The rationale for this finding comes from behavioral ecology studies that, in pre-modern hu- man population, social standing affects lifetime reproductive success through reproduction, which depends on both fertility and the survival of offspring, rather than through one’s own survival. In- corporating this effect into a biological model, Kageyama (2012a) shows that, as resources increase, reproduction and, thus, relative standing become more decisive to lifetime reproductive success, leading to the spread of preferences that induce greater concerns for relative standing when income increases.
This is rooted in the findings in Kageyama
（2012） , which empirically showed that the relationships between LEGAP and these happiness indicators are bidirectional. In one direction, LEGAP negatively affects both HPN and HPGAP . An increase in LEGAP raises women ’ s widowhood ratio, and, since widows are, on average, less happy, it lowers women ’ s average happiness, HPN, and HPGAP . We call this effect the “ marital-status composition effect ” as the marital-status composition plays a central role.
1.6 The FALSE statement below is:
(a). the extra revenue a firm gets from an extra unit of capital equals the marginal product of capital times the price of output.
(b). the extra revenue a firm gets from an extra unit of labor equals the marginal product of labor times real wage. (c). a perfectly competitive firm’s capital demand curve is the MPK schedule.
“participate,” and make the preference dependent on the number of “participants,” which is the externalities we consider.
To the best of our knowledge, Sasaki and Toda (1996) and Hafalir (2008) are the only papers that investigate a two-sided matching model with externalities. Both papers consider a very general form of externalities. Analyzing such matching models is di¢cult because preference is de…ned over the set of assignments rather than matchings. Hence, regular de…nition of “stability” or “deviation” are not su¢cient to analyze such a model because a deviating pair’s preference also depends on how other agents would react to their deviation, not just their matching. To model how other agents would react to a player’s deviation, both papers use what they call the estimation function approach. Estimation functions specify the expectations on the assignment (i.e., what the matching among all players would be) after each deviation. They prove that a strong requirement on the estimation function is necessary in order to guarantee the existence of stable matching. Based on their estimation function approach while considering a particular form of externalities (the payo¤ depends only on the number of operating …rms in the market), we show the existence and provide characterizations.
This paper also contributes to the empirical literature on the effect of financial con- straints on firm investment (e.g., Fazzari et al. (1988) ; Hoshi et al. (1991) ; Kaplan and Zin-
gales (1997) ). Empirical papers on the effect of financial constraints use various observed
measures of financial constraint, such as cash flow, firm size, and years of establishment, to examine their effect on investment. It is often difficult, however, to interpret these em- pirical results because such measures of financial constraint can be viewed as endogenous variables and correlated with the firm’s efficiency measure, which also explains investment. For example, a positive estimate of the cash flow coefficient could just reflect its positive correlation with firm efficiency. This paper examines how the BCR of the bank with which a firm has a relationship influences the firm’s investment decisions. To the extent that the BCR measure is viewed as more exogenous than other measures of financial constraint, this paper’s results shed further light on the impact of financial constraints on investment.
Advertising and research expenditures. Sutton’s (1991, 1998) seminal work focuses on two main
types of endogenous sunk costs: advertising and research outlays. Thus, a natural question is whether these costs can account for the empirical patterns documented. In particular, Park (2008) documents that advertising expenditure has increased over time in the mutual fund industry (in particular, for no-load funds), and this increase may have fostered concentration. However, it is unlikely that this alternative hypothesis can explain all our empirical findings, for several reasons. First, we wish to emphasize that our empirical model is designed to precisely control for spurious correlations due to unobserved factors, including advertising and research expenditures. In particular, the instruments that we employ in the empirical analysis exploit exogenous variations in the number of funds and number of categories offered. Second, many families offer funds in both the retail and the institutional segments of the market and, presumably, the effects of advertising and research (in particular) are not confined to a single segment of the market. Thus, it is not immediately obvious why market conduct and market structure respond differently to the same input. Third, and perhaps most important, Gallaher, Kaniel, and Starks (2009) investigate patterns of advertising in the mutual fund industry and find that families with funds in more objective classes advertise less than families with fewer objectives. Hence, if advertising were the key determinant of market structure and concentration in the retail segment, their finding would imply that families with funds in more objective classes should have a larger market share, in sharp contrast to the results of our analysis. Thus, we conclude that advertising and research expenditures cannot explain our empirical findings.
Mutual funds typically are members of a fund family. A fund’s risk factor loadings could also be influenced by the strategic consideration at the level of fund family. Nanda et al. (2004) show that a star fund attracts greater cash inflow not only to the fund but also to other funds in the family. Such a “spillover” effect can give incentives to a fund fam- ily to increase its chance of having a star fund. Goetzmann and Ibbotson (1993) argue that fund fam- ily can maximize the probability of having a star fund by lowering the cross-fund return correlation within the family. With differentiation in risk factor load- ings, product differentiation over the states of nature indeed lowers the correlation of returns across funds in the family and increases its chance of having a star fund in a certain state. It is conceivable that, compared to the standalone funds, the additional benefit of the spillover effect from having a star fund in a fund fam- ily would give funds in the family more incentives to engage in product differentiation over the state space. Our analysis considers the product differentiation over the state space through the differentiation in returns attributed to risk factor loadings. Differen- tiation over the state space could also be achieved through the differentiation in funds’ idiosyncratic returns. However, differentiation through idiosyn- cratic returns might not be as effective as differentia- tion through returns attributed to risk factor loadings. The average R-square of the four-risk-factor model in explaining mutual fund returns is around 90% (e.g., Carhart 1997), suggesting that return attributed to risk
Let us consider a panel data set of banks’ funding costs that would rationalize their bids in the Main Refinancing Operations of the European Central Bank that we obtain as in Section 3.5.1. Given this panel of banks’ revealed short-term funding costs, the network structure can be estimated by looking at the covariation of a given bank’s funding cost with other system banks’ (lagged) funding costs, controlling for various sources of comovement due to common asset exposures, such as holdings of sovereign bonds. Implementing this through traditional linear regression, however, yields a dimensionality problem: the number of covariates in the regressions potentially exceeds the number of data points available for each bank in the data set. To combat this problem, which is commonly encountered in Genetics and Machine learning, Bonaldi et al. (2013) utilize the adaptive elastic net method. This is a modification of the Least Absolute Shrinkage and Selection Operator (LASSO) method, which has been shown to have several desirable properties. In particular, as opposed to other methods for model selection, like the standard LASSO, the elastic net performs well under the presence of highly correlated covariates, as measured by prediction accuracy, which is a feature that likely applies in the case of financial networks.