, X3, X4, X5). The variables used in our demand forecast model were selected for three basic reasons: general availability, easy predictability, and relevance. Here are the variables selected. Q is total sales of life insurance per year that is the dependent variable. It is also the only available data for sales, which is taken from the Life Insurance Fact Book. From X1 to X5 are average price of life insurance, number of marriages, personal income, residence population. Number of marriages: it has lag effect, which assumed to be eight years in this study. Same thing for number of birth but the lag is supposed to be 2 years All of these variables are easily obtainable for many years. Total life insurance sales were obtained from the Life Insurance Fact Book for years 1929-1964. The other data were readily available in various editions of the Statistical Abstract of the United States, for the same period. The Demand Equation The computer program used automatically computed an equation of the form Q = a + bX1 + cX2 + dX3 + eX4 + fX5

THE LATENCY EFFECT IN LIFE INSURANCE

By using the data from 1929-1964 data, here is the result Y=6878.54+7414.79X1 1363.66X2 1254.61X3 + 0.34578X4 + 472.99X5 Then if we are going to forecast next year, we have already got the parameters and since birth and marriage are lag variables, we can still use past data for these two in the forecast function. Other three variables are widely and quite accurately predicted by many groups of people. We can just use the predicted value from them to run the forecast. The forecast results seem to be pretty concise and accurate. On the other hand, it is also very crucial for insurers to know which factors would influence households demand for life insurance. Li, M. (2008)s paper - Factors Influencing Households Demand for Life Insurance aims to examine both the type and amount of life insurance purchased by households. A comprehensive models of households demand for life insurance were developed in the end, which included demographic variables (age, education, employment status, health status, number of children, marital status, and race), economic and assets variables (income, homeownership, debts, as well as portfolio elements such as liquid assets, certificates of deposit, mutual funds, bonds, stocks, individual retirement accounts, annuities, other miscellaneous financial assets, and nonfinancial assets), and psychographic variables (attitude toward risk, attitude toward leaving a bequest and ones expected life expectancy). The effects of these factors on either term or cash value life insurance purchased by households were examined separately. The results of the two-stage model indicated that some variables in the likelihood of purchasing life insurance model and the amount of life insurance model different in their significances. In addition to race, life expectancy, CDs, and annuities, all other hypothesized factors had significantly positive or negative impacts on term life insurance demand of

THE LATENCY EFFECT IN LIFE INSURANCE

households. Employment of head, race, and life expectancy did not significantly affect cash value life insurance demand of households, while other factors were shown to be significant contributors. This study provides three contributions. First, the results proved that most of assets categories associate with the purchase of life insurance by households. Second, using Heckman two-stage selection model is supported in this study because factors influenced the probability of owning life insurance and the amount life insurance held were different. Finally, the fact that variables associated with the demand for term life insurance and the demand for cash value life insurance were different support the view that term life and cash value life insurance should be examined separately. Another research in household behaviour is from Headen, R. S., and Lee, J. F. (1974). Life Insurance Demand and Household Portfolio Behaviour discusses the influences of short-run financial market behaviours and consumer expectations on purchases of ordinary life insurance. In this paper, analysing household portfolio behaviours is the crucial tool to figure out the structural determinants of life insurance demand. The authors recommend and estimate a model that can interconnect demand for various financial assets. From my perspective, the purpose of this article is to relate life insurance sales to the financial market. The relationship is completely presented and analyzed in this paper. The results resoundingly modify the previous demand function that has been proven to be unreliable. Due to the importance of life insurance companies in the economy, the demand for life insurance requires more systematic study and analysis. Hence, this research surely gives useful references to all the insurance firms. To manifest the feature of the demand for life insurance in the United States, it will be extremely helpful to compare the demand for life insurance with other countries. In the first

THE LATENCY EFFECT IN LIFE INSURANCE

place, based on Truett, B., and Truett, J. (2010)s study - The Demand for Life Insurance in Mexico and the United States: A Comparative Study, it compares the difference between demands for life insurance in Mexico with that in the United States. From a historical perspective, it demonstrates on the growth of life insurance purchases in the two countries and employs regression analysis to estimate life insurance demand functions. The principal results are the age, education, and level of income affect the demand for life insurance and the income elasticity of demand for life insurance are much higher in Mexico than in the United States. The most outstanding observation one can make when comparing the estimated demand functions for Mexico and the United States is that the estimated income elasticity of demand for life insurance is much greater for Mexico than for the United States. In fact, this estimated elasticity was more than three times as great for Mexico as for the United States in all corresponding cases. Since income elasticity of demand for life insurance is very low in the United States, income should not have much impact on the sales of life insurance. This result would be consistent with the hypothesis that the income elasticity of demand for life insurance is much higher at lower levels of income than at higher income levels. Such a situation would seem reasonable since a high income family would likely have accumulated greater wealth to protect the standard of living of the family if a major income-earner died. Nonetheless, much more studies have been conducted in this area. The paper - An International Analysis of Life Insurance Demand from Browne, M. J., Kim, K. (1993) is one of them. The service sector of the world economy has grown substantially since World War II. The worldwide insurance industry has had an average annual growth rate of over 10% since 1950. During the mid-1980s, the international life insurance industry grew at an average annual rate greater than 25%. This article identifies factors that lead to variations in the demand for life

THE LATENCY EFFECT IN LIFE INSURANCE

insurance across countries. Important factors are found to be the dependency ratio, national income, government spending on social security, inflation, the price of insurance, education, life expectancy and whether Islam is the predominant religion in a country. To summarize, these literatures and previous researches give us extremely precious resources to build up our model. From Headen, R. S., and Lee, J. F.(1968)s study, we got the basic model of this paper. According other literatures, we edited and changed some of the factors so that it is more suitable for the situation in the United States at present. For example, the personal income is not a good variable to compute the demand for life insurance in the United States, based on Truett, B., and Truett, J. (2010)s study. They found out the income elasticity for life insurance is very low in America, which indicated that variable does not have much impact on the sales of life insurance. Under the influences of all the literatures, our model was created following deliberate logical and reasonable assumptions.

THE LATENCY EFFECT IN LIFE INSURANCE

10

III.

Selection of Variables

The two major rules to select variables for the demand forecast model: general availability and relevance. The first requirement is very important because large financial resources are extremely necessary to generate new data. Fortunately, there are enough data for this model, have been published. Except average price of life insurance, other variables are easily obtainable for many years. They were obtained from the statistical abstract of the United States for year 1979-2007. What is more, the scopes of all the data in this study cover the whole country. Variables selected were as follows: Q (billion dollars) is total sales of life insurance per year that is the dependent variable. Table 1. The line and symbol graph of the sales of life insurance in the United States

THE LATENCY EFFECT IN LIFE INSURANCE

11

As the diagram shows, the trend of demand for life insurance is generally keeping raising up since 1979 to 2000. Even though it was suffering sluggish development in 1980s, it goes up and back to the trend after that. Since the year of 2000, it is suffering sluggish development again. Table 2. Histogram and Statistic of the sales of life insurance

Both mean and median are around 94 and the mode is between 70 and 80. In Table 1, the sales of life insurance were stuck in 1980s and the sales are also around 70 to 80 billion dollars. Recall American history, during this period of time, America are sustaining sever inflation and exploded federal debts due to blind believing in the theory of the supply side of economics and call off several necessary taxes. Moreover, mistaking the current position of the Laffer-curve

THE LATENCY EFFECT IN LIFE INSURANCE

12

made the economy even worse at that time. Thus, we can get this conclusion that the sales of life insurance is highly positive correlated to the economy. X1 (dollars) is average price of life insurance is the only variable does not get directly. Since we have already got the sales of life insurance and the number of life insurance are sold in each year, we can simple use the total sales divided by number of life insurance to get the average price. Even though it seems to be reasonable and should not be a big problem, there are still potentially some inaccuracies for this data and it may lead our final result biased. X2 is number of marriages (1000): married men (or women) are supposedly better insurance buying prospects than single men (or women), and as total marriages increase, life insurance sales should also increase. This variable might lagged several years. In our models, the lag was separately assumed to be zero, two, three, six and eight years. If the lag is bigger than 0, we will use the past data to replace the current one. This problem of lags for demographic variables will be discussed in more detail below. X3 is number of births (1000) in this model. It has been argue for many years that the time to sell a man (or woman) an insurance policy is just after he or she has had a new child. Actually, it turns out that several years after the birth of a child. We experimented some years in the model. According to our model, two or three years is proved to be the best time. X4 is college education (1000), which is measured by both private and public college students enrollment. Generally, people with higher education in the United States are more likely to have safer jobs and wont threaten their lives in some dangerous situations. Therefore, they have less incentive to purchase life insurance than others. For these people who received less education, however, are more likely to have jobs such as driller or farm worker. They are

THE LATENCY EFFECT IN LIFE INSURANCE

13

working at more danger condition. Therefore, less educated people in the United States are more willing to buy life insurance than well educated people. X5 is resident population (1000). The larger the population, the more insurance will be sold. Due to if there are more people, there are more potential consumers in the market and thus, insurance companies have more opportunities to sell their products.

THE LATENCY EFFECT IN LIFE INSURANCE

14

IV.

Models and Methods

In this paper, we use Ordinary Least Square (OLS) to run the model. Essentially, our Ordinary Least Square (OLS) analysis took the form of an equation as follows: Q=f(X1, X2, X3, X4, X5), where Q is the sales of life insurance of U.S. companies(the dependent variable), and X1 though X5 are independent variables, which directly affect the quantity of insurance sold. Here is our Ordinary Least Square (OLS) model: Y = 0 + 1X1+ 2X2+ 3X3 +4X4 + 5X5 + u Tobit, Probit and Logit models are impropriate in this situation because these models require large sample size, which is very hard to compare each year. What is more, the Tobit model has a notable limitation. In the Tobit model the same set of variables and the same coefficients determine both the probability that an observation will be censored and the value of the dependent variable. That is, the Tobit model does not allow for the differences between the decision of whether to purchase life insurance and the decision about the amount of purchase. Previous research has used ordinary least square (OLS) to forecast the sales of life insurance. Using OLS regression produces unbiased parameter estimates when the dependent variable and one of the independent variables are correctly recorded. On the other hand, the ordinary least square (OLS) appears reasonable. In the paper demand for life insurance, one of the demand projections, which is conduct by Headen, R. S., and Lee, J. F.(1968), has been made on the basis of this sort analysis as well. Moreover, their result seems to be very accurate. Since our data range is limited in the United States and with better model. The results are expected to be even more concise. In order to find the effect of latency of some factors, which are the number of marriages and the number of births in this case, we did many experiments to compare the results. As an

THE LATENCY EFFECT IN LIFE INSURANCE

15

illustration, if one is calculating insurance sales in 2008 and we assume the lag effect of the number of births is two years and the lag effect of the number of marriages is eight years, the number of birth will respectively take place in 2006. For marriages, it will separately take place in 2002. In our study, we are going to experiment with various combinations of the same independent variables with different years lag for birth and marriage to find out whether these two factors have hysteresis effects on the sales of life insurance or not and how many years these variables will take to affect the sales of life insurance in the United States. The results should be very helpful for these insurance companies to amend their data for the demand for life insurance and they will get more accurate and concise results in the end. Here are the twelve models that are elaborately selected from these combinations. The first model, we assume there are no lag effect of both number of births and number of marriages. It indicated that peoples life insurance purchasing decision will be affected just after he or she got married or got his or her child. In reality, this is a very strong assumption but the reason why we still keep it as one of the models is to compare and contrast with others and demonstrate the latency effect of some factors exists. From model two to model six, we use four models to compare with the first one. In this place, we assume that number of marriages does not have lag effect on the sales of life insurance in the model two, three and four. For the model five and six, we assume that number of birth does not have lag effect on the sales of life insurance. Therefore, we can directly see how these lag effects influence our independent variable and our final result. Here are the rest models, the second model, we choose two years lag effect for number of births and no lag effect for number of marriages. While the third model, we decided is three years lag effect for number of births and

THE LATENCY EFFECT IN LIFE INSURANCE

16

no lag effect for number of marriages. The fourth one has four years lag effect for number of births and no lag effect for number of marriages. For the fifth model, on the other hand, we use no lag effect for number of births and six years of lag effect for number of marriages. The six models assume no lag effect for number of births and eight years lag effect for number of marriages. From comparing these five models, we should have a much more clear view towards the latency effect and figure out how these variables influence our result. Another part is from model seven to model twelve. In this part, we selected combinations of different lag effect for both number of births and number of marriages. To figure out how many years of lag effect generate the best result is the major goal of this part. Here are the models. For the seventh one, we assume two year lag effect for number of births and two years of lag effect for number of marriages. We use three year lag effect for number of births and two years lag effect for number of marriages in the eighth model. The ninth model, we choose to use two years lag effect for number of births and three years lag effect for number of marriages. The tenth model utilizes both three years lag effect for number of births and for number of marriages. The last model, we assume four years lag effect for number of births and three years of lag effect for number of marriages. All of these models are selected from various combinations and proved to be better models than others. By comparing these models, we will find out the real latency effect for both number of births and number of marriages.

THE LATENCY EFFECT IN LIFE INSURANCE

17

V.

Results The data in this study covers from 1979 to 2007 of the whole country and six variables in

total. Using the data, we get our regression results in table 3 and table 4. In the table, we got coefficient and P-value for each variable in all models. We can also got adjusted R-square to detect the reasonability of these models and compare these models. In both table 3 and table 4, we got all the results as we expected. This indicated that the explanations in section III are more likely to be correct. We can combine our previous explanations and the regression results to interpret the meaning of our results. The P-values of average price for all the models are lower than ten percent. It means this variable does have impact onto our dependent variable that is the sales of life insurance. The sales of life insurance will goes up as the average price of life insurance goes up. The coefficient of average price of life insurance is very big and much more than other variables, which indicated this variable has the dominant effect on the sales of life insurance. Another reason the orders of magnitude of other variables are huge and even though the coefficients of these variables are small, the effects of these variables are actually tremendous. Thus, other variables are still indispensible in this model. For residence population in both Table 3 and Table 4, all the P-values of this variable are zero percent. It indicated that this variable has a significant influence on the dependent variable the sales of life insurance. Furthermore, the coefficients for this variable in all the models are positive that means the larger the population is, the more insurance will be sold. Due to if there are more people, there are more potential consumers in the market and thus, insurance companies have more opportunities to sell their products.

THE LATENCY EFFECT IN LIFE INSURANCE

18

In both Table 3 and Table 4, all the P-values of college education are less than ten percent, which means this variable also has a significant influence on the dependent variable the sales of life insurance. Moreover, all the coefficients of this variable are negative. It is consist with the explanation in section III selection of variables. Less educated people usually have more dangerous jobs than people whom are well educated. In the next part, we are going to use control variable method to test the latency effect of two independent variables number of births and number of marriages. In model two, model three and model four, we assume number of marriages does not have lag effect and compare the effects on our results, which are caused by the latency effect of number of births and in model five and six, we assume number of births does not have lag effect and compare the effects on our results, which are caused by the latency effect of number of marriages. From the results shown in table 3, we can see the P-value for number of births in model one is 0.7266, which is much bigger than ten percent. This seems to be unrelated to the sales of life insurance. The explanation for this part is that number of births does have lag effect on the sales of life insurance. Therefore, we use the current years value of number of births and number of marriages directly to run the regression lead to incorrect results. Compared to model one, the number of births in model two to model four has different level of lag effect. Judged from their P-value, their P-values are lower than 0.9178 and all under ten percent. Thus, they are indeed much more accurate than the first one. What is more, we can get the same result from the coefficient of number of births as well. In model one, the coefficient of number of births is just 0.002948. However, in model two the coefficient of number of births is 0.011426. In model three, the coefficient of number of births is 0.013834. The coefficient of number of births is 0.012415 in model four. They are all much bigger than 0.002948, which means number of birth influence more on the results in model two,

THE LATENCY EFFECT IN LIFE INSURANCE

19

Table 3. Model 1(0,0) Model 2(2,0) Model 3(3,0) Model 4(4,0) Model 5(0,6) Model 6(0,8)

Variables

Average Price Coefficient P-value 4716.172 0.0996 5510.613 0.0228 6406.14 0.0066 8368.068 0.0014 6408.577 0.0194 9887.615 0.0082

Birth Coefficient P-value 0.002948 0.7266 0.011426 0.0499 0.013834 0.013 0.012415 0.0337 0.008148 0.2691 0.012771 0.1166

Marriage coefficient P-value 0.001074 0.9559 -0.005107 0.734 -0.006257 0.6658 -0.005311 0.7196 0.022404 0.1931 0.018951 0.3446

Residence Population coefficient P-value 0.001642 0 0.001524 0 0.001436 0 0.001358 0 0.001544 0 0.001666 0

College education coefficient P-value Adjusted squared R0.9903032 0.992085 0.9922 0.991246 0.991167 0.989795 -0.005306 0.0089 -0.004765 0.0032 -0.004204 0.006 -0.004135 0.0113 -0.006834 0.0127 -0.010095 0.0068

THE LATENCY EFFECT IN LIFE INSURANCE

20

model three and model four. Judging from Adjusted R-squared, model two, three and four are also better than model one. Thus, we can tell the latency effect does exist for number of births. For the number of marriages, we find the similar phenomenon. From model two to model four, we get negative coefficient of number of marriages, which should be positive according to our explanations. Since we assume number of marriages does not have lag effect in this three models, the coefficient may caused by the incorrect assumption. Therefore, we create model five and model six. P-value for number of marriages in model one is 0.9559, which is very big. In model five, the P-value for number of marriages reduce to 0.193. The P-value for number of marriages decreases to 0.3446 in model six. On the other hand, the coefficients of number of marriages in model five and six are 0.022404 and 0.018951, which are positive as we expected. Moreover, the coefficients of number of marriages in model five and six are much larger than 0.001074 in model one. Hence, we proved that number of marriages also has latency effect on our dependent variable that is the sales of life insurance. We have already proved that both number of births and number of marriages do have latency effect on our dependent variable - the sales of life insurance. However, how many years it will usually take for birth and marriage? According to the paper demand for life insurance by Headen, R. S., and Lee, J. F.(1968), it will takes about two years to affect their parents life insurance purchasing decision after the birth of their children and after eight years of marriage, couples will usually start to buy life insurance. However, the study was completed more than 40 years ago. The preference towards life insurance and the latency effect of these two variables maybe changed during this period of time.

THE LATENCY EFFECT IN LIFE INSURANCE

21

Table 4.

Variables

Model 7(2,2)

Model 8(3,2)

Model 9(2,3)

Model 10(3,3)

Model 11(4,3)

Average Price coefficient P-value 5575.354 0.021 6439.177 0.0064 4727.587 0.0442 5453.438 0.0115 10122.89 0.0031

Birth coefficient P-value 0.11133 0.0525 0.013556 0.0138 0.012093 0.0308 0.014904 0.0039 0.012997 0.0149

Marriage coefficient P-value 0.003687 0.8371 0.005106 0.7676 0.030928 0.0945 0.035035 0.0424 0.007467 0.6507

Residence Population coefficient P-value 0.001532 0 0.001451 0 0.001606 0 0.00154 0 0.001316 0

College education coefficient P-value -0.004559 0.0056 -0.003977 -0.003872 0.0142 -0.003379 0.015 -0.004474 0.0316

Adjusted R-squared

0.992056

0.99216

0.992309

0.993524

0.99196

THE LATENCY EFFECT IN LIFE INSURANCE

22

In table 4, we set up five models with different combinations of latency effect of both number of births and number of marriages and find out how many years number births and number marriages will take to affect the sales of life insurance. These models are elaborately selected and all generate good results. The results turns out that when the latency effect of birth and marriage is both around two to four years shows the much better result than others. However, from model 7, model 8 and model 11, we can still find at least have one variables P-value is more than 10%. In model 7, the P-value of number of marriages is 0.8371. P-value of number of marriage is 0.7676 in model 8 and it is 0.6507. In these three models, we should reject the effect of number of marriages, which means the year assumptions for latency effect of number of birth and number of marriages are incorrect. On the other hand, according to Adjusted R-squared, model 9 and model 10 have highest number, which are 0.992309 and 0.993524. What is more, the P-values of all the variables in both model 9 and model 10 are all under 5%. This shows model 9 and model 10 that are the best models among these five selected models. The lag effect of number of births is two years and the lag effect of number of marriages is three years in the model 9. The lag effect of number of births is three years and the lag effect of number of marriages is three years in the model 10. Therefore, results from table 4 show the latency effect of birth is approximately two or three years and the latency effect of marriage is about three years.

THE LATENCY EFFECT IN LIFE INSURANCE

23

VI.

Conclusions

In this paper, we consider the latency effect of two variables that are number of births and number of marriages by using an Ordinary Least Square (OLS) model. By comparing models with different combinations of latency effect of one variable with the other stay no lag effect to model one, which is our reference model, we find these two variables number of births and number marriages do have latency effect to some extent on our independent variable the sales of life insurance. In another part of this paper, several models were built to figure out how many years the two variables will take to influence the sales of life insurance. We find out the lag effect of number of births is two or three years and the lag effect of number of marriages is three years generate the best results in adjusted R-square and in these two models, the P-value of all the variables are lower than ten percent. However, according to the paper demand for life insurance by Headen, R. S., and Lee, J. F.(1968), it will take about two years to affect their parents life insurance purchasing decision after the birth of their children and after eight years of marriage, couples will usually start to buy life insurance. But, the research was completed more than 40 years ago. The preference towards life insurance and the latency effect of these two variables might have changed during this period of time. By using the data from 1979 to 2007, the lag effect of number of births are approximately two or three years and the lag effect of number of marriages are about three years. The results described above do seem to offer a fairly accurate latency effect of number of births and number of marriages. With the concise lag effect of these two variables, we can get more exact and practice demand function for the sales of life insurance. Thus, this study will enable life insurers to have a better understanding of customers life insurance behaviors and

THE LATENCY EFFECT IN LIFE INSURANCE

24

have a new demand function for life insurance, which would be more concise and accurate. As a result, life insurance corporations will be more precisely to anticipate the demand of life insurance in the future.

THE LATENCY EFFECT IN LIFE INSURANCE

25

References Anderson, D. R, & Nevin, J. R. (1975). Determinants of young marrieds life insurance purchasing behavior. An empirical investigation. Journal of Risk and Insurance, 42, 375387. Devaney, S. S., & Keaton, E. F. (1986). Determing purchasers of whole life insurance: Using a classification tree. Journal of the Society of Insurance Research, 7(2), 33-45. Headen, R. S., and Lee, J. F. Life Insurance Demand and Household Portfolio Behaviour. American Risk and Insurance Association: 685-98. Web. 9 Jan. 2010. Life Insurance Marketing and Research Association. (2006, November). Facts about life 2006. Retrieved from http://www.rosemaryinsurance.com/factsaboutlifeins.pdf Life Insurance Marketing and Research Association. (2007, November). Retrieved from http://www.limra.com/PDFs/NewsCenter/Materials/07USFAQ.PDF Mantis, G, and farmer, R. N. (1968). Demand for life insurance. American Risk and Insurance Association, 35, 247-56. Li, M. (2008). Factors Influencing Households Demand for Life Insurance. Retrieved from http://www.scribd.com/doc/21342015/Factors-Influencing-Households%E2%80%99Demand-For-Life-Insurance-%E2%80%93-Min-Li-2008#fullscreen:on Truett, B., and Truett, J. (2010). The Demand for Life Insurance in Mexico and the United States: A Comparative Study. American Risk and Insurance Association, 57, 321-28. Browne, M. J., Kim, K. (1993). An International Analysis of Life Insurance Demand. The Journal of Risk and Insurance, 60, 616-634.

THE LATENCY EFFECT IN LIFE INSURANCE

26

Headen, R. S., and Lee, J. F. (1974). Life Insurance Demand and Household Portfolio Behavior. American Risk and Insurance Association, 41, 685-98.