دانلود مقاله پیش‌بینی تغییرات حقوق و دستمزد غیرکشاورزی در سال‌های 2008 و 2020 و تأثیر شوک کرونا
خلاصه
1. مقدمه
2. بررسی ادبیات
3. داده ها
4. روش شناسی
5. یافته ها
6. سخنان پایانی
اعلامیه منافع رقابتی
پیوست A
ضمیمه B
پیوست C
منابع
Abstract
1. Introduction
2. Literature review
3. Data
4. Methodology
5. Findings
6. Concluding remarks
Declaration of Competing Interest
Appendix A
Appendix B
Appendix C
References
چکیده
ما یک پنل نامتعادل از پیش‌بینی‌های ماهانه تغییرات حقوق و دستمزد غیرکشاورزی (NFP) بین ژانویه 2008 و دسامبر 2020 را که منبع آن از بلومبرگ است، تجزیه و تحلیل می‌کنیم. جای تعجب نیست که ما متوجه می شویم که کیفیت پیش بینی در بین اقتصاددانان متفاوت است و ما فرضیه توانایی پیش بینی برابر را رد می کنیم. در تجزیه خطا، شواهدی از پیش‌بینی‌های مغرضانه قابل‌توجهی پیدا می‌کنیم. میزان مشارکت در نظرسنجی بر این سوگیری تأثیر می گذارد. ما متوجه شدیم که شرکت‌کنندگان در نظرسنجی از دست دادن شغل در مواقع آشفتگی بازار را کمتر پیش‌بینی می‌کنند و در عین حال بهبودی پس از آن، به‌ویژه در طول شوک کارگری COVID19 را کمتر پیش‌بینی می‌کنند. برای پیش‌بینی تغییرات NFP، مدل‌های اتورگرسیو با یک شبکه حافظه کوتاه‌مدت یادگیری عمیق بهتر عمل می‌کنند. با این حال، پیش‌بینی اجماع، پیش‌بینی‌های بهتری را نسبت به رویکردهای مبتنی بر مدل به دست می‌دهد و با ترکیب پیش‌بینی‌های بهترین اقتصاددانان با عملکرد بیشتر بهبود می‌یابد. شوک کار COVID19 اثرات نامطلوبی بر عملکرد پیش‌بینی اقتصاددانان دارد. با این حال، همه اقتصاددانان به یک اندازه تحت تأثیر قرار نمی گیرند.
توجه! این متن ترجمه ماشینی بوده و توسط مترجمین ای ترجمه، ترجمه نشده است.
Abstract
We analyze an unbalanced panel of monthly predictions of nonfarm payroll (NFP) changes between January 2008 and December 2020 sourced from Bloomberg. Unsurprisingly, we find that prediction quality varies across economists and we reject the hypothesis of equal predictive ability. In an error decomposition, we find evidence of significantly biased forecasts. Participation rate in the survey is affecting this bias. We find that survey participants under-predict job losses in times of market turmoil while also under-predicting the recovery thereafter, especially during the COVID19 labor shock. For prediction of NFP changes, autoregressive models are outperformed by a deep learning long short-term memory network. However, the consensus forecast yields better forecasts than model-based approaches and are further improved by combining the forecasts of the best performing economists. The COVID19 labor shock is shown to have adverse effects on the prediction performance of economists. However, not all economists are affected equally.
Introduction
Nonfarm payroll (NFP) figures and monthly changes thereof are important and immediate indicators of the development of the economy in the U.S., particularly the labor market itself. Published by the Bureau of Labor Statistics (BLS) on a monthly basis, nonfarm payroll represents the number of payroll jobs and its month-to-month changes. The NFP covers most of the non-agricultural industry contributing roughly 80% of the GDP. As such, the monthly development in the labor market is an important precursor to the development and publication of other macroeconomic variables. Monthly NFP releases cause short- and medium term reactions to stock, bond, and FX markets which is documented in literature (Fleming, Remolona, 1999, Dungey, McKenzie, Smith, 2009, Dungey, Hvozdyk, 2012). The released numbers are perceived with a signaling effect, in particular when released numbers exceed or fall short of (market) expectations. Measuring and correctly quantifying these expectations—as for any micro- or macroeconomic variable—are of relevance in view of their impact and more importantly, their economic implications.
Concluding remarks
We analyze an unbalanced panel of nonfarm payroll predictions from January 2008 to December 2020 from 181 forecasters. Based on the framework of Davies and Lahiri (1995), we decompose the forecasting error of each forecaster into three components, of which two are further studied. Firstly, we focus on the temporal shock component that affects all forecasters equally per forecasting period. These shocks, a general over- or under-prediction of all forecasters for a particular month represents a news effect where an under-prediction of job increases is considered a positive shock and vice versa. From these estimated shocks, we find that the sample of predicting economists under-estimate job losses in times of prolonged market turmoil. In addition, recovery phases are under-predicted as well, leading to positive shocks.
In general, we find that the mean predictions are rather stable, causing the shock estimate to alternate regularly. Secondly, we focus on the individual bias, which describes a systematic over- or under-prediction of a particular forecaster. We find the bias of several forecasters to be statistically significant. More importantly, we find that with increasing participation rate, the individual bias is decreasing, yielding a lower prediction error. This indicates that economists that regularly make predictions are incorporating differing information sets than those with very few predictions. If we decompose the forecast errors based on a more precise measure for job market figures, the most recent publication, we observe a downward shift and a generally negative bias, underlining a tendency to under-predict true or more precise values of NFP changes. This suggests that forecasters make limited use of subsequent revisions of NFP changes and their focus remains on the initial and preliminary numbers. In view of the applied framework, we find that the impact of these revisions affects the temporal shock to a lesser extent than the individual bias.