Personal experiences in shaping expectations--论文代写范文精选

这种差异逐渐消失值到1990年代,经过多年温和的通货膨胀。我们的模型解释这种差异是由于年轻个体感知的通货膨胀,直到1980年代初,通货膨胀率随之下降。下面的essay代写范文进一步阐述。

Abstract 
Our estimation results show that past experiences have an economically important effect on inflation expectations. Individuals of different ages differ significantly in their inflation expectations, and these differences are well explained by differences in their inflation experiences. The heterogeneity is particularly pronounced following periods of high surprise inflation. Consider again the strong divergence in expectations between younger and older cohorts during the late 1970s and early 1980s displayed in Figure 1. The higher expectations of younger individuals are consistent with their experience being dominated by the highinflation years of the 1970s, while older individuals also experienced the low-inflation years of the 1950s and 1960s. 

The discrepancy faded away only slowly by the 1990s, after many years of moderate inflation. Our model explains this difference as the result of younger individuals perceiving inflation to be (i) higher on average and (ii) more persistent when inflation rates were high until the early 1980s, but less persistent when inflation rates dropped subsequently. Our estimates of the gain parameter further imply that when individuals weight their accumulated life-time experiences, recent data receives higher weight than experiences earlier in life, though experiences from 20 to 30 years ago still have some measurable long-run effects. Going one step further, we also link the effect of experiences on beliefs to actual household decision-making. The experience-induced disagreement about future inflation leads to disagreement about real interest rates on assets and liabilities with nominally fixed long-term interest rates. 

Consistent with this idea, we show that the learning-from-experience model predicts the borrowing and investment decisions of households in the Survey of Consumer Finances. Households that forecast higher inflation according to the learning-from-experience model are more inclined than other individuals to borrow using fixed-rate mortgages, but not using variable-rate mortgages. They are also less likely (but with only marginal statistical significance) to invest in long-term bonds. Taken together, our results show that households with higher experience-induced inflation expectations tilt their exposure towards liabilities with nominally fixed rates rather than assets with nominally fixed rates. 

The effects are economically large. For instance, a one percentage point difference in the learningfrom-experience forecast of one-year inflation affects the mortgage balance by as much as a one-standard-deviation change in log income. Finally, we link learning from experience to aggregate expectations, as existing macro models with adaptive learning are commonly fit to the time series of aggregate (mean) in- flation expectations. We show that the cross-sectional average of the fitted learning-fromexperience forecasts at each point in time matches the average survey expectations closely. The similarity is remarkable because our estimation did not utilize any information about the level of the average survey expectations, only information about cross-sectional differences between cohorts. Learning from experience thus provides a natural micro-foundation for constant-gain learning algorithms that are popular in macroeconomics. In fact, the average learningfrom-experience forecast can be approximated quite closely with constant-gain learning: The constant-gain parameter that best matches the weighting of past data implied by our estimated learning-from-experience model, γ = 0.0180, is quantitatively very similar to those that Orphanides and Williams (2005a) and Milani (2007) have estimated by fitting the parameters to macroeconomic data and aggregate survey expectations, 0.02 and 0.0183, respectively. This similarity is remarkable because we did not calibrate the learning-from-experience rule to match the average level of inflation expectations or any macroeconomic data.

At the same time, learning from experience and constant-gain learning differ fundamentally in their motivation. The down-weighting of past data in constant-gain learning models is typically motivated as the response to structural changes in macroeconomic time series. Learning from experience, instead, is based on the notion that memory of past data is lost as older generations die and new ones are born. We show that the structural-change based explanation is hard to reconcile with the empirical evidence: households discount past inflation data much less than the degree of structural change in the time-series of inflation would call for. 

Moreover, macroeconomic time series, such as GDP and inflation, differ in the degree of structural change, but these differences does not seem to be reflected in the gains implied by survey expectations. Learning from experience also differs from existing models of adaptive learning in terms of identification. The gain parameter in our estimations is identified purely from crosssectional variation. The econometric advantage, compared to using aggregate time-series data to identify the gain parameter, is twofold. First, using a new dimension of data to pin down the parameters of individuals’ learning rule helps alleviate the concern about nonstandard (or, boundedly rational) learning models involving too many degrees of freedom and, hence, not being falsifiable, as expressed in Sargent (1993) and Marcet and Nicolini (2003). Second, empirically, the identification from cross-sectional data offers some advantage in capturing the dynamics of survey expectations. We show that estimating the gain parameter purely from cross-sectional variation provides a better out-of-sample fit to mean inflation expectations than estimating a constant-gain rule from time-series data, even though the estimation within the learning-from-experience framework does not use any data on the level of mean expectations.

Our paper connects to several strands of literature. A large macroeconomic literature analyzes the formation of expectations. While the crucial role of expectations for macroeconomic outcomes and asset prices is well understood at least since Keynes (1936), the empirical knowledge how economic agents form their beliefs about the future is more limited. The liter- 6 ature on adaptive learning (see Bray (1982); Sargent (1993); Evans and Honkapohja (2001)) views individual agents as econometricians who make forecasts based on simple forecasting rules estimated on historical data. Fuster, Laibson, and Mendel (2010) and Fuster, Hebert, and Laibson (2011) propose a model of “natural expectations” and demonstrate its ability to match hump-shaped dynamics in different economic time series（essay代写)

51Due网站原创范文除特殊说明外一切图文著作权归51Due所有；未经51Due官方授权谢绝任何用途转载或刊发于媒体。如发生侵犯著作权现象，51Due保留一切法律追诉权。
更多essay代写范文欢迎访问我们主页 www.51due.com 当然有essay代写需求可以和我们24小时在线客服 QQ:800020041 联系交流。-X（essay代写)