Organizational Change and Workers’ Health

Get Complete Project Material File(s) Now! »

A eld experiment on the demand for prevention among elderly


Which policy is the most e ective to increase demand for prevention among the elderly ? As a result of advances in medicine and major improvements in living conditions, life expectancy has increased during the last decades (Cervellati and Sunde, 2005). Together with a decline in the fertility rate, this demographic evolu-tion creates an important change in the age structure. It has led to an increase in the relative size of the dependent population and consequently put great pressure on public nances by increasing the amount devoted to social expenditures (Hage-mann and Nicoletti, 1989). In France, public spending allocated to the dependent elderly accounted for 1,3% of the total GDP in 2010 2. The Retirement Insurance agency has developed a new preventive program. Part of this program aims at adap-ting housing in order to maintain the self-su cient elderly at home. Despite some evidence of the program’s e ciency, the demand for the program remains very low among the elderly. In 2011, only 0,2% of the retired people living in Ile-de-France, the capital region, bene ted from preventive measures aimed at adapting housing. Acquiring a better understanding of demand-side barriers that hinder the uptake of prevention among elderly and nding an e cient way of increasing demand are today a main concern for policymakers.
In developed countries, States have a real nancial interest in reducing the bur-den of societal support for the elderly (Heller, 1989). In this context, maintaining self-su cient elderly at home would appear to be a way to limit costs to the pu-blic purse. Most social actors agree that priority should be given to prevention (Harwood, Sayer, and Hirschfeld, 2004). Several pieces of evidence suggest that the adoption of preventive measures could reduce dependence among the elderly by hel-ping people cope with impairments (Cutler, 2001). Falls are a major cause of injury and loss of self-su ciency among the elderly and account for 70 percent of acci-dental deaths in persons over 75 (Assessment, 2000). Tinetti et al. (1994) emphize that around 30% of people over 65 experience a fall each year with psychological  trauma (Close, Ellis, Hooper, Glucksman, Jackson, and Swift, 1999), physical in-jury (Nevitt, Cummings, Kidd, and Black, 1989) or functional deterioration (Dunn, Rudberg, Furner, and Cassel, 1992) as major consequences. This high probability of falling lead to increased levels of dependency, and higher costs associated with falls (Beard et al., 2006) whereas several randomized control trials on the impact of prevention program nd positive e ects 3.
The low level of demand for prevention programs might be explained in a rst instance by a lack of information. Providing the elderly with information about the program could be both a way to overcome failures in knowledge (Hsieh and Lin, 1997), and moreover, a way to improve risk perception. Indeed, a large number of studies nd that beliefs about risks are often biased, although the signs of the bias may not be obvious. For instance, in a study about smoking and the risk of lung cancer, Viscusi (1990) nds that people tend to underestimate the likelihood of events with strong probability to occur and to overestimate the likelihood of events with low probability to occur. Viscusi and Hakes (2008) con rm these re-sults and generalize them to the perception of mortality risks and life expectancy loss. Providing information might be a way to achieve a better risk perception and consequently a more optimal level of prevention e ort. Avery, Kenkel, Lillard, and Mathios (2006) nd a causal e ect of a Canadian information campaign on tabacco consumption among smokers. Moreover, compliance with prevention may be highly sensitive to the way the information is provided. For example, in their eld expe-riment on food choices, Downs, Loewenstein, and Wisdom (2009) point out that manipulating information is more e cient than providing objective information on calorie content.
However, some empirical studies are more skeptical about the use of information to optimize prevention e orts. For example, Khwaja, Silverman, Sloan, and Wang (2009) show that risky behaviors like smoking can not be explained by a sweked Because individuals have limited cognitive skills, providing too much information can have a negative e ect on behaviors (Simon, 1955; Norman and Bobrow, 1975; Marois and Ivano , 2005). Empirical studies also detect an aversion to information, caused by a fear of facing the fact of disease, which undermines its deterrence e ect on preventive behaviors. For instance, Lerman et al. (1996) show that 40% of patients having a high risk of getting breast or ovarian cancer refuse free testing. Individuals tend to delay a consultation or postpone a medical examination in case of breast cancer symptoms (Meechan, Collins, and Petrie, 2002) or melanoma (Richard et al, 2000). Similar results were found in developing countries for HIV. While making people informed about their HIV status is a necessity to reduce the prevalence of HIV, a large percentage of people do not return to pick up their test results (Thornton, 2008).
Price is another major potential determinant of demand for health prevention services. Its impact on take-up is one of the most contentious policy issues and has often been discussed in the economic literature. Many randomized evaluations were conducted in developing countries to determine how price a ects purchase decisions for health services. Most of them nd e ects in line with the standard economic models on human capital investment (Holla and Kremer, 2009). In the theoretical framework developed by Grossman (1972b), prevention, for instance having one’s dweling adapted, can be considered as an investment realized to overcome health capital depreciation. The choice of prevention results from the equalization of the marginal bene ts of health capital with the price of the investment. Results, mostly in developed countries, tend to nd a very low price sensitivity of demand for medical care (Fuchs, 1972) as well as for insurance (Manning, Newhouse, Duan, Keeler, and Leibowitz, 1987). Conclusions are di erent for prevention behaviors. In a randomized experiment conducted in Kenya, Cohen and Dupas (2010) show that demand for insecticide-treated bedding, a preventive measure against malaria, is very sensitive to price and drops signi cantly after a price increase. Kremer and Miguel (2007) nd similar results about drugs o ering prevention against intestinal worms : the introduction of even a small cost-sharing component reduced take-up by 80% despite mobilization interventions. Focusing on rubber shoes for children, Meredith, Robinson, Walker, and Wydick (2011) nd that 78% of the decision to purchase is explained by variations in price. Moreover, several studies show that incentives may be a way to increase the demand for health products and point out non-linearities in its e ect (Holla and Kremer, 2009). Small subsidies may make a big di erence in take-up. For example in the randomized experiment in Malawi, small incentives strongly increased the probability of picking up test results (Thornton, 2008). Byrne and Thompson (2001) give a theoretical dimension to this result, showing that the level of prevention e ort is suboptimal and that a simple subsidy could be e cient to attain the rst best level of prevention. The non-linearity of this e ect could also be explained by time-inconsistent preferences and procrastination behaviors (O’Donoghue and Rabin, 1999). Some people, knowing they are present-biaised, tend to have a preference for commitment (Du o, Kremer, and Robinson, 2009) and a demand for control devices (Kan, 2007). Imposing deadlines to bene t from a small subsidy can reduce procrastination behaviors and increase take-up.
In this paper, we contribute to the existing debate on factors in uencing demand for prevention by studying the take-up of a housing adaptation program that o ers assistance for home adaptations. We use a unique opportunity to test the impact of several manipulations on take-up for this program. We conducted a large scale eld experiment and sent over 40,000 yers to French retired people urging them to call the Retirement Insurance agency in order to bene t from the housing adap-tation program. We implement two main categories of treatment. First, following the methodology applied by Cohen and Dupas (2010), we randomly increase the amount of the subsidy in the program and also test the impact of additional re-ductions conditional on commitment. Secondly, we examine the e ect of providing information about risk. Seven yers were designed to contain randomized framing manipulations. We test, for example, the existence of loss aversion and belief revi-sion mechanisms.
We are therefore able to estimate the impact of the di erent treatments on intention to demand and risk perception as well as on the e ective demand. The main issue is to determine whether the determinants are the same as those in uencing the e ective demand.
We provide evidence that giving more information on risk is the most e ective treatment to increase the take-up of the program. In line with Bertrand, Karlan, Mullainathan, Sha r, and Zinman (2010), our results also suggest that a crucial role is played by the framing of the information. A message about risk framed as a personalized letter is the most e ective way to induce take-up, whereas the same message framed as statistical information leads to the lowest take-up, even lower than the control yer. On the other hand, we do not nd strong evidence for price sensitivity. Only a 100% subsidy appears to be e ective at inducing demand, and in any case this e ect is still smaller than the impact of information. This result departs from the ndinggs of other studies. Indeed several randomized experiments conducted in other contexts, mainly in developing countries, show price as the main determinant of investments in prevention (Meredith, Robinson, Walker, and Wy-dick, 2011; Cohen and Dupas, 2010).
Using the survey information, we also show that the determinants of intention to demand and risk perception are the same. But, we provide evidence that they are di erent from the determinants of actual take-up. For example, considering men, we nd that a loss versus gain frame a ects strongly and positively the intention to demand and risk perception, whereas it a ects the e ective demand negatively. More generally, the whole response pro le is not the same for the intention to demand and risk perception as it is for take-up. This last result highlights the fact that there is a gap between intention and action. The implicit general understanding of the decision-making process as a single process with a unique set of determinants is not validated by our experimental data. They suggest, on the contrary, the existence of a dynamic in the decision-making process, with di erent determinants playing a role at di erent steps. One remaining issue is to determine to what extent intentions to demand and risk perception are prior to the true e ective demand that will be realized in the future.
This paper is organized as follows. Section 2 provides background information on the health product we focus on and presents randomizations and the sample. Section 3 presents results on take-up. Section 4 provides additional ndings on determinants of the decision-making process while section 5 concludes.

READ  Node Activity Scheduling and In-Network Data Storage: Impact of Introducing duty-cycling on DA  VBC 71

Experimental design

The program

The health product we focus on in this paper is a new prevention program called Personalized Actions Programs (PAP) implemented by the Retirement Insurance agency since 2004. An evaluation of the needs of the elderly person is performed by a caseworker mandated by the Retirement Insurance agency during a home visit. After observing the living environment and interviewing the elderly person, the caseworker recommends support and services all aimed at reducing the risk of dependency. The Retirement Insurance agency supplies nancial support for a large range of services : home help, tele-assistance, housing adaptation, bringing of meals or prevention workshops on di erent subjects.
In this study, we focus on home adaptation. This program o ers nancial sup-port for minor housing adaptations that help elderly people to perform activities independently and safety at home. It aims at limiting physical environmental bar-riers, facilitating mobility and reducing the risk of falling (Iwaksson and Isacsson, 1996). Grab bars, external handrails, stair-rails, ramps or alterations to steps are examples of the proposed adaptations. We focus on home adaptation for several reasons. First, the presence of home hazards is important in predicting falls at home, specially among the more vigorous retired people (Northridge, Nevitt, Kel-sey, and Link, 1995). Despite the e ctiveness of home adaptation measures (Hey-wood, 2001), it is still one of the less-demanded services among those o ered by the Personalized Actions Programs (PAP). 350,000 retired people bene t from a PAP each year but only 15,000 among them accept having their home adapted (around demand for prevention among elderly 4%). Caseworkers face huge di culties in convincing elderly persons to adapt their homes.
To bene t from the program, persons have to contact the Retirement Insurance agency to submit an application. When accepted 4, the demand is transmitted to one of the mandated service providers. Needs are evaluated by a caseworker visiting the seniors’ place of residence. However, the senior herself takes the decision to implement the recommendations and is in charge of the practical aspects. The Retirement Insurance agency o ers a means-tested percentage of total outlays as nancial support.

The randomizations

In partnership with the Retirement Insurance agency, we sent 14 yers randomly in three waves (September, October and November 2011). All individuals in the treatment groups (42,079 retired people) received a yer. Each yer corresponds to a speci c treatment aiming at increasing take-up for the program. The 14 yers provide di erent incentives to contact the Retirement Insurance agency. The phone number of the Retirement Insurance agency was clearly indicated on each yer (see yers in Appendix). Calling this number to ask for the application form is the rst step in the demand process. A description of each yer is provided in Table 2.1.

Table of contents :

1 Introduction générale 
1.1 Développement de l’ore et insusance de la demande
1.2 Comprendre les determinants de la demande
1.2.1 Des biens trop co^uteux ?
1.2.2 Un manque d’information ?
1.2.3 Des individus peu rationnels ?
1.2.4 Quelques elements de reponse
1.3 Evaluer des programmes visant a augmenter la demande
1.3.1 Methode d’evaluation
1.3.2 Le r^ole de l’information
1.3.3 L’approche communautaire
1.3.4 L’espoir d’une amelioration ?
1.4 Mesurer l’eet de changements organisationnels sur la sante
1.4.1 Depreciation du capital sante
1.4.2 Des dicultes methodologiques
1.4.3 Strategie d’identication
1.4.4 Contributions
2 A eld experiment on the demand for prevention among elderly
2.1 Introduction
2.2 Experimental design
2.2.1 The program
2.2.2 The randomizations
2.2.3 Sample selection
2.3 Results
2.3.1 Overview
2.3.2 Acting on demand
2.3.3 Changing the cost
2.3.4 Providing information
2.4 Additional Findings on the Decision-making process
2.4.1 Non-response and selection bias
2.4.2 Summary statistics
2.4.3 Decision-making process
2.5 Conclusion
3 An evaluation of a Community-based Information Campaign on Health Demand in Mali
3.1 Introduction
3.2 Background
3.2.1 The Malian health system
3.2.2 The program
3.3 Data
3.4 Identication Strategy
3.5 Results
3.6 Conclusion
4 Organizational Change and Workers’ Health : Lessons from the 2000 Reform in the French Energy Utilities 
4.1 Introduction
4.2 Institutional background
4.3 Data and descriptive statistics
4.3.1 The GAZEL database
4.3.2 Descriptive Statistics
4.4 Empirical approach
4.4.1 Identication strategy
4.4.2 Econometric model
4.5 Results
4.5.1 Impact on working conditions
4.5.2 Impact on health
4.6 Robustness checks
4.7 Conclusion


Related Posts