## 统计代写|经济统计代写Economic Statistics代考|ECO380

statistics-lab™ 为您的留学生涯保驾护航 在代写经济统计Economic Statistics方面已经树立了自己的口碑, 保证靠谱, 高质且原创的统计Statistics代写服务。我们的专家在代写经济统计Economic Statistics代写方面经验极为丰富，各种代写经济统计Economic Statistics相关的作业也就用不着说。

• Statistical Inference 统计推断
• Statistical Computing 统计计算
• (Generalized) Linear Models 广义线性模型
• Statistical Machine Learning 统计机器学习
• Longitudinal Data Analysis 纵向数据分析
• Foundations of Data Science 数据科学基础

## 统计代写|经济统计代写Economic Statistics代考|Forecasting with a Random Forest Algorithm

We leverage a random forest algorithm to evaluate whether Yelp measures can provide gains in nowcasting CBP measures before the release of official statistics. We are interested in the ability of Yelp to predict changes in overall CBP establishments and restaurants over and above the prediction power generated by lagged CBP data. Consequently, we begin our prediction task by regressing the change in CBP establishments on the two lags of changes in CBP establishments and zip code and year fixed effects. We then work with the residual quantity. Given the two lags of the CBP, our sample spans years 2012 to 2015 . We use a relatively simple first stage regression because we have a limited number of years, and because modest increases in complexity add little predictive power.

We assign the last year of our dataset (2015) to the test set, which represents 25 percent of our sample, and the rest to the training set. We then examine the ability of lagged and contemporaneous Yelp data to predict residual changes in CBP number of establishments in a given zip code and year. We include the following Yelp measures in the feature set: contemporaneous and lagged changes in, and absolute count of, the total number of open, opened, and closed businesses; aggregate review counts; and the average rating of businesses, all in terms of total numbers and broken down by lowest and highest price level, along with the year and total number of businesses that closed within one year. The number of trees in the forest is set to 300 , and the gains to increasing this number are marginal, yielding very similar results. Using an off-the-shelf random forest algorithm on models with limited feature sets, our analyses represent basic exercises to evaluate the usefulness of Yelp data, rather than to provide the most precise forecasts.
Table $9.4$ shows the prediction results. The first column shows our results for CBP establishments overall, while the second column shows the results for restaurants. We evaluate the predictive power of our model in two ways. Using the 2012-2014 data, we use an “out-of-bag” estimate of the prediction accuracy. We also use the 2015 data as a distinct testing sample.

The first row shows that the model has an $R^2$ of $0.29$ for predicting the 2014-2015 CBP openings for all businesses and an $R^2$ of $0.26$ for restaurants. Since the baseline data were already orthogonalized with respect to year, this implies that the Yelp-based model can explain between one-quarter and one-third of the variation across zip codes in the residualized CBP data.
The second row shows the out-of-bag estimates of $R^2$, based on the training data. In this case, the $R^2$ is $0.21$ for both data samples. The lower $R^2$ is not surprising given that out-of-bag estimates can often understate the predictive power of models. Nonetheless, it is useful to know that the fit of the model is not particular to anything about 2015.

There appears to be a wide range of predictive ability-but on average bounded within approximately half a standard deviation for businesses, with $8.0$ mean absolute error (MAE) and $3.9$ median absolute error, compared to a mean of $3.4$ and a standard deviation of 15.1. The mean and median absolute errors for restaurants are substantially smaller than for businesses, at $1.7$ and 1.1, respectively, but the mean and standard deviation for restaurant growth are also substantially lower than for businesses, at $0.54$ and $2.9$, respectively.

Yelp’s predictive power is far from perfect, but it does provide significant improvement in our knowledge about the path of local economies. Adding Yelp data can help marginally improve predictions compared to using only prior CBP data.

## 统计代写|经济统计代写Economic Statistics代考|Interactions with Area Attributes

Table $9.5$ shows results from regressions where changes in Yelp’s open business numbers are interacted with indicators for geographic characteristics. We use indicator variables that take on a value of one if the area has greater than the median level of population density, income, and education, and zero otherwise. Population density estimates are from the 2010 Census, while measures of median household income and percentage with a bachelor’s degree are from the 2015 American Community Survey five-year estimates. We present results just for total establishments and begin with the simple specification of regression (2) in table 9.2.

In this first regression, we find that all three interaction terms are positive and statistically significant. The interaction with high population density is $0.14$, while the interaction with high income is $0.30$, and the interaction with high education is $0.09$. Together, these interactions imply that the coeffi-cient on contemporaneous Yelp openings is $0.2$ in a low-density, low-income and low-education zip code, and $0.73$ in a high-density, high-income, and high-education zip code. This is an extremely large shift in coefficient size, perhaps best explained by far greater usage of Yelp in places with higher density, higher income, and higher education. If higher usage leads to more accuracy, this should cause the attenuation bias to fall and the estimated coefficient to increase.

In the second regression, we also add lagged Yelp openings. In this case,the baseline coefficient is negative, but again all three interactions are positive. Consequently, the estimated coefficient on lagged Yelp openings is $-0.1$ in low-density, low-income, and low-education locales, but $0.24$ in highdensity, high-income, and high-education areas. Again, decreased attenuation bias is one possible interpretation of this change. The third regression includes changes in Yelp closings and the number of Yelp reviews.

These interactions suggest that the predictive power of Yelp is likely to be higher in places with more density, education, and income. However, it is not true that adding interactions significantly improves the overall $R^2$. There is also little increase in $R^2$ from adding the lag of Yelp openings or the other Yelp variables, just as in table 9.2. While contemporaneous Yelp openings is the primary source of explanatory power, if policy makers want to use Yelp openings to predict changes in establishments, they should recognize that the mapping between contemporaneous Yelp openings and CBP openings is different in different places.

## 统计代写|经济统计代写Economic Statistics代考|Forecasting with a Random Forest Algorithm

Yelp 的预测能力远非完美，但它确实大大提高了我们对当地经济发展路径的了解。与仅使用之前的 CBP 数据相比，添加 Yelp 数据有助于略微改进预测。

## 有限元方法代写

tatistics-lab作为专业的留学生服务机构，多年来已为美国、英国、加拿大、澳洲等留学热门地的学生提供专业的学术服务，包括但不限于Essay代写，Assignment代写，Dissertation代写，Report代写，小组作业代写，Proposal代写，Paper代写，Presentation代写，计算机作业代写，论文修改和润色，网课代做，exam代考等等。写作范围涵盖高中，本科，研究生等海外留学全阶段，辐射金融，经济学，会计学，审计学，管理学等全球99%专业科目。写作团队既有专业英语母语作者，也有海外名校硕博留学生，每位写作老师都拥有过硬的语言能力，专业的学科背景和学术写作经验。我们承诺100%原创，100%专业，100%准时，100%满意。

## MATLAB代写

MATLAB 是一种用于技术计算的高性能语言。它将计算、可视化和编程集成在一个易于使用的环境中，其中问题和解决方案以熟悉的数学符号表示。典型用途包括：数学和计算算法开发建模、仿真和原型制作数据分析、探索和可视化科学和工程图形应用程序开发，包括图形用户界面构建MATLAB 是一个交互式系统，其基本数据元素是一个不需要维度的数组。这使您可以解决许多技术计算问题，尤其是那些具有矩阵和向量公式的问题，而只需用 C 或 Fortran 等标量非交互式语言编写程序所需的时间的一小部分。MATLAB 名称代表矩阵实验室。MATLAB 最初的编写目的是提供对由 LINPACK 和 EISPACK 项目开发的矩阵软件的轻松访问，这两个项目共同代表了矩阵计算软件的最新技术。MATLAB 经过多年的发展，得到了许多用户的投入。在大学环境中，它是数学、工程和科学入门和高级课程的标准教学工具。在工业领域，MATLAB 是高效研究、开发和分析的首选工具。MATLAB 具有一系列称为工具箱的特定于应用程序的解决方案。对于大多数 MATLAB 用户来说非常重要，工具箱允许您学习应用专业技术。工具箱是 MATLAB 函数（M 文件）的综合集合，可扩展 MATLAB 环境以解决特定类别的问题。可用工具箱的领域包括信号处理、控制系统、神经网络、模糊逻辑、小波、仿真等。

## 统计代写|经济统计代写Economic Statistics代考|ECON227

statistics-lab™ 为您的留学生涯保驾护航 在代写经济统计Economic Statistics方面已经树立了自己的口碑, 保证靠谱, 高质且原创的统计Statistics代写服务。我们的专家在代写经济统计Economic Statistics代写方面经验极为丰富，各种代写经济统计Economic Statistics相关的作业也就用不着说。

• Statistical Inference 统计推断
• Statistical Computing 统计计算
• (Generalized) Linear Models 广义线性模型
• Statistical Machine Learning 统计机器学习
• Longitudinal Data Analysis 纵向数据分析
• Foundations of Data Science 数据科学基础

## 统计代写|经济统计代写Economic Statistics代考|Comparing Restaurant Coverage on Yelp

We first compare Yelp and CBP restaurant numbers to paint a more detailed picture of Yelp coverage across geography. In 2015 (the last year of CBP data available), 27,074 zip codes out of 33,120 ZCTAs listed in the US in 2010 had at least one restaurant in either the CBP or Yelp. ${ }^5$ The CBP listed 542,029 restaurants in 24,790 zip codes, and Yelp listed 576,233 restaurants in 22,719 zip codes. There were 2,284 zip codes with at least one Yelp restaurant but no CBP restaurants, and 4,355 zip codes with at least one CBP restaurant and no Yelp restaurants.

We focus on Yelp coverage ratios, which are defined as the ratio of Yelp restaurants to CBP restaurants. Since we match the data by geography and not by establishment, there is no guarantee that the same establishments are being counted in the two data sources. Nationwide, the Yelp coverage ratio is $106.3$ percent, meaning that Yelp captures more establishments, presumably disproportionately smaller ones, than it misses. ${ }^6$ Approximately 95 percent of the population in our sample live in zip codes where the number of Yelp restaurants is at least 50 percent of the number of CBP restaurants, and over 50 percent of the population in our zip code sample live in zip codes with more Yelp restaurants than CBP restaurants (see figure 9.3).

Yelp coverage of CBP restaurants is strongly correlated with population density. In the 1,000 most sparsely populated zip codes covered by the CBP, mean Yelp coverage is 88 percent (median coverage $=67$ percent), while in the 1,000 densest zip codes, mean coverage is 126 percent (median coverage $=123$ percent $)$. Figure $9.4$ shows the relationship between Yelp coverage of CBP restaurants and population density across all zip codes covered by the CBP, plotting the average Yelp/CBP ratio for each equal-sized bin of population density. The relationship is at first negative and then positive for population density levels above 50 people per square mile.

The nonmonotonicity may simply reflect a nonmonotonicity in the share of restaurants with no employees, which in turn reflects offsetting supply and demand side effects. In zip codes with fewer than 50 people per square mile, Yelp tends to report one or two restaurants in many of these areas whereas the CBP reports none. Extremely low-density levels imply limited restaurant demand, which may only be able to support one or two small establishments. High-density levels generate robust demand for both large and small establishments, but higher-density areas may also have a disproportionately abundant supply of small-scale, often immigrant entrepreneurs.

## 统计代写|经济统计代写Economic Statistics代考|Regression Analysis

Table $9.2$ shows results from regressing changes in CBP business numbers on prior CBP and Yelp measures. Column (1) regresses changes in the CBP’s number of businesses in year $t$ on two lags of the CBP. The addition of one CBP establishment in the previous year is associated with an increase of $0.27$ businesses in year $t$, showing that there is positive serial correlation in the growth of businesses at the zip code level. The correlation is also strongly positive with a two-year lag of CBP business openings. Together, the two lags of changes in CBP establishments explain $14.8$ percent of the variance (as measured by adjusted $R^2$ ).

Column 2 of table $9.2$ regresses changes in CBP business numbers in year $t$ on two lags of the CBP and the contemporaneous change in Yelp business numbers. Adding contemporaneous Yelp business numbers increases the variance explained to $22.5$ percent. A one-unit change in the number of Yelp businesses in the same year is associated with an increase in the number of CBP businesses of $0.6$. This coefficient is fairly precisely estimated, so that with 99 percent confidence, a one-unit increase in the number of Yelp establishments is associated with an increase between $0.55$ and $0.66$ in CBP establishments in the same year, holding two years of lagged CBP establishment growth constant.

The prediction of a purely accounting model of establishments is that the coefficient should equal one, but there are at least two reasons why that prediction will fail. First, if there is measurement error in the Yelp variable, that will push the coefficient below one due to attenuation bias. Second, Yelp does not include many CBP establishments, especially in industries other than retail. If growth in retail is associated with growth in other industries, then the coefficient could be greater than one, which we term spillover bias and expect to be positive. The estimated coefficient of $0.61$ presumably reflects a combination of attenuation and spillover bias, with spillover bias dominating.

Columns 3 and 4 of table $9.2$ show that lagged Yelp data, as well as other Yelp variables including the number of closures and reviews, are only mildly informative in explaining the variance of CBP business number growth. Growth in CBP establishments is positively associated with a one-year lag in the growth in the number of Yelp establishments, and including that variable causes the coefficient on contemporary establishment growth to drop to 0.44. Regression (4) also shows that increases in the number of Yelp closings are negatively correlated with growth in the number of CBP establishments, and that the number of Yelp reviews is not correlated with growth in the number of CBP establishments. Some of these extra Yelp variables are statistically significant, but they add little to overall explanatory power. The adjusted $R^2$ only rises from $0.225$ to $0.229$ between regression (2) and regression (4). The real improvement in predictive power comes from the inclusion of contemporaneous Yelp openings, not from the more complex specification. This suggests that simply looking at current changes in the number of Yelp establishments may be enough for most local policy makers who are interested in assessing the current economic path of a neighborhood.

## 统计代写|经济统计代写Economic Statistics代考|Comparing Restaurant Coverage on Yelp

CBP 餐厅的 Yelp 覆盖率与人口密度密切相关。在 CBP 覆盖的 1,000 个人口最稀少的邮政编码中，Yelp 的平均覆盖率为 88%（覆盖率中位数=67百分比），而在 1,000 个最密集的邮政编码中，平均覆盖率为 126%（中位数覆盖=123百分). 数字9.4显示了 CBP 餐厅的 Yelp 覆盖率与 CBP 覆盖的所有邮政编码的人口密度之间的关系，绘制了每个等大小人口密度箱的平均 Yelp/CBP 比率。对于每平方英里 50 人以上的人口密度水平，该关系首先为负，然后为正。

## 有限元方法代写

tatistics-lab作为专业的留学生服务机构，多年来已为美国、英国、加拿大、澳洲等留学热门地的学生提供专业的学术服务，包括但不限于Essay代写，Assignment代写，Dissertation代写，Report代写，小组作业代写，Proposal代写，Paper代写，Presentation代写，计算机作业代写，论文修改和润色，网课代做，exam代考等等。写作范围涵盖高中，本科，研究生等海外留学全阶段，辐射金融，经济学，会计学，审计学，管理学等全球99%专业科目。写作团队既有专业英语母语作者，也有海外名校硕博留学生，每位写作老师都拥有过硬的语言能力，专业的学科背景和学术写作经验。我们承诺100%原创，100%专业，100%准时，100%满意。

## MATLAB代写

MATLAB 是一种用于技术计算的高性能语言。它将计算、可视化和编程集成在一个易于使用的环境中，其中问题和解决方案以熟悉的数学符号表示。典型用途包括：数学和计算算法开发建模、仿真和原型制作数据分析、探索和可视化科学和工程图形应用程序开发，包括图形用户界面构建MATLAB 是一个交互式系统，其基本数据元素是一个不需要维度的数组。这使您可以解决许多技术计算问题，尤其是那些具有矩阵和向量公式的问题，而只需用 C 或 Fortran 等标量非交互式语言编写程序所需的时间的一小部分。MATLAB 名称代表矩阵实验室。MATLAB 最初的编写目的是提供对由 LINPACK 和 EISPACK 项目开发的矩阵软件的轻松访问，这两个项目共同代表了矩阵计算软件的最新技术。MATLAB 经过多年的发展，得到了许多用户的投入。在大学环境中，它是数学、工程和科学入门和高级课程的标准教学工具。在工业领域，MATLAB 是高效研究、开发和分析的首选工具。MATLAB 具有一系列称为工具箱的特定于应用程序的解决方案。对于大多数 MATLAB 用户来说非常重要，工具箱允许您学习应用专业技术。工具箱是 MATLAB 函数（M 文件）的综合集合，可扩展 MATLAB 环境以解决特定类别的问题。可用工具箱的领域包括信号处理、控制系统、神经网络、模糊逻辑、小波、仿真等。

## 统计代写|经济统计代写Economic Statistics代考|ECON202

statistics-lab™ 为您的留学生涯保驾护航 在代写经济统计Economic Statistics方面已经树立了自己的口碑, 保证靠谱, 高质且原创的统计Statistics代写服务。我们的专家在代写经济统计Economic Statistics代写方面经验极为丰富，各种代写经济统计Economic Statistics相关的作业也就用不着说。

• Statistical Inference 统计推断
• Statistical Computing 统计计算
• (Generalized) Linear Models 广义线性模型
• Statistical Machine Learning 统计机器学习
• Longitudinal Data Analysis 纵向数据分析
• Foundations of Data Science 数据科学基础

## 统计代写|经济统计代写Economic Statistics代考|Machine Learning

The vector outputs from Doc2Vec models lend themselves well to unsupervised classification techniques such as clustering. They can also function as features (independent variables) in supervised machine learning algorithms. After matching our data to the BR, we get the actual NAICS sector codes for each establishment matched, which we use as our dependent variable. We build a Random Forest model-based classifier to predict the NAICS sector of each establishment, where the independent variables are the generated vectors for business name, user reviews, and websites, as well as a series of binary variables indicating the Google Type tag for each establishment. Random Forests are a method of classification techniques derived from Decision Tree classifiers but are relatively immune to overfitting that often impacts Decision Trees. In some cases, Random Forests outperform more common approaches such as logistic regression in class-imbalanced circumstances (Muchlinski et al. 2016). The 120,000 records are split into 80 percent training and 20 percent validation set for model training and evaluation.

In order to ensure our model selection is both replicable and maximizes accuracy, we performed an analysis of 1,000 different model configurations. We randomly alter the number of vectors a Doc 2 Vec model produces, as well as how many, and how deep, the trees are in the Random Forest model. We then tested how those different model configurations altered the accuracy and repeat this process. Minimum log-loss is chosen as the model comparison criteria, as log-loss is a penalizing function that allows us to weigh the trade-off between the prediction and its certainty. Log-loss penalizes incorrect predictions with high predicted probabilities but does not penalize less certain incorrect assumptions. For our purposes, this is an ideal trade-off, as the comparable SSA Autocoder does not assign NAICS codes if the predicted probability is less than $0.638$ (Kearney and Kornbau 2005). Hence, any system based on our model will need to be sensitive to the need to prevent assigning incorrect codes without high levels of certainty.

## 统计代写|经济统计代写Economic Statistics代考|Predictive Accuracy

The findings here discuss our best fitting model, which utilizes 119 trees in the Random Forest, with 20 vectors for business name, 8 for user reviews, and 16 for websites. Overall, across all NAICS sectors, and for SU establishments only, our model predicts approximately 59 percent of cases accurately. This places our model substantially below the current autocoding methods used by the SSA; however, it is at a similar level to initial match rates for the SSA method, and shows comparable performance to similar exercises in other countries (Kearney and Kornbau 2005; Roelands, van Delden, and Windmeijer 2017). The model also exhibits considerable variation, with some NAICS codes (Information, Manufacturing) seeing fewer than 5 percent of observations correctly predicted, while Accommodation and Food Services has approximately 83 percent of establishments correctly predicted into their NAICS sector. Given the unbalanced nature of our sample, evaluating strictly on accuracy may be misleading – it would encourage a model to overfit to only large NAICS codes. Instead, we use the F1 score to evaluate our model. ${ }^9$

Figure $8.6$ shows a scatter plot of the average number of words unique to the NAICS sector in our data (from figure 8.3) on the $\mathrm{x}$-axis, and the $\mathrm{F} 1$ Score for each NAICS sector on the y-axis. Clearly, Accommodation and Food Services, and Retail Trade have the highest F1 scores, and corresponding highest percentage of unique words. Similarly, F1 scores for Information, Wholesale Trade, and Manufacturing sectors are exceedingly low and also have the least percentage of unique words appearing in those NAICS codes. This clear relationship demonstrates encouraging signs of this modeling and approach-words that are unique to a certain NAICS code represent a better signal for a model to use as a classifier. Therefore, we argue that our model performance will improve with additional data from undersampled sectors. Although the increase in number of unique words may not be linear compared to the number of observations, our findings point directly to our model not able to correctly predict businesses in a sector from a relatively small number of unique words, which may be ameliorated with a broader search.

## 统计代写|经济统计代写Economic Statistics代考|Machine Learning

Doc2Vec 模型的矢量输出非常适合无监督分类技术，例如聚类。它们还可以在监督机器学习算法中充当特征（自变量）。将我们的数据与 BR 匹配后，我们得到每个匹配机构的实际 NAICS 部门代码，我们将其用作因变量。我们构建了一个基于随机森林模型的分类器来预测每个机构的 NAICS 部门，其中自变量是为企业名称、用户评论和网站生成的向量，以及一系列指示 Google Type 标签的二进制变量每个机构。随机森林是一种源自决策树分类器的分类技术方法，但相对不受经常影响决策树的过度拟合的影响。在某些情况下，随机森林优于更常见的方法，例如类不平衡情况下的逻辑回归（Muchlinski 等人，2016 年）。120,000 条记录分为 80% 的训练集和 20% 的验证集，用于模型训练和评估。

## 有限元方法代写

tatistics-lab作为专业的留学生服务机构，多年来已为美国、英国、加拿大、澳洲等留学热门地的学生提供专业的学术服务，包括但不限于Essay代写，Assignment代写，Dissertation代写，Report代写，小组作业代写，Proposal代写，Paper代写，Presentation代写，计算机作业代写，论文修改和润色，网课代做，exam代考等等。写作范围涵盖高中，本科，研究生等海外留学全阶段，辐射金融，经济学，会计学，审计学，管理学等全球99%专业科目。写作团队既有专业英语母语作者，也有海外名校硕博留学生，每位写作老师都拥有过硬的语言能力，专业的学科背景和学术写作经验。我们承诺100%原创，100%专业，100%准时，100%满意。

## MATLAB代写

MATLAB 是一种用于技术计算的高性能语言。它将计算、可视化和编程集成在一个易于使用的环境中，其中问题和解决方案以熟悉的数学符号表示。典型用途包括：数学和计算算法开发建模、仿真和原型制作数据分析、探索和可视化科学和工程图形应用程序开发，包括图形用户界面构建MATLAB 是一个交互式系统，其基本数据元素是一个不需要维度的数组。这使您可以解决许多技术计算问题，尤其是那些具有矩阵和向量公式的问题，而只需用 C 或 Fortran 等标量非交互式语言编写程序所需的时间的一小部分。MATLAB 名称代表矩阵实验室。MATLAB 最初的编写目的是提供对由 LINPACK 和 EISPACK 项目开发的矩阵软件的轻松访问，这两个项目共同代表了矩阵计算软件的最新技术。MATLAB 经过多年的发展，得到了许多用户的投入。在大学环境中，它是数学、工程和科学入门和高级课程的标准教学工具。在工业领域，MATLAB 是高效研究、开发和分析的首选工具。MATLAB 具有一系列称为工具箱的特定于应用程序的解决方案。对于大多数 MATLAB 用户来说非常重要，工具箱允许您学习应用专业技术。工具箱是 MATLAB 函数（M 文件）的综合集合，可扩展 MATLAB 环境以解决特定类别的问题。可用工具箱的领域包括信号处理、控制系统、神经网络、模糊逻辑、小波、仿真等。

## 统计代写|经济统计代写Economic Statistics代考|ECON121

statistics-lab™ 为您的留学生涯保驾护航 在代写经济统计Economic Statistics方面已经树立了自己的口碑, 保证靠谱, 高质且原创的统计Statistics代写服务。我们的专家在代写经济统计Economic Statistics代写方面经验极为丰富，各种代写经济统计Economic Statistics相关的作业也就用不着说。

• Statistical Inference 统计推断
• Statistical Computing 统计计算
• (Generalized) Linear Models 广义线性模型
• Statistical Machine Learning 统计机器学习
• Longitudinal Data Analysis 纵向数据分析
• Foundations of Data Science 数据科学基础

## 统计代写|经济统计代写Economic Statistics代考|Related Literature

Ours is not the first paper to make use of ADP payroll data. Several papers study the National Employment Report (NER), ADP’s publicly available monthly estimate of US payroll gains constructed jointly with Moody’s Analytics. Importantly, NER estimates are derived from a model including not only ADP microdata but also other contemporaneous and lagged indicators of US economic activity. The existing literature finds that the NER moves closely with CES (Phillips and Slijk 2015) and has some ability to forecast CES, though it does not appear to improve forecasts based on other available information, such as existing consensus forecasts (Gregory and Zhu 2014; Hatzius et al. 2016).

As noted above, we do not use the NER but instead focus on the ADP microdata. A number of recent papers explore these data. Cajner et al. (2018) analyze the representativeness of ADP microdata (relative to CES and QCEW) and construct an ADP payroll index that can improve forecasts of CES; we employ that index in the present paper. Ozimek, DeAntonio and Zandi (2017) use ADP’s linked employer-employee microdata to study the negative effect of workforce aging on aggregate productivity growth. Grigsby, Hurst, and Yildirmaz (2021) study wage rigidity in the same data, finding that the high-frequency microdata can be useful for shedding light on a key business cycle question. Cho (2018) uses ADP microdata to study the employment and wage effects of the 2009 American Recovery and Reinvestment Act.

Our approach in the present paper is different from those above in that we explicitly investigate the usefulness of $\mathrm{ADP}$ as a supplement to CES data for tracking the underlying state of the labor market. In this respect, our work is inspired by Aruoba et al. (2016), who note difficulties in assessing the growth of aggregate output in real time given limitations on the comprehensiveness and timeliness of GDP measures. Two independent measures of GDP exist – the commonly reported expenditure-side approach and the income-based approach-and both are prone to measurement errors arising from various sources. Aruoba et al. (2016) combine the two measures using a state-space framework, recovering an underlying state of output growth which they label “gross domestic output.” We follow this general approach with a focus on employment rather than output.

## 统计代写|经济统计代写Economic Statistics代考|Structure of the ADP Microdata

ADP provides human capital management services to firms, including payroll processing. Processing payroll for a client firm involves many tasks, including maintaining worker records, calculating taxes, and issuing paychecks. ADP processes payroll for about 26 million US workers each month (about 20 percent of total US private employment). The structure of the microdata is determined by the business needs of ADP. ADP maintains records at the level of payroll account controls (PAC), which often correspond to business establishments (but may sometimes correspond to firms) as defined by the Census Bureau and BLS. Each PAC updates their records at the end of each pay period. The records consist of the date payroll was processed, employment information for the pay period, and many time-invariant PAC characteristics (such as an anonymized PAC identifier, NAICS industry code, zip code, etc.). PAC records include both the number of individuals employed (“active employees”) and the number of individuals issued a paycheck in a given pay period (“paid employees”). Active employees include wage earners with no hours in the pay period, workers on unpaid leave, and the like. Paid employees include any wage or salary workers issued regular paychecks during the pay period as well as those issued bonus checks and payroll corrections. In this paper we focus exclusively on active employment, having found that it is substantially less volatile, more closely resembles officially published aggregates, and performs better in forecasting exercises, though we plan to further investigate the active/paid distinction in the future. ${ }^6$

The data begin in July $1999 .{ }^7$ In terms of frequency, the files we use are weekly snapshots of individual PAC records, taken every Saturday since July 2009 (snapshots were taken semimonthly between May 2006 and June 2009 and monthly before May 2006). Each snapshot contains the most recent pay date for each PAC, the relevant employment counts, and the other information described above. As few firms regularly process payroll more than once per week, the weekly snapshots provide a comprehensive history of PAClevel employment dynamics. ${ }^8$

We can compare ADP payroll microdata to the QCEW and CES data in terms of pay frequency, region, establishment size, and industry composition. Most notably, ADP has significantly more employment in midsized units than does CES, with a distribution that looks reasonably similar to QCEW. ${ }^9$

## 统计代写|经济统计代写经济统计代考|相关文献

ADP为企业提供人力资本管理服务，包括工资处理。为客户公司处理工资单涉及许多任务，包括维护工人记录、计算税收和签发工资单。ADP每月为大约2600万美国工人处理工资单(约占美国私人就业总量的20％)。微数据的结构由ADP的业务需求决定。ADP保持工资账户控制(PAC)级别的记录，PAC通常对应于人口普查局和劳工统计局定义的商业机构(但有时也对应于公司)。每个PAC在每个付款期结束时更新其记录。这些记录包括处理工资单的日期、支付期间的就业信息和许多时不变的PAC特征(例如匿名的PAC标识符、NAICS行业代码、邮政编码等)。PAC记录包括被雇用的人数(“在职雇员”)和在给定的工资期间发放工资的人数(“有薪雇员”)。在职雇员包括在有薪期间没有工作时间的领工资者、无薪休假的工人等等。受薪员工包括在支付期间签发定期工资支票的员工，以及签发奖金支票和工资更正的员工。在这篇论文中，我们专门关注积极就业，因为我们发现它的波动性更小，更接近官方公布的总量，在预测练习中表现更好，尽管我们计划在未来进一步研究积极/付费的区别。${ }^6$

## 有限元方法代写

tatistics-lab作为专业的留学生服务机构，多年来已为美国、英国、加拿大、澳洲等留学热门地的学生提供专业的学术服务，包括但不限于Essay代写，Assignment代写，Dissertation代写，Report代写，小组作业代写，Proposal代写，Paper代写，Presentation代写，计算机作业代写，论文修改和润色，网课代做，exam代考等等。写作范围涵盖高中，本科，研究生等海外留学全阶段，辐射金融，经济学，会计学，审计学，管理学等全球99%专业科目。写作团队既有专业英语母语作者，也有海外名校硕博留学生，每位写作老师都拥有过硬的语言能力，专业的学科背景和学术写作经验。我们承诺100%原创，100%专业，100%准时，100%满意。

## MATLAB代写

MATLAB 是一种用于技术计算的高性能语言。它将计算、可视化和编程集成在一个易于使用的环境中，其中问题和解决方案以熟悉的数学符号表示。典型用途包括：数学和计算算法开发建模、仿真和原型制作数据分析、探索和可视化科学和工程图形应用程序开发，包括图形用户界面构建MATLAB 是一个交互式系统，其基本数据元素是一个不需要维度的数组。这使您可以解决许多技术计算问题，尤其是那些具有矩阵和向量公式的问题，而只需用 C 或 Fortran 等标量非交互式语言编写程序所需的时间的一小部分。MATLAB 名称代表矩阵实验室。MATLAB 最初的编写目的是提供对由 LINPACK 和 EISPACK 项目开发的矩阵软件的轻松访问，这两个项目共同代表了矩阵计算软件的最新技术。MATLAB 经过多年的发展，得到了许多用户的投入。在大学环境中，它是数学、工程和科学入门和高级课程的标准教学工具。在工业领域，MATLAB 是高效研究、开发和分析的首选工具。MATLAB 具有一系列称为工具箱的特定于应用程序的解决方案。对于大多数 MATLAB 用户来说非常重要，工具箱允许您学习应用专业技术。工具箱是 MATLAB 函数（M 文件）的综合集合，可扩展 MATLAB 环境以解决特定类别的问题。可用工具箱的领域包括信号处理、控制系统、神经网络、模糊逻辑、小波、仿真等。

## 统计代写|经济统计代写Economic Statistics代考|ECON502

statistics-lab™ 为您的留学生涯保驾护航 在代写经济统计Economic Statistics方面已经树立了自己的口碑, 保证靠谱, 高质且原创的统计Statistics代写服务。我们的专家在代写经济统计Economic Statistics代写方面经验极为丰富，各种代写经济统计Economic Statistics相关的作业也就用不着说。

• Statistical Inference 统计推断
• Statistical Computing 统计计算
• (Generalized) Linear Models 广义线性模型
• Statistical Machine Learning 统计机器学习
• Longitudinal Data Analysis 纵向数据分析
• Foundations of Data Science 数据科学基础

## 统计代写|经济统计代写Economic Statistics代考|Adjustments to the First and Last Month of the Constant-Merchant Sample

Before we combine information from the overlapping 14-month merchant samples, we need to correct for a bias at the beginning and end of the samples. For each month in the dataset (excepting the first 13 months and the most recent 13 months), there are exactly fourteen 14-month samples that have a sales estimate for that month, and thirteen 14-month samples that have a monthly sales growth estimate for that month (which requires that months $t$ and $t-1$ be in the sample). Although the monthly level of sales in each sample is highly dependent on the merchant births, deaths, and business acquisitions between overlapping 14-month merchant samples, we find that the estimates of monthly growth in different samples are, on average, similar, with two notable exceptions: The first monthly growth estimate from a 14-month merchant sample is biased upwards, and the last monthly growth estimate is biased downwards. To make things more explicit, call $g_t^{t+j}$ the estimate of monthly growth in time $t$ that comes from the 14-month sample ending in month $t+j$. For each month $t$, we construct the average growth rate, $gt$ using all 14-month samples that include an estimate of the growth rate in $t$ : $$g_t=\frac{1}{13} \sum{j=0}^{12} g_t^{t+j} \text {. }$$
Next, we calculate the deviation of the growth estimate $t$ from a merchant sample $t+j$ relative to the average across all samples:
deviation from mean $(j, t)=g_t^{(+j}-g_t$.
In figure 4B.1, we plot the distribution of deviations in all calendar months in the dataset, based on where the growth estimate falls in the merchant sample window (the index $j$ ). ${ }^{22}$ The upward bias at the beginning of the 14month sample – that is, the growth rate at time $t$ for the sample that runs from $t-1$ through $t+12$-comes from a “birthing” bias due to firms that were just born and who are therefore ramping up sales. Equivalently, the downward bias at the end of a sample – the growth rate that runs from $t-13$ through $t$-is from the fact that firms that are about to die (say in time $t+1$, just after the sample ends) tend to have falling sales.

## 统计代写|经济统计代写Economic Statistics代考|Mathematical Derivation of Birth and Death Bias

The main disadvantage of the constant-merchant methodology described above is that we cannot capture true economic births and deaths. To show the bias that may result, we introduce some notation. In a given month $t$ let $x_t$ be the total consumer spending in that month so that the true monthly growth rate of consumer spending is simply:
$$g_t=\frac{x_t}{x_{t-1}}-1 .$$
Some set of firms transact in both period $t$ and $t-1$ and we can call the spending at these firms in time $t, s_t^{-}$(where the minus denotes that these are the firms that existed in both that period and the previous one, so $t$ and $t-1$ ) and, in time $t-1, s_{t-1}^{+}$(where the plus denotes the firms that existed in both that period and the following one, so $t-1$ and $t$ ). The growth rate of spending for merchants who transact in both periods, what we will refer to as “constant-merchant” growth, is simply:
$$\hat{g}t=\frac{s_t^{-}}{s{t-1}^{+}}-1 .$$
However, we know that in every period new establishments are born, and we assume that they make up some fraction $b_t$ of the sales in the previous period so that their total sales in the current period $t$ are $b_2 x_{t-1}$. Similarly, some fraction, $d_t$, of total sales are by firms that die at the end of the period such that total sales in period $t-1$ can be expressed as:
$$x_{t-1}=\frac{s_{t-1}^{+}}{\left(1-d_{t-1}\right)} .$$
And sales in period $t$ can be written as:
$$x_t=s_t^{-}+b_t \frac{s_{t-1}^{+}}{\left(1-d_{t-1}\right)} .$$

## 统计代写|经济统计代写经济统计代考|出生和死亡偏差的数学推导

$$g_t=\frac{x_t}{x_{t-1}}-1 .$$

$$\hat{g}t=\frac{s_t^{-}}{s{t-1}^{+}}-1 .$$

$$x_{t-1}=\frac{s_{t-1}^{+}}{\left(1-d_{t-1}\right)} .$$

$$x_t=s_t^{-}+b_t \frac{s_{t-1}^{+}}{\left(1-d_{t-1}\right)} .$$

## 有限元方法代写

tatistics-lab作为专业的留学生服务机构，多年来已为美国、英国、加拿大、澳洲等留学热门地的学生提供专业的学术服务，包括但不限于Essay代写，Assignment代写，Dissertation代写，Report代写，小组作业代写，Proposal代写，Paper代写，Presentation代写，计算机作业代写，论文修改和润色，网课代做，exam代考等等。写作范围涵盖高中，本科，研究生等海外留学全阶段，辐射金融，经济学，会计学，审计学，管理学等全球99%专业科目。写作团队既有专业英语母语作者，也有海外名校硕博留学生，每位写作老师都拥有过硬的语言能力，专业的学科背景和学术写作经验。我们承诺100%原创，100%专业，100%准时，100%满意。

## MATLAB代写

MATLAB 是一种用于技术计算的高性能语言。它将计算、可视化和编程集成在一个易于使用的环境中，其中问题和解决方案以熟悉的数学符号表示。典型用途包括：数学和计算算法开发建模、仿真和原型制作数据分析、探索和可视化科学和工程图形应用程序开发，包括图形用户界面构建MATLAB 是一个交互式系统，其基本数据元素是一个不需要维度的数组。这使您可以解决许多技术计算问题，尤其是那些具有矩阵和向量公式的问题，而只需用 C 或 Fortran 等标量非交互式语言编写程序所需的时间的一小部分。MATLAB 名称代表矩阵实验室。MATLAB 最初的编写目的是提供对由 LINPACK 和 EISPACK 项目开发的矩阵软件的轻松访问，这两个项目共同代表了矩阵计算软件的最新技术。MATLAB 经过多年的发展，得到了许多用户的投入。在大学环境中，它是数学、工程和科学入门和高级课程的标准教学工具。在工业领域，MATLAB 是高效研究、开发和分析的首选工具。MATLAB 具有一系列称为工具箱的特定于应用程序的解决方案。对于大多数 MATLAB 用户来说非常重要，工具箱允许您学习应用专业技术。工具箱是 MATLAB 函数（M 文件）的综合集合，可扩展 MATLAB 环境以解决特定类别的问题。可用工具箱的领域包括信号处理、控制系统、神经网络、模糊逻辑、小波、仿真等。

## 统计代写|经济统计代写Economic Statistics代考|ECN329

statistics-lab™ 为您的留学生涯保驾护航 在代写经济统计Economic Statistics方面已经树立了自己的口碑, 保证靠谱, 高质且原创的统计Statistics代写服务。我们的专家在代写经济统计Economic Statistics代写方面经验极为丰富，各种代写经济统计Economic Statistics相关的作业也就用不着说。

• Statistical Inference 统计推断
• Statistical Computing 统计计算
• (Generalized) Linear Models 广义线性模型
• Statistical Machine Learning 统计机器学习
• Longitudinal Data Analysis 纵向数据分析
• Foundations of Data Science 数据科学基础

## 统计代写|经济统计代写Economic Statistics代考|The Partial Government Shutdown in 2019

In December 2018 and January 2019 , heightened turmoil in global financial markets raised concerns about the pace of economic activity; as a result, policymakers were acutely focused on the incoming economic data to inform their decisions. Unfortunately, a government shutdown delayed the publication of many official statistics, including December retail sales-ordinarily one of the timeliest indicators of consumer spending-leaving policymakers with less information to assess current economic conditions.

The First Data spending index remained available during the shutdown. In contrast to the worrying signs in financial markets, the December reading from First Data indicated only a modest decline in retail spending, as shown in figure 4.9.

When the shutdown ended and Census published its first estimate of December retail sales (on February 14, a month later than usual), it showed an exceptionally large decline. At that point, however, the January First Data reading was also available, and it pointed to a solid rebound in spending. Indeed, the first Census reading for January also popped back up when it was eventually published on March 11 .

## 统计代写|经济统计代写Economic Statistics代考|Hurricanes Harvey and Irma in 2017

Another useful application of our data is for assessing the impact of severe weather events, like hurricanes. The disruptions to spending during a storm are often severe but localized and short-lived, so that the lost spending is hard to quantify with monthly national statistics where the sampling frame may be inadequate to capture geographic shocks. Moreover, policymakers ultimately care about the extent to which swings in aggregate spending reflect the effect of a large, short-run disruption like a hurricane versus a change in the underlying trend in spending.

The 2017 Atlantic hurricane season was unusually active, with 17 named storms over a three-month period. Two of these hurricanes-Harvey and Irma-were especially large and severe. On August 28, Hurricane Harvey made landfall in Texas. Historic rainfall and widespread flooding severely disrupted life in Houston, the fifth largest metropolitan area in the United States. Less than two weeks later, Hurricane Irma made landfall in South Florida after causing mass destruction in Puerto Rico, and then proceeded to track up the western coast of the state, bringing heavy rain, storm surge, and flooding to a large swath of Florida and some areas of Georgia and South Carolina. By Monday, September 11, 2017, more than 7 million US residents of Puerto Rico, Florida, Georgia, and South Carolina were without power. ${ }^{18}$ In figure $4.10$, panel A depicts the path of the two hurricanes and panel B the Google search intensity during the two storms.

Using daily, state, and MSA-level indexes, we examined the pattern of activity in the days surrounding the landfalls of Hurricanes Harvey and Irma. To quantify the size of the hurricane’s effect, we estimated the following regression specification for each affected state:
\begin{aligned} \ln (\text { Spending })=& \sum_{i=-7}^{i-14} \beta_i * H_{t-i}+\sum_{w=\text { Mon }}^{w-\text { Sun }} \delta_w * I\left(\text { Day }t=w\right) \ &+\sum{m=\text { July }}^{m-\text { Nov }} \delta_m * I\left(\text { Month }t=m\right)+T_t+\varepsilon_t . \end{aligned} The state-specific hurricane effects are captured by the coefficients on the indicator variables, $H{t-i}$, which equal one if the hurricane occurred on day $t-i$, and zero otherwise. The regression also controls for variation in spending due to the day of week, the month of year, and a linear time trend $\left(T_t\right)$. The coefficient $\beta_0$ is thus the estimated effect on ( $\left.\log \right)$ spending in that state on the day the hurricane struck.

## 统计代写|经济统计代写经济统计代考| 2019年部分政府关闭

.经济统计

2018年12月和2019年1月，全球金融市场动荡加剧，引发了人们对经济活动节奏的担忧;因此，政策制定者非常关注即将到来的经济数据，以便为他们的决策提供依据。不幸的是，政府关闭推迟了许多官方统计数据的发布，包括12月零售销售——通常是消费者支出最及时的指标之一——这使得政策制定者评估当前经济状况的信息更少

First Data支出指数在关闭期间仍然可用。与金融市场令人担忧的迹象形成对比的是，First Data的12月数据显示，零售支出仅小幅下降，如图4.9所示

## 统计代写|经济统计代写经济统计代考| 2017年飓风哈维和厄玛

2017年的大西洋飓风季异常活跃，在三个月的时间里发生了17次风暴。其中的两个飓风——哈维和厄玛——特别大，特别严重。8月28日，飓风哈维在德克萨斯州登陆。史无前例的降雨和大范围的洪水严重扰乱了美国第五大大都市休斯顿的生活。不到两个星期后，飓风“厄玛”在波多黎各造成大规模破坏后登陆南佛罗里达州，然后继续向该州西海岸移动，给佛罗里达州大片地区以及乔治亚州和南卡罗来纳州的部分地区带来暴雨、风暴潮和洪水。截至2017年9月11日(周一)，波多黎各、佛罗里达州、乔治亚州和南卡罗来纳州的700多万美国居民断电。${ }^{18}$在图$4.10$中，面板A描述了两个飓风的路径，面板B描述了两个风暴期间谷歌搜索强度

\begin{aligned} \ln (\text { Spending })=& \sum_{i=-7}^{i-14} \beta_i * H_{t-i}+\sum_{w=\text { Mon }}^{w-\text { Sun }} \delta_w * I\left(\text { Day }t=w\right) \ &+\sum{m=\text { July }}^{m-\text { Nov }} \delta_m * I\left(\text { Month }t=m\right)+T_t+\varepsilon_t . \end{aligned}特定于州的飓风影响由指标变量$H{t-i}$上的系数捕获，如果飓风发生在$t-i$日，则该系数为1，否则为0。回归还控制了支出的变化，由于一周的日子，一年的月份，和线性时间趋势$\left(T_t\right)$。因此，系数$\beta_0$是对($\left.\log \right)$在飓风袭击当天在那个州的支出的估计影响

## 有限元方法代写

tatistics-lab作为专业的留学生服务机构，多年来已为美国、英国、加拿大、澳洲等留学热门地的学生提供专业的学术服务，包括但不限于Essay代写，Assignment代写，Dissertation代写，Report代写，小组作业代写，Proposal代写，Paper代写，Presentation代写，计算机作业代写，论文修改和润色，网课代做，exam代考等等。写作范围涵盖高中，本科，研究生等海外留学全阶段，辐射金融，经济学，会计学，审计学，管理学等全球99%专业科目。写作团队既有专业英语母语作者，也有海外名校硕博留学生，每位写作老师都拥有过硬的语言能力，专业的学科背景和学术写作经验。我们承诺100%原创，100%专业，100%准时，100%满意。

## MATLAB代写

MATLAB 是一种用于技术计算的高性能语言。它将计算、可视化和编程集成在一个易于使用的环境中，其中问题和解决方案以熟悉的数学符号表示。典型用途包括：数学和计算算法开发建模、仿真和原型制作数据分析、探索和可视化科学和工程图形应用程序开发，包括图形用户界面构建MATLAB 是一个交互式系统，其基本数据元素是一个不需要维度的数组。这使您可以解决许多技术计算问题，尤其是那些具有矩阵和向量公式的问题，而只需用 C 或 Fortran 等标量非交互式语言编写程序所需的时间的一小部分。MATLAB 名称代表矩阵实验室。MATLAB 最初的编写目的是提供对由 LINPACK 和 EISPACK 项目开发的矩阵软件的轻松访问，这两个项目共同代表了矩阵计算软件的最新技术。MATLAB 经过多年的发展，得到了许多用户的投入。在大学环境中，它是数学、工程和科学入门和高级课程的标准教学工具。在工业领域，MATLAB 是高效研究、开发和分析的首选工具。MATLAB 具有一系列称为工具箱的特定于应用程序的解决方案。对于大多数 MATLAB 用户来说非常重要，工具箱允许您学习应用专业技术。工具箱是 MATLAB 函数（M 文件）的综合集合，可扩展 MATLAB 环境以解决特定类别的问题。可用工具箱的领域包括信号处理、控制系统、神经网络、模糊逻辑、小波、仿真等。

## 统计代写|经济统计代写Economic Statistics代考|ECN329

statistics-lab™ 为您的留学生涯保驾护航 在代写经济统计Economic Statistics方面已经树立了自己的口碑, 保证靠谱, 高质且原创的统计Statistics代写服务。我们的专家在代写经济统计Economic Statistics代写方面经验极为丰富，各种代写经济统计Economic Statistics相关的作业也就用不着说。

• Statistical Inference 统计推断
• Statistical Computing 统计计算
• (Generalized) Linear Models 广义线性模型
• Statistical Machine Learning 统计机器学习
• Longitudinal Data Analysis 纵向数据分析
• Foundations of Data Science 数据科学基础

## 统计代写|经济统计代写Economic Statistics代考|Reengineering the Data Architecture

The opportunities created by the ubiquitous digitization of transactions can only be realized with a new architecture for data collection. The aim is for the statistical system to use all the relevant detail provided by transactions data. There are a number of issues the new data architecture would need to address (see Jarmin 2019). These include issues of privacy, confidentiality, and value of husiness data; cost to husinesses and the statistical agencies of the new architecture; and the technical and engineering issues of building a new architecture.

There are multiple potential modes for businesses providing such data. All have advantages and disadvantages. We expect that the new architecture should support multiple approaches to providing and collecting data. The agencies will need to be flexible.

Direct feed of transaction-level data. The agencies could get transactionlevel data directly from firms and do the calculations necessary to aggregate them. This approach has already been implemented by the Australian Bureau of Statistics for its retail food price index. While the agencies should be receptive to such arrangements, it is unlikely to be practical in the US context because of unwillingness of companies to provide such granular data and the difficulty for the agencies of handling the volume of data that it would entail.

Direct feed of (detailed) aggregate measures of price, quantity, and sales via APIs. Alternatively, and probably more practical in the US context, firms (e.g., retailers) could do the calculations needed to produce detailed but aggregated measures of price, quantity, and sales that could then be transmitted to the statistical agencies. Surveys and enumerations could be replaced by APIs. The agencies – in collaboration with businesses – would have to design a large, but finite, number of APIs that would mesh with would have a substantial fixed cost, but then provide much improved data at low marginal cost.

## 统计代写|经济统计代写Economic Statistics代考|Capabilities and Mandates of the Statistical Agencies

This paper envisions a new architecture for economic statistics that would build consistent measurement of price and quantity from the ground up. Currently, the collection and aggregation of data components is spread across three agencies. Implementing the new architecture we envision undoubtedly will be a challenge. Moving away from a survey-centric form of data collection for retail prices and quantities to computing statistics from detailed transaction-level data requires an approach that would have businesses providing their data in a unified way. The institutional arrangements that fundamentally separate the collection of data on prices and quantities would need to be changed. There have long been calls for reorganizing BEA, BLS, and Census to help normalize source data access, improve efficiencies, and foster innovation. Regardless of whether the agencies are realigned or reorganized, they need to review the current structure given how the production of statis-tics is evolving. Having one agency negotiate access to transaction-level data will be difficult enough. Having multiple agencies doing so unduly burdens both businesses and the taxpayer. Importantly, under the current statistical system structure, no agency has the mandate to collect data on both price and quantities, so implementing the data architecture to measure price and quantity simultaneously is not in scope for any agency. ${ }^{31}$

There are also ditticult questions about the legal and policy structure needed to govern how statistical agencies access private data assets for statistical uses. For instance, a key question is whether companies would seek to charge for access to the type of data described above and, if so, whether the associated fees would be within the budgetary resources of the statistical agencies.

To further test, develop, and implement a solution such as we are proposing here, the statistical agencies must expand their general data science capabilities. Whether transaction level data are transmitted to the agencies or whether retailers provide intermediate calculations, an important point of focus for the statistical agencies will be not only the acquisition but the curation of new types of unstructured data. The ingestion, processing, and curation of these new sources introduces scalability concerns not present in most survey contexts. Also, negotiating access will require the agencies to hire more staff with the skills to initiate and manage business relationships with data providers.

## 有限元方法代写

tatistics-lab作为专业的留学生服务机构，多年来已为美国、英国、加拿大、澳洲等留学热门地的学生提供专业的学术服务，包括但不限于Essay代写，Assignment代写，Dissertation代写，Report代写，小组作业代写，Proposal代写，Paper代写，Presentation代写，计算机作业代写，论文修改和润色，网课代做，exam代考等等。写作范围涵盖高中，本科，研究生等海外留学全阶段，辐射金融，经济学，会计学，审计学，管理学等全球99%专业科目。写作团队既有专业英语母语作者，也有海外名校硕博留学生，每位写作老师都拥有过硬的语言能力，专业的学科背景和学术写作经验。我们承诺100%原创，100%专业，100%准时，100%满意。

## MATLAB代写

MATLAB 是一种用于技术计算的高性能语言。它将计算、可视化和编程集成在一个易于使用的环境中，其中问题和解决方案以熟悉的数学符号表示。典型用途包括：数学和计算算法开发建模、仿真和原型制作数据分析、探索和可视化科学和工程图形应用程序开发，包括图形用户界面构建MATLAB 是一个交互式系统，其基本数据元素是一个不需要维度的数组。这使您可以解决许多技术计算问题，尤其是那些具有矩阵和向量公式的问题，而只需用 C 或 Fortran 等标量非交互式语言编写程序所需的时间的一小部分。MATLAB 名称代表矩阵实验室。MATLAB 最初的编写目的是提供对由 LINPACK 和 EISPACK 项目开发的矩阵软件的轻松访问，这两个项目共同代表了矩阵计算软件的最新技术。MATLAB 经过多年的发展，得到了许多用户的投入。在大学环境中，它是数学、工程和科学入门和高级课程的标准教学工具。在工业领域，MATLAB 是高效研究、开发和分析的首选工具。MATLAB 具有一系列称为工具箱的特定于应用程序的解决方案。对于大多数 MATLAB 用户来说非常重要，工具箱允许您学习应用专业技术。工具箱是 MATLAB 函数（M 文件）的综合集合，可扩展 MATLAB 环境以解决特定类别的问题。可用工具箱的领域包括信号处理、控制系统、神经网络、模糊逻辑、小波、仿真等。

## 统计代写|经济统计代写Economic Statistics代考|ECO380

statistics-lab™ 为您的留学生涯保驾护航 在代写经济统计Economic Statistics方面已经树立了自己的口碑, 保证靠谱, 高质且原创的统计Statistics代写服务。我们的专家在代写经济统计Economic Statistics代写方面经验极为丰富，各种代写经济统计Economic Statistics相关的作业也就用不着说。

• Statistical Inference 统计推断
• Statistical Computing 统计计算
• (Generalized) Linear Models 广义线性模型
• Statistical Machine Learning 统计机器学习
• Longitudinal Data Analysis 纵向数据分析
• Foundations of Data Science 数据科学基础

## 统计代写|经济统计代写Economic Statistics代考|Quality- and Appeal- Adjusted Price Indexes

The promise of digitized data goes beyond the ability to produce internally consistent price and nominal revenue data. The item-level price and quantity data, which are often accompanied by information on item-level attributes, offer the prospect of novel approaches to quality adjustment. Currently, the BLS CPI implements hedonic quality adjustment on a relatively small share of consumer expenditures (about 5 percent). For the remaining items, a matched model approach is used with ad hoc quality adjustments when feasible (e.g., if a new model of an item has more features than a prior matched item, then an attempt is made to adjust the prices to account for the change in features). The sample of products in the CPI consumption basket is rotated every four years and no quality adjustment is made to prices when a new good enters the index due to product rotation.

The digitized data offer the possibility of accounting for the enormous product turnover observed in item-level transactions data. For the Nielsen scanner data, the quarterly rates of product entry and exit are $9.62$ percent and $9.57$ percent, respectively. By product entry and exit, we mean the entry and exit of UPCs from the data. Some of the product turnover at the UPC code level in the scanner data involves minor changes in packaging and marketing, but others represent important changes in product quality.
We consider two approaches for capturing the variation in quality in price indexes using transactions data. The first approach is based on consumer demand theory and has been developed by Redding and Weinstein (2018, 2020) who build on the earlier work by Feenstra (1994). The second approach uses hedonic methods, following the insights of Pakes $(2003,2005)$ and Erickson and Pakes (2011). While these hedonic approaches are already partly in use by BLS and BEA, the item-level transactions data offer the potential for implementing these approaches with continuously updated weights and with methods to avoid selection bias arising from product entry and exit and – equally importantly – at scale. Bajari et al. (2021) is an initial attempt to implement hedonics at scale using a rich set of product attributes. We draw out the many different issues that must be confronted for practical implementation of these modern methods by the statistical agencies. Since both methods are part of active research agendas, we emphasize that our discussion and analysis is exploratory rather than yielding ultimate guidance for implementation.

## 统计代写|经济统计代写Economic Statistics代考|The Demand Residual

The large declines in the UPI, even for product categories such as soft drinks that are not obvious hotbeds of technological innovation, raise the question of whether the implied estimates are reasonable, and if so, how best to interpret them.

Redding and Weinstein (2018) take a strong view in formulating the UPI: they treat all of the measured residual demand variation not accounted for by changing prices as reflecting changes in product appeal or quality. The UPI exactly rationalizes observed prices and expenditure shares by treating the entire error in an estimated demand system as reflecting such changes. In contrast, other approaches such as hedonics or the Feenstra (1994) approach, leave an estimated residual out of the price index calculation. Although hedonic approaches can in principle capture much of the variation from changing product quality and appeal, the $R^2$ in period-by-period hedonic regressions is typically substantially less than one. Conceptually, therefore, although both the UPI and hedonics capture time-varying quality and appeal valuations from both product turnover and continuing products, the UPI is arguably more general because it comprehensively captures the error term from the underlying demand system in the price index.

The debate over whether it is appropriate to treat the entire error term from an estimated consumer demand system as reflecting changes in product quality and appeal that affect the cost of living is very much in its infancy, however. The measured error term from the estimated demand system may reflect measurement or specification error from several sources. Specification error may reflect not only functional form but also a misspecified degree of nesting or level of aggregation. Presumably, those errors would ideally be excluded from the construction of a price index.

Another possible source of specification error relates to permitting richer adjustment dynamics in consumer demand behavior. Diffusion of product availability, diffusion of information about products, habit formation, and learning dynamics will show up in the error term from estimation of specifications of static CES demand models. A related but distinct possibility is that the underlying model of price and quantity determination should reflect dynamic decisions of the producing firms (through endogenous investments in intangible capital like customer base as well as related marketing, promotion, and distribution activity by firms). It is important to remember that the approaches being used to estimate the elasticity of substitution are jointly estimating the demand and supply system, so misspecification of either the demand or supply equations can yield specification error.

## 统计代写|经济统计代写Economic Statistics代考|The Demand Residual

UPI 的大幅下降，即使是软饮料等不是技术创新的明显温床的产品类别，也引发了隐含估计是否合理的问题，如果是，如何最好地解释它们。

Redding 和 Weinstein（2018 年）在制定 UPI 时持强烈观点：他们将所有测量的剩余需求变化视为反映了产品吸引力或质量的变化，而不是通过价格变化来解释。UPI 通过将估计需求系统中的整个误差视为反映此类变化，从而准确地使观察到的价格和支出份额合理化。相比之下，其他方法，如享乐主义或 Feenstra (1994) 方法，将估计残差排除在价格指数计算之外。尽管享乐方法原则上可以捕捉到产品质量和吸引力变化带来的大部分变化，但R2在逐周期特征回归中，通常远小于 1。因此，从概念上讲，尽管 UPI 和享乐主义都从产品周转率和持续产品中捕捉到随时间变化的质量和吸引力估值，但 UPI 可以说更普遍，因为它从价格指数的潜在需求系统中全面捕捉了误差项。

## 有限元方法代写

tatistics-lab作为专业的留学生服务机构，多年来已为美国、英国、加拿大、澳洲等留学热门地的学生提供专业的学术服务，包括但不限于Essay代写，Assignment代写，Dissertation代写，Report代写，小组作业代写，Proposal代写，Paper代写，Presentation代写，计算机作业代写，论文修改和润色，网课代做，exam代考等等。写作范围涵盖高中，本科，研究生等海外留学全阶段，辐射金融，经济学，会计学，审计学，管理学等全球99%专业科目。写作团队既有专业英语母语作者，也有海外名校硕博留学生，每位写作老师都拥有过硬的语言能力，专业的学科背景和学术写作经验。我们承诺100%原创，100%专业，100%准时，100%满意。

## MATLAB代写

MATLAB 是一种用于技术计算的高性能语言。它将计算、可视化和编程集成在一个易于使用的环境中，其中问题和解决方案以熟悉的数学符号表示。典型用途包括：数学和计算算法开发建模、仿真和原型制作数据分析、探索和可视化科学和工程图形应用程序开发，包括图形用户界面构建MATLAB 是一个交互式系统，其基本数据元素是一个不需要维度的数组。这使您可以解决许多技术计算问题，尤其是那些具有矩阵和向量公式的问题，而只需用 C 或 Fortran 等标量非交互式语言编写程序所需的时间的一小部分。MATLAB 名称代表矩阵实验室。MATLAB 最初的编写目的是提供对由 LINPACK 和 EISPACK 项目开发的矩阵软件的轻松访问，这两个项目共同代表了矩阵计算软件的最新技术。MATLAB 经过多年的发展，得到了许多用户的投入。在大学环境中，它是数学、工程和科学入门和高级课程的标准教学工具。在工业领域，MATLAB 是高效研究、开发和分析的首选工具。MATLAB 具有一系列称为工具箱的特定于应用程序的解决方案。对于大多数 MATLAB 用户来说非常重要，工具箱允许您学习应用专业技术。工具箱是 MATLAB 函数（M 文件）的综合集合，可扩展 MATLAB 环境以解决特定类别的问题。可用工具箱的领域包括信号处理、控制系统、神经网络、模糊逻辑、小波、仿真等。

## 统计代写|经济统计代写Economic Statistics代考|ECON227

statistics-lab™ 为您的留学生涯保驾护航 在代写经济统计Economic Statistics方面已经树立了自己的口碑, 保证靠谱, 高质且原创的统计Statistics代写服务。我们的专家在代写经济统计Economic Statistics代写方面经验极为丰富，各种代写经济统计Economic Statistics相关的作业也就用不着说。

• Statistical Inference 统计推断
• Statistical Computing 统计计算
• (Generalized) Linear Models 广义线性模型
• Statistical Machine Learning 统计机器学习
• Longitudinal Data Analysis 纵向数据分析
• Foundations of Data Science 数据科学基础

## 统计代写|经济统计代写Economic Statistics代考|Existing Architecture

Table $1.1$ summarizes the source data and statistics produced to measure real and nominal consumer spending. ${ }^3 \mathrm{~A}$ notable feature of the current architecture is that data collection for total retail sales (Census) and for prices (BLS) are completely independent. The consumer price index program collects prices based on (1) expenditure shares from the Consumer Expenditure Survey (BLS manages the survey and Census collects the data), (2) outlets selected based on the Telephone Point of Purchase Survey, and (3) a relatively small sample of goods at these outlets that are chosen probabilistically (via the Commodities and Services Survey). The Census Bureau collects sales data from retailers in its monthly and annual surveys. The monthly survey is voluntary and has suffered from declining response rates. In addition, the composition of the companies responding to the monthly survey can change over time, which complicates producing a consistent time series. Store-level sales data are only collected once every five years as part of the Economic Census.

Integration of nominal sales and prices by BEA is done at a high level of aggregation that is complicated by the availability of product class detail for nominal sales that is only available every five years from the Economic Census. In the intervening periods, BEA interpolates and extrapolates based on the higher frequency annual, quarterly, and monthly surveys of nominal sales by the Census Bureau. These higher frequency surveys are typically at the firm rather than establishment level. Moreover, they classify firms by major kinds of business. For example, sales from the Census Monthly Retail Trade Survey (MRTS) reflect sales from “Grocery Stores” or “Food and Beverage Stores.” Such stores (really firms) sell many items beyond food and beverages, complicating the integration of the price indexes that are available at a finer product-class detail.

This complex decentralized system implies that there is limited granularity in terms of industry or geography in key indicators such as real GDP. BEA’s GDP by industry provides series for about 100 industries, with some 4-digit (NAICS) detail in sectors like manufacturing, but more commonly 3-digit and 2-digit NAICS detail. The BEA recently released county-level GDP on a special release basis, a major accomplishment. However, this achievement required BEA to integrate disparate databases at a high level of aggregation with substantial interpolation and extrapolation. Digitized transactions data offer an alternative, building up from micro data in an internally consistent manner.

## 统计代写|经济统计代写Economic Statistics代考|Using Item- Level Transactions Data

In the results presented here, we focus on two sources of transactions data summarized to the item level. One source is Nielsen retail scanner data, which provide item-level data on expenditures and quantities at the UPC code level for over 35,000 stores, covering mostly grocery stores and some mass merchandisers. ${ }^4$ Any change in product attributes yields a new UPC code so there are no changes in product attributes within the item-level data we use. The Nielsen data cover millions of products in more than 100 detailed product groups (e.g., carbonated beverages) and more than 1,000 modules within these product groups (e.g., soft drinks is a module in carbonated beverages). While the Nielsen scanner item-level data are available weekly at the store level, our analysis aggregates the item-level data to the quarterly, national level.5 Since the weeks may split between months, we use to monthly data. The NRF calendar places complete weeks into months and controls for changes in the timing of holidays and the number of weekends per month, and we use the months to create the quarterly data used in this paper. For more than 650,000 products in a typical quarter, we measure nominal sales, total quantities, and unit prices at the item level. We use the Nielsen scanner data from 2006:1 to 2015:4. The NPD Group (NPD) ${ }^6$ data cover more than 65,000 general merchandise stores, including online retailers, and include products that are not included in the Nielson scanner data. We currently restrict ourselves to the analysis of one detailed product module: memory cards. ${ }^7$ The NPD raw data are at the item-by-store-bymonth level; NPD produces the monthly data by aggregating weekly data using the NRF calendar, as we do ourselves with the Nielsen data. Again, for our analysis we aggregate the data to the quarterly, national item level. For example, the item-level data for memory cards tracks more than 12,000 item-by-quarter observations for the 2014:1 to $2016: 4$ sample period. As with the Nielsen data, we measure nominal sales, total quantities, and unit prices at the item-level by quarter.

## 统计代写|经济统计代写Economic Statistics代考|Existing Architecture

BEA 对名义销售额和价格的整合是在高水平的聚合中完成的，由于名义销售额的产品类别详细信息的可用性而复杂化，而名义销售额的产品类别详细信息仅在经济普查中每五年提供一次。在中间期间，BEA 根据人口普查局对名义销售额的较高频率的年度、季度和月度调查进行内插和外推。这些较高频率的调查通常是在公司而不是机构层面进行的。此外，他们按主要业务类型对公司进行分类。例如，人口普查每月零售贸易调查 (MRTS) 的销售额反映了“杂货店”或“食品和饮料店”的销售额。此类商店（实际上是公司）销售食品和饮料以外的许多商品，从而使价格指数的整合变得复杂，这些价格指数可以在更精细的产品级细节中获得。

## 有限元方法代写

tatistics-lab作为专业的留学生服务机构，多年来已为美国、英国、加拿大、澳洲等留学热门地的学生提供专业的学术服务，包括但不限于Essay代写，Assignment代写，Dissertation代写，Report代写，小组作业代写，Proposal代写，Paper代写，Presentation代写，计算机作业代写，论文修改和润色，网课代做，exam代考等等。写作范围涵盖高中，本科，研究生等海外留学全阶段，辐射金融，经济学，会计学，审计学，管理学等全球99%专业科目。写作团队既有专业英语母语作者，也有海外名校硕博留学生，每位写作老师都拥有过硬的语言能力，专业的学科背景和学术写作经验。我们承诺100%原创，100%专业，100%准时，100%满意。

## MATLAB代写

MATLAB 是一种用于技术计算的高性能语言。它将计算、可视化和编程集成在一个易于使用的环境中，其中问题和解决方案以熟悉的数学符号表示。典型用途包括：数学和计算算法开发建模、仿真和原型制作数据分析、探索和可视化科学和工程图形应用程序开发，包括图形用户界面构建MATLAB 是一个交互式系统，其基本数据元素是一个不需要维度的数组。这使您可以解决许多技术计算问题，尤其是那些具有矩阵和向量公式的问题，而只需用 C 或 Fortran 等标量非交互式语言编写程序所需的时间的一小部分。MATLAB 名称代表矩阵实验室。MATLAB 最初的编写目的是提供对由 LINPACK 和 EISPACK 项目开发的矩阵软件的轻松访问，这两个项目共同代表了矩阵计算软件的最新技术。MATLAB 经过多年的发展，得到了许多用户的投入。在大学环境中，它是数学、工程和科学入门和高级课程的标准教学工具。在工业领域，MATLAB 是高效研究、开发和分析的首选工具。MATLAB 具有一系列称为工具箱的特定于应用程序的解决方案。对于大多数 MATLAB 用户来说非常重要，工具箱允许您学习应用专业技术。工具箱是 MATLAB 函数（M 文件）的综合集合，可扩展 MATLAB 环境以解决特定类别的问题。可用工具箱的领域包括信号处理、控制系统、神经网络、模糊逻辑、小波、仿真等。