|
on Discrete Choice Models |
| By: | Michelle Sovinsky (University of Mannheim); Liana Jacobi (University of Melbourne); Alessandra Allocca (LMU Munich); Tao Sun (University of Melbourne) |
| Abstract: | As illicit substances move into the legal product space, substitution patterns with legal products become more salient. In particular, marijuana legalization may have implications for the use of other legal “sin” goods. We estimate a structural model of multi-product use of illegal and legal substances considering joint use, limited access to illicit products, and persistence in use. We focus on a young person’s choice to consume marijuana, alcohol or cigarettes (and possible combinations), and we find that sin goods are complements. Furthermore, our findings emphasize the necessity of accounting for joint consumption and access to obtain correct price sensitivity estimates. Post-legalization, youth marijuana use would increase from 25% to 37%. However, counterfactual results show that a combination of (reasonable) tax increases on all goods along with enforcement against illegal use can potentially revert use to pre-legalization levels. The earlier the tax increases are implemented the more effective they are at curbing future use. Our results inform the policy debate regarding the impact of marijuana legalization on the long-term use of sin goods. |
| Keywords: | complementarity, marijuana legalization, limited choice sets, data restrictions, discrete choice models; marijuana legalization; limited choice sets; data restrictions; discrete choice models; |
| JEL: | C11 D12 L15 K42 L66 C35 |
| Date: | 2023–12–19 |
| URL: | http://d.repec.org/n?u=RePEc:rco:dpaper:487&r=dcm |
| By: | Sören Blomquist |
| Abstract: | Estimated labor supply functions are important tools when designing an optimal income tax or calculating the effect of tax reforms. It is therefore of large importance to use estimation methods that give reliable results and to know their properties. In this paper Monte Carlo simulations are used to evaluate two different methods to estimate labor supply functions; the discrete choice method and a nonparametric method suggested in Blomquist and Newey (2002). The focus is on the estimators’ ability to predict the hours of work for a given tax system and the change in hours of work when there is a tax reform. The simulations show that the DC method is quite sensitive to misspecifications of the likelihood function and to measurement errors in hours of work. A version of the Blomquist Newey method shows the overall best performance to predict the hours of work. |
| Keywords: | labor supply, tax reform, predictive power, estimation methods, Monte Carlo simulations |
| JEL: | C40 C52 C53 H20 H30 |
| Date: | 2023 |
| URL: | http://d.repec.org/n?u=RePEc:ces:ceswps:_10827&r=dcm |
| By: | Elisabeth Honka; Stephan Seiler; Raluca Ursu |
| Abstract: | Researchers are increasingly able to observe consumers’ behavior prior to a purchase, such as their navigation through a store or website and the products they consider. Such pre-purchase (or search) data can be valuable to researchers in a variety of ways: as an additional source of information to estimate consumer preferences, to understand how firms can influence the search process through marketing mix variables, and to analyze how limited information about products shapes market outcomes. We provide an overview of these three areas with a particular emphasis on online and offline retailing. |
| Keywords: | consumer search, limited information, consideration sets, retailing |
| JEL: | D43 D83 L13 |
| Date: | 2023 |
| URL: | http://d.repec.org/n?u=RePEc:ces:ceswps:_10786&r=dcm |
| By: | Kramer, Berber; Porter, Maria; Wassie Bizuayehu, Solomon |
| Abstract: | Index insurance is considered an important strategy to reduce agricultural risk and increase smallholder farmers’ investments. However, insured farmers may develop mistrust of insurance if they experience crop losses and do not receive a payout, for instance because index insurance covers only a subset of covariate risks. At the same time, insurance for idiosyncratic risks would introduce differences in payouts within social networks, which might be considered unfair, introduce jealousy, and further depress demand for insurance. We conduct lab-in-the-field experiments with farmers in Ethiopia to examine the effects of a novel insurance approach that ensures insurance payouts for farmers with crop losses due to idiosyncratic events. We also examine the effects of informing farmers about their neighbors’ experiences alongside their own. We find that such social comparison increases perceived fairness of weather index insurance. In addition, providing complete insurance coverage for crop losses increases farmers’ perceived fairness of outcomes and willingness to pay, without introducing jealousy over neighbors receiving different payouts. Finally, we find that the increase in willingness to pay for complete insurance is concentrated among men and risk averse respondents. |
| Keywords: | Agricultural Finance, Risk and Uncertainty |
| Date: | 2023–12–18 |
| URL: | http://d.repec.org/n?u=RePEc:ags:assa24:339075&r=dcm |
| By: | von der Heyde, Leah (LMU Munich); Haensch, Anna-Carolina; Wenz, Alexander (University of Mannheim) |
| Abstract: | The recent development of large language models (LLMs) has spurred discussions about whether LLM-generated “synthetic samples” could complement or replace traditional surveys, considering their training data potentially reflects attitudes and behaviors prevalent in the population. A number of mostly US-based studies have prompted LLMs to mimic survey respondents, finding that the responses closely match the survey data. However, several contextual factors related to the relationship between the respective target population and LLM training data might affect the generalizability of such findings. In this study, we investigate the extent to which LLMs can estimate public opinion in Germany, using the example of vote choice as outcome of interest. To generate a synthetic sample of eligible voters in Germany, we create personas matching the individual characteristics of the 2017 German Longitudinal Election Study respondents. Prompting GPT-3 with each persona, we ask the LLM to predict each respondents’ vote choice in the 2017 German federal elections and compare these predictions to the survey-based estimates on the aggregate and subgroup levels. We find that GPT-3 does not predict citizens’ vote choice accurately, exhibiting a bias towards the Green and Left parties, and making better predictions for more “typical” voter subgroups. While the language model is able to capture broad-brush tendencies tied to partisanship, it tends to miss out on the multifaceted factors that sway individual voter choices. Furthermore, our results suggest that GPT-3 might not be reliable for estimating nuanced, subgroup-specific political attitudes. By examining the prediction of voting behavior using LLMs in a new context, our study contributes to the growing body of research about the conditions under which LLMs can be leveraged for studying public opinion. The findings point to disparities in opinion representation in LLMs and underscore the limitation of applying them for public opinion estimation without accounting for the biases in their training data. |
| Date: | 2023–12–15 |
| URL: | http://d.repec.org/n?u=RePEc:osf:socarx:8je9g&r=dcm |
| By: | Maria Elena Bontempi; Jan Ditzen |
| Abstract: | The generalized method of moments (GMM) estimator applied to equations in levels, GMM-lev, has the advantage of being able to estimate the effect of measurable time-invariant covariates using all available information. This is not possible with GMM-dif, applied to equations in each period transformed into first differences, while GMM-sys uses little information, as it adds the equation in levels for only one period. The GMM-lev, by implying a two-component error term containing the individual heterogeneity and the shock, exposes the explanatory variables to possible double endogeneity. For example, the estimation of true persistence could suffer from bias if instruments were correlated with the unit-specific error component. We propose to exploit the \citet{Mundlak1978}'s approach together with GMM-lev estimation to capture initial conditions and improve inference. Monte Carlo simulations for different panel types and under different double endogeneity assumptions show the advantage of our approach. |
| Date: | 2023–12 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2312.00399&r=dcm |
| By: | Philipp Lergetporer (TU München); Katharina Wedel (ifo Institut); Katharina Werner (ifo Institut) |
| Abstract: | We study how beliefs about the automatability of workers' occupation affect labor-market expectations and willingness to participate in further training. In our representative online survey, respondents on average underestimate the automation risk of their occupation, especially those in high-automatability occupations. Randomized information about their occupations’ automatability increases respondents’ concerns about their professional future, and expectations about future changes in their work environment. The information also increases willingness to participate in further training, especially among respondents in highly automatable occupation (+five percentage points). This uptick substantially narrows the gap in willingness to train between those in high- and low-automatability occupations. |
| Keywords: | automation; further training; labor-market expectations; survey experiment; information; |
| JEL: | J24 O33 I29 D83 |
| Date: | 2023–12–20 |
| URL: | http://d.repec.org/n?u=RePEc:rco:dpaper:488&r=dcm |
| By: | Luca Gaegauf (University of Zurich); Simon Scheidegger (University of Lausanne); Fabio Trojani (University of Geneva; University of Turin; Swiss Finance Institute) |
| Abstract: | We introduce a comprehensive computational framework for solving dynamic portfolio choice problems with many risky assets, transaction costs, and borrowing and short-selling constraints. Our approach leverages the synergy between Gaussian process regression and Bayesian active learning to efficiently approximate value and policy functions with a novel, formal way of characterizing the irregularly-shaped no-trade region; we then embed this into a discrete-time dynamic programming algorithm. This combination allows us to study dynamic portfolio choice problems with more risky assets than was previously possible. Our results indicate that giving the agent access to more assets may alleviate some illiquidity resulting from the presence of transaction costs. |
| Keywords: | Machine learning, computational finance, computational economics, Gaussian process regression, dynamic portfolio optimization, transaction costs, liquidity premia |
| JEL: | C61 C63 C68 E21 |
| Date: | 2023–11 |
| URL: | http://d.repec.org/n?u=RePEc:chf:rpseri:rp23114&r=dcm |
| By: | Roger J. A. Laeven; Mitja Stadje |
| Abstract: | This paper axiomatizes, in a two-stage setup, a new theory for decision under risk and ambiguity. The axiomatized preference relation $\succeq$ on the space $\tilde{V}$ of random variables induces an ambiguity index $c$ on the space $\Delta$ of probabilities, a probability weighting function $\psi$, generating the measure $\nu_{\psi}$ by transforming an objective probability measure, and a utility function $\phi$, such that, for all $\tilde{v}, \tilde{u}\in\tilde{V}$, \begin{align*} \tilde{v}\succeq\tilde{u} \Leftrightarrow \min_{Q \in \Delta} \left\{\mathbb{E}_Q\left[\int\phi\left(\tilde{v}^{\centerdot}\right)\, \mathrm{d}\nu_{\psi}\right]+c(Q)\right\} \geq \min_{Q \in \Delta} \left\{\mathbb{E}_Q\left[\int\phi\left(\tilde{u}^{\centerdot}\right)\, \mathrm{d}\nu_{\psi}\right]+c(Q)\right\}. \end{align*} Our theory extends the rank-dependent utility model of Quiggin (1982) for decision under risk to risk and ambiguity, reduces to the variational preferences model when $\psi$ is the identity, and is dual to variational preferences when $\phi$ is affine in the same way as the theory of Yaari (1987) is dual to expected utility. As a special case, we obtain a preference axiomatization of a decision theory that is a rank-dependent generalization of the popular maxmin expected utility theory. We characterize ambiguity aversion in our theory. |
| Date: | 2023–12 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2312.05977&r=dcm |