1、吸烟论文The Empirical Analysis of the Factors Affecting Adult SmokingAbstract:By using the 2006 CHNS cross-sectional data, the article tries to figure out the main factors influencing smoking behavior. The method includes two IV, one for income (IV is number of staff in jobs organization) and the other
2、for health (IV is whether getting medical insurance), a proxy variable, dummy interaction term, LPM and logit model. The conclusion is age and health both have quadratic relationship with smoke, and the most significant variables are male, alcohol and education. Student: 马超凡 ID number:41110042.Intro
3、ductionNowadays, the number of smokers in china has been over 300 million. And there are total about 540 million smokers including passive smokers. The close relationship between smoking and health contribute smoking to be the fourth most dangerous factor that influences health. Every year, there we
4、re about 1 million people died from tobacco related diseases. There is no doubt that smoking really brings grim challenge to human society.Smoking, as an addictive behavior, has complex contributing factors. Some of them are evident like male, education, alcohol, health condition and age. Some of th
5、em just exert subtle influence on smoking behavior like married condition, number of siblings and sleep. Now we are in a period of social transition. In the fast pace of life, these factors are working together to produce more smokers. To solve this serious problem, numerous relevant researches have
6、 been conducted. Most researches are about the smoking behaviors of teenagers and college students, for they are the most potential smokers. Ma Xiaobin, a professional doctor, find that peer smoking behaviors are contagious. Once a student forms this bad habit, he will attract and affect his friends
7、 to imitate him and try to smoke. And subsequently this bad behavior can be spread widely. Teenagers often have no awareness of serious harm of smoking. They once start smoking, then they will indulge in it for the moments pleasure. On the other hand, many students just regard smoking as showing uni
8、que personality, which encourages the bad atmosphere.Xie Jia, a postgraduate student of Zhejiang University, mainly focuses on the effects of mental pressure on smoking. Stress measurement results show that the majority population is facing greater stress, especially the people with low income, low
9、educations. Therefore, effectively solve the problem of excessive stress will be a powerful factor to reduce urban residents smoking rate. It is suggested that the government should focus on this group of people, not only concerned about their smoking behavior, but also to care about living and work
10、ing pressure. After reading so many papers about smoking, I find most of them are about teenagers. Even though there are a few papers about adults, these papers are almost about the survey results in macroeconomics. They just fail to dig out the deep reasons behind the smoking behaviors. In my paper
11、, the research focuses on adult. First figure out the comprehensive contributing factors of smoking, and then study how they lead to smoking. .Data description:Variable description:1)Smoke (dependent variable): the number of cigarettes smoked every day2)Ifsmoke (dependent variable): a dummy variable
12、. “1” means the person has smoked.”0” means the person never smoked.3)Education: the number of years of education the person has received. I hold the belief that people who received more education tend to smoke less.4)Male: a dummy variable with “1” representing male. In general men are more like to
13、 start smoke and smoke more cigarettes.5)City: a dummy variable. “1” means the person lives in city. “0” means he person lives in country. Its hard to say which group smokes more.6)Age: the years old the person7)Sports: a dummy variable. Define people who dont participate or dislike walking, tai chi
14、, all kinds of ball games and fitness activities as “0”. The other is defined as “1”.8)Income: it is the quantity of total income in one year. It is calculated by the following equation: (first professional monthly wage+ monthly subsidy)*12 + first professional yea-end bonus + (second professional m
15、onthly wage + monthly subsidy)*12 + second professional yea-end bonus + farming income + Animal husbandry income + fishing income + business income.9)Starttime: if the person has smoked, this is the starting age of smoking. Maybe starttime is positively correlated with smoke 10)Alcohol: this is a du
16、mmy variable. “1” means the person who drinks alcohol. “0” means the person doesnt drink alcohol. People who drink alcohol are more likely smoke.11)Sleep: the hours the person spends on sleeping per day.12)Health: this is the result of self-evaluation. “1”means very good,”2” means good,”3” means ord
17、inary, and “4” means bad.13)Married: a dummy variable.”1” means the person has married and is still has a marriage relationship.”0”means the person who is unmarried or divorced or his spouse has been dead.14)Siblings:the number of siblings the person has.15)Midnsur: a dummy variable, ”1” means the p
18、erson has a medical insurance.16)Workers: the number of staff the persons job firm or organization have.Possible omitted variables:1)The cigarettes brand and price each smoker buys. It is related with income and city and maybe education. So this omitted variable is bound to produce bias and inconsis
19、tency. We assume that people who buy low price cigarettes can smoker more every day, so cigarette price is positively related with smoking quantity. This omitted variable has a positive coefficient. Then we know people who have more income, is in city, and have higher education tend to buy higher pr
20、ice cigarettes. So income, city and education all have positive correlations with cigarette price. So the biases (coefficient of price*correlation) of coefficients of these three variables (income, city and education) are all positive. 2)The working pressure. Because it is unobservable, I find a pro
21、xy variable.Whours: the usual working hours in one week. In my view, people with more working pressure tend to smoke more. Although his proxy variable can generate bias on three parts: intercept, working pressure and error term, it insures that the all other important variables are unbiased. Without
22、 this proxy variable, all explanatory variables will be biased because of omitted variable.Sample selection:1) I have dropped the observations whose any variable is negative. 2) Drop all the observations whose values are beyond the range set for each variable.2) Then I deliberately have dropped the
23、observations whose income100. For one thing, it is impossible for a person has income of a ye1ar less than 100, so these observations are outliers. For another, although this can cause nonrandom samples, it has no effect on unbiased and consistent estimators in the population model, for the regressi
24、on function E (smoke | income, age, male and so on) is the same for any subset of the population, which is exogenous sample selection.After doing these two steps, the sample size is reduced from 9528 to 3179 observations.Model specification:The original model is Smoke=0+1education+2male+3age+4income
25、 +5alcohol+6married+7siblings+8sleep+9city+10sports+11health+uNext, it is needed to be modified.1)Firstly, now that functional misspecification can also be due to omitted variable, in the equation, whours (working hours per week) as a proxy variable of working pressure should be added.2)Secondly, ch
26、ecking whether quadratic or cubic items are needed. I guess age has a quadratic relationship with smoke, so I add age2(marked as age2) variable. This graph describes cigarettes smoked per day on average for different periods of age. The data used are all 3179 observations including 2268 nonsmokers w
27、hose cigarettes smoked per day are 0.The graph shows that smoke is increasing as age increases until about 50-55 years-old. Then smoke decreases as age increases.Using RESET test, I estimate two models for smoke. The first one has all variables in level form. The second one adds age2. The RESET stat
28、istic for equation (1) turns out to be 26.7, this is the value of an F (2,3130), and the associated p-value is 0.0000,this is the evidence of functional misspecification in (1). The RESET statistic for (2) is 24.6, with p-value=0.0000. So we can see the equation(2) is better, but the functional miss
29、pecification is still a problem.Then use the 3 significant variables square (health, education, siblings) respectively added to the equation to see whether the function is improved. Then health square gives a significant t=-2.96. So remaining the health square into the equation is necessary.3) Third
30、ly, check whether log items is needed with. Income should be used as ln(income), because income ranges from 100 to 960,720 yuan. So we should try to shorten the range. After doing this, income ranges from 4.60517 to 13.77544. And the Mizon&Richasrd test also shows lnincome is better. I try to use ln
31、(whours) to replace whours, but 756 observations have 0 working hours, so lnwhours is not available.4) Fourthly, use chow-test to check whether interaction terms are needed. At first, I use married*other 12 variables respectively to add to the function. Then F(12,3116)=1.46 with p-value 0.13, so we
32、cant reject the null hypothesis at 5% significance level and even at 10% significance level. So the interaction terms are jointly insignificant.Then I use male*other 12 variables respectively to add to the function. The F=7.2 with p-value 0.0000. So there is no doubt that some male interaction terms must be added. Then I choose four very significant interaction terms male*education, male*age, male*age2 and male*alcohol, with t-statistics are over 5 to add to the equation. Unfortunately, originally significant variables like education, male, age,
copyright@ 2008-2022 冰豆网网站版权所有
经营许可证编号:鄂ICP备2022015515号-1