Data: An Explanation of Real US Interest Rates with an Exchange Economy

The hypothesis is that an asset pricing model can explain real US short term bond interest rates. This is tested by using data to construct the three shocks of the model and inputting the shocks back into the model to produce the model generated US real bond interest rate from 1975-2020. This is then compared to the actual US data. The notable results are that the data matches the model generated data with a high correlation and relative volatility near one, indicating a close fit. The quarterly data set presents all variables used to fit the model to the data, for 1975Q1 to 2020Q4. All data series after construction are transformed by taking natural logarithms and detrending them to be in deviations from their respective trends. In the filtered results, we used a Hodrick-Prescott filter with λ=1600. In constructing real output, consumption, investment, government expenditures, and net exports real per capita series from raw data we follow Chari et al. (2007). [Chari, V. V., Kehoe, Patrick J., and McGrattan, Ellen R., 2007. "Business Cycle Accounting", Econometrica, vol. 75(3), pp. 781--836, May.] The final output series used is then obtained by deducted government expenditures and net exports from the total output to be consistent with our model. Quarterly employment and physical capital are obtained by interpolating annual data using the method in Chari et al. (2007). The goods sector labor share is measured by the Total Full-Time and Part-Time Employees minus the Full-Time and Part-Time Employees in Finance and Insurance Services (FIS) and divided by the Civilian Noninstitutional Population. The proxy for the banking time share is the same as employees in FIS as divided by the Civilian Noninstitutional Population. Leisure is then the residual share. The quarterly physical capital stock is constructed as the sum of the annual Current cost net stock of consumer durables and fixed assets interpolated. It is transformed into real terms by normalizing with the implicit price deflator for durable goods. The inflation measure is the CPI index, quarterly, with the percentage change from the year before (on an annual basis). Velocity measures are constructed by dividing real consumption with the respective real money stocks. The nominal series for exchange credit and deposits are transformed to real terms by normalizing with the CPI index. The data can be used in conjunction with the Matlab Code files for the model, which are attached here. This allows one to replicate the model results.本研究提出的核心假设为：某资产定价模型可以解释美国短期实际债券利率。该假设的验证流程为：利用数据集构建该模型的三类冲击，并将这些冲击重新输入模型，以生成1975年至2020年间的美国实际债券利率模拟值，随后将其与真实美国经济数据进行对比。研究得到的显著结果为：模拟数据与真实数据具有较高的相关性，且相对波动率接近1，表明模型拟合度极佳。 本数据集为1975年第一季度至2020年第四季度的季度数据，涵盖了用于模型拟合的全部变量。所有构建完成的序列均经过自然对数变换与去趋势处理，以得到各序列与其长期趋势的偏离值。在滤波处理环节，我们采用了参数λ=1600的霍德里克-普雷斯科特（Hodrick-Prescott）滤波器。 在从原始数据构建实际人均产出、消费、投资、政府支出以及净出口序列时，我们遵循了查里等人（Chari et al., 2007）的方法。[查里, V. V., 基奥, 帕特里克·J., 麦格拉坦, 埃伦·R., 2007. 《商业周期核算》, 《计量经济学》(Econometrica), 第75卷第3期, 第781--836页, 5月。] 最终使用的产出序列为总产出扣除政府支出与净出口后得到的结果，以与研究模型保持一致。 季度就业与物质资本存量数据通过对年度数据进行插值得到，插值方法同样来自Chari et al. (2007)。商品部门的劳动份额通过以下方式计算：用全体全职与兼职雇员人数减去金融与保险服务业（FIS）的全职及兼职雇员人数，再除以民用非机构化人口数。银行业时间份额的代理变量为金融与保险服务业雇员人数除以民用非机构化人口数。闲暇份额则为剩余部分。季度物质资本存量通过对年度当前成本计价的耐用消费品净存量与固定资产净存量求和并插值得到，并通过耐用消费品隐含价格平减指数进行标准化，以转换为实际值。 通胀指标采用季度消费者物价指数（CPI），以同比百分比变动的形式呈现。货币流通速度指标通过实际消费额除以对应实际货币存量计算得到。信贷与存款的名义序列通过除以CPI指数转换为实际值。 本数据集可与配套提供的模型Matlab代码文件结合使用，以实现研究结果的复现。