Last updated: 19 September 2012
Nova School of Business and Economics
Campus de Campolide
Abstract not available.
In this presentation, we aim to investigate the impact of drug decriminalization in Portugal using the synthetic control method. The applied econometric methodology compares Portuguese drug-related variables with the ones extracted from a convex combination of similar European countries. The results suggest that the policy change contributed to a decrease in the number of heroine and cocaine seizures, a decrease in the number of offenses and drug-related deaths, and a decrease in the number of clients entering treatment.
In the last 10 years, the Portuguese governments have created more than 50 policy measures aimed at developing the pharmaceutical market for generic drugs, which was nearly nonexistent. In this presentation, we propose to study the impact of all the measures adopted, which has not been done until now. Using information for generic drug sales and market share (in value and volume), we estimate a diffusion curve using a parametric approach for the period from 2000 to 2010. This curve is estimated using logistic regressions in which policy measures are included in the diffusion rate using dummy variables, one for each measure.
Our results show that the development of the generics market has been following its own dynamic. Indeed, only a small part of the measures had an impact in the market. Moreover, this impact was not always positive. In addition, statistically significant effects are more easily found when the market is considered in prices rather than in volume, as expected, because several policies address pharmaceutical drug prices. The market in terms of value was found to be already reaching its limit, while in terms of volume it is still in expansion. All in all, there is no single policy measure that has an unquestionable impact on the generics market. Nevertheless, the measures taken by governments in the last 10 years do not seem to be fulfilling their objectives.
We study the determinants of MRI use across Portuguese NHS hospitals for patients belonging to specific diagnosis-related groups (DRGs).
Using data on individual hospital admissions, we estimate a probit model including individual-, hospital-, time- and region-specific variables to explain the probability of a patient being sent for an MRI.
Results convey a tightening effect on the hospital’s budget constraint in the end of each year. Hospitals seem to account for regional characteristics when defining adoption patterns. Individual-specific variables are good predictors of MRI use. Measures taken by the government only impact the short run. Finally, the gains from an MRI scan, as far as the probability of death is concerned, occur mainly for less severe patients.
In this presentation, I aim to assess whether hospitals react to random demand pressure by discharging patients earlier than expected. As a matter of fact, combining an unpredictable demand for medical services with limited and, to some extent, fixed medical resources generates strong incentives to discharge patients earlier than expected when demand is high—increasing the risk of readmission and decreasing the benefit from treatment. This work was conducted as a way to determine whether those incentives actually affect discharging decisions. Analysis of Portuguese hospitals data shows that hospital utilization levels at the time of admission, prior to admission, and post admission do have a negative impact over the length of stay in hospital, although this impact is quantitatively irrelevant. More than that, larger utilization levels have a positive impact over the probability of being discharged at certain days of the week, indicating that a problem may exist with early discharges.
Technological innovation in health care has become the main catalyst of development in the health sector. It has allowed the modernization of health care, increasing its quality, safety, timeliness, and efficiency. Notwithstanding these positive benefits, there is not yet an international consensus concerning the potential benefits and cost savings that can be achieved with this innovation.
Accordingly, I aim primarily to understand the actual impact of technological innovations and other determinants on health care costs, expanding several studies by using a panel-data framework and by direct observation through the creation of a health care technological index. Unfortunately, this methodology raises several problems with missing data, because we deal with health data (mainly expenditures and resources) from the 1970s.
To overcome these problems and the resulting estimate bias (Rubin 1987), I use a multiple-imputation technique. The recent Stata multiple-imputation package (available from Stata 11) allowed the unprecedented conclusion that innovation increases health expenditures but at decreasing rates. More interesting, it is argued that most developed countries nowadays are beyond the “turning point”, where the weight of innovation on health care costs becomes increasingly lower.
Rubin, D. B. 1987. Multiple Imputation for Nonresponse in Surveys. New York: Wiley.
It seems that the news of the death of jobs for life is premature. In this study, we use quantile regression methods to analyze the changes in job duration distribution in Portugal by using matched employer–employee data. We use a decomposition method proposed by Machado and Mata (2005) to disentangle the contribution of the compositional changes and the structural changes. Our findings indicate that there is a decrease in job durations. Both compositional changes and structural changes play a role, albeit in opposite directions. We find that the decrease in job duration is an illusion brought about by the bigger share of the external services industry and by the changing relationship between firm size and job duration.
Machado, J. A. F., and Mata, J. 2005. Counterfactual decomposition of changes in wage distributions using quantile regression. Journal of Applied Econometrics 20: 445–465.
It is possible to calculate the optimal performance time (the ideal length of the hazard cycle) for all events whose hazard function is time dependent and where there is a preventive action that restores the initial hazard level. This time length depends on the difference between the preventive action and the event occurrence costs and on the hazard function.
The classical single-variable hazard functions can then be generalized to a multivariate probability model. By using single variables, an event gets one optimal hazard cycle length. By adding fixed covariates, we can measure the impact of each covariate on the event optimal cycle length and cost values.
Using time-dependent covariates, the optimal cycle length and cost are always changing. In this way, we obtained a set of results, using the Cox models, demonstrating this decision response to the proportional hazard different variations over time. The time can be replayed for any other counter.
This methodology is applied to maintenance management. The application to the medicine is direct because there is a perfect analogy in the decisions. Another main goal is to discover other areas of application.
In this presentation, I show a new Stata command, scoredum, that implements a score test for nonlinear models that can be used to test for fixed effects in Poisson and logit regression models. The command can also be used to implement Pearson chi-squared tests on binary or count data at the individual level. The command is partially coded in Mata.
A staple for academic researchers, Stata is known for its comprehensive range of facilities and simple, powerful language. It is also highly regarded amongst developers. Because of its open architecture, it is remarkably flexible to extend to new applications and to implement new statistical procedures. Given these advantages and Stata's affordable pricing, it seems surprising that Stata is arguably less accepted outside of academia. One potential reason might be its reputation of being overly command-line driven and less user-friendly than other programs.
By adding support for the widely used Microsoft Excel file format in release 12, however, the gap between academia and practice has considerably narrowed. Therefore, in this talk, I aim to facilitate the use of Stata by practitioners by demonstrating how users now can easily generate automated reports in Excel.
Visualizing the true effect of a predictor over a range of values can be difficult for models that are not parameterized in their natural metric, such as for logistic or (even more so) probit models. Interaction terms in such models cause even more fogginess. In this talk, I show how both the margins and the marginsplot commands can make for much clearer explanations of effects for both nonstatisticians and statisticians alike.
Pedro Pita Barros, Universidade Nova de Lisboa
João Cerejeira, Universidade do Minho
Sónia Félix, Universidade Nova de Lisboa
Miguel Portela, Universidade do Minho
Luís Catela Nunes, Universidade Nova de Lisboa
Antonio Gouveia de Oliveira, Universidade Nova de Lisboa
Patricia Xufre, Universidade Nova de Lisboa
Timberlake Consultores, the official distributor of Stata in Portugal.