Benchmark Revisions Research Articles

Improved statistical consistency: The effect of data revisions on the energy use gap between China and its provinces

National energy consumption data are typically adjusted upward and provincial revision downward based on the Third Economic Census, thus substantially reducing energy statistical discrepancy between China and its provinces. Examining revision effects can facilitate better understanding whether and how the revision process eliminated the fundamental roots of the discrepancy. Through multiple index decomposition methods, this study traced the overall discrepancy to raw coal consumption by energy source, coal washing by transforming process and industrial consumption by economic sector. The reduction of the discrepancy was primarily based on improvements in the data consistency for raw coal, including the input by coal washing and industrial raw coal consumption. At provincial level, unexpected operations were detected during benchmark revisions, particularly inner-provincial inconsistency and inter-provincial incoordination. We conclude that the data discrepancy was institutionally rooted and cannot be solved by benchmark revision alone. Shifting the statistical system from a hierarchical regime to a more centralized one, which would allow greater coordination during energy data accounting and revisions, can help resolve this issue fundamentally.

Predicting benchmarked US state employment data in real time

US payroll employment data come from a survey and are subject to revisions. While revisions are generally small at the national level, they can be large enough at the state level to alter assessments of current economic conditions. Users must therefore exercise caution in interpreting state employment data until they are “benchmarked” against administrative data 5–16 months after the reference period. This article develops a state-space model that predicts benchmarked state employment data in real time. The model has two distinct features: (1) an explicit model of the data revision process and (2) a dynamic factor model that incorporates real-time information from other state-level labor market indicators. We find that the model reduces the average size of benchmark revisions by about 11 percent. When we optimally average the model’s predictions with those of existing models, the model reduces the average size of the revisions by about 14 percent.

Months for benchmark dominance: A new accuracy measure for state employment data

Employment growth is one of the most commonly used measures of the economic health of U.S. states. While the CES jobs data is timely and available by industry, for many states the early estimates are subject to large revisions, significantly reducing the usefulness of the series. Much of the error, however, cancels out when taking growth over several months. In this paper we evaluate the benchmark revisions by state and produce a quality guidepost that determines the months of growth needed to be calculated such that the early estimates are more representative of the benchmark growth than due to error.

The statistical reconciliation of time series of accounts between two benchmark revisions

The 2003–2007 U.S. annual input–output accounts, GDP‐by‐industry accounts, and expenditure‐based GDP are reconciled with the 2002 and 2007 quinquennial benchmarks and all contemporaneous constraints of the input–output accounts for the in‐between years. The reconciliation is performed at a very detailed level (six‐digit NAICS) according to feasible statistical procedures able to deal with very large systems of accounts. Our objective is to adjust the preliminary levels of the annual 2003–2007 series such that they are consistent with the quinquennial benchmarks available, fulfill all the accounting relationships for any given year, and show movements that are as close as possible to the preliminary information. We use a simultaneous least‐squares procedure based on the proportional first difference criterion, a well known movement preservation principle proposed by Denton. We evaluate the possible adoption of (i) a pure proportional (PROP) adjustment for small series and series with both negative and positive values that deteriorate the meaningfulness of growth rates, and (ii) a priori assumptions for groups of variables according to their different reliabilities, where this can reasonably be imposed.

Biased data revisions: Unintended consequences of China's energy-saving mandates

In China energy-saving mandates issued by the central government can trigger a trade-off between faithful implementation and kinds of strategic responses at the local level. This paper examines the biased data revision pattern of Chinese provinces under incentives created by energy-saving mandates. The data revisions conducted by provinces in the treatment group are evaluated against those carried out by their counterparts in the control group. Empirical evidence is provided that some provinces in the treatment group conducted biased data revisions by taking advantage of benchmark revisions of historical data following the 2008 Economic Census. These provinces adjusted their base-year energy consumption figure upward in seeking to reach the 2010 energy intensity reduction targets numerically. In institutional terms, strong political incentives are formed by assigning energy-saving indicators a veto power in cadre performance assessment. In a self-reported statistical regime, checks and balances are particularly weak regarding benchmark revisions of historical data. The findings of this paper suggest that the data quality of associated official statistics at the local level is undermined by energy-saving mandates. To improve the data quality of the official statistics, new checks and balances should be installed for local statistics.

Is Economics Responding to Critique? What do the UK 2015 QAA Subject Benchmarks Indicate?

The Quality Assurance Agency for Higher Education provides subject benchmarks which inform but do not determine the content of university and college academic programmes in the United Kingdom. These are revised every few years and have recently been completed in economics for the first time since the global financial crisis. Given the extensive criticism of mainstream economics since the crisis, one might anticipate the benchmark revisions to be extensive. However, this has not been the case. This article explores why this is so. The analysis may also be considered of broader significance because the conditions under which the review has occurred involve general processes that will be familiar, albeit with local variation, to heterodox economists elsewhere. In the conclusion, a more fundamental reconstruction of the benchmarks is provided. These will also be of interest as general orienting statements for a different kind of economics.

Dating Business Cycles in Historical Perspective: Evidence for Switzerland

Summary In this study we suggest a chronology of the classical business cycle in Switzerland based on dating algorithms suggested in Artis et al. (2004) and Harding and Pagan (2002). A further contribution of our study is that we determine the sensitivity of the chronology with respect to the particular GDP vintage used. For this purpose we employ a real-time database that contains 59 vintages of GDP data starting from 1997Q4 and ending in 2012Q2. We show that major changes in identified phases of the classical business cycle in Switzerland can be well traced to several benchmark revisions to the national accounts. In the absence of benchmark revisions the vintage-to-vintage variation exerts a comparatively minor effect on identified phases of the classical business cycle.

Chinese Statistics: Output Data

China is the world’s second-largest economy and its output data are being closely watched. The release of the latest GDP data by China’s National Bureau of Statistics can be felt on stock markets around the globe, and may influence a broad range of economic decisions ranging from companies’ investment strategies to monetary policy. But China’s GDP data are poorly understood. GDP in one year may be revised upward by 16.8 percent, while rural household consumption falls by 26.6 percent and government consumption rises by 41.1 percent. Data series appear to be revised every few years, even outside official benchmark revisions. The labor share in income is retrospectively revised to fall by more than 10 percent, and a few years later it is raised back to near its original level. Nominal data are retrospectively revised when an economic census newly discovers previously uncounted economic activities in the census year, but real growth rates are not retrospectively revised, implying a revised deflator for periods many years earlier. Data series end while series with similarly sounding labels newly start. The list is long. But explanations of what is happening to the data are almost always missing. This manuscript first specifies what official data are available for: GDP, including sectoral value-added, with a subsector breakdown for the tertiary sector as well as for industry (in the latter case limited to certain types of enterprises); aggregate expenditures, including the different types of expenditures, and national income, including income components. It then proceeds to examine the quality of these data: it explains statistical breaks, resolves inconsistencies to the extent possible, and describes the limitations of different series. Third, it recommends particular output series for use in economic analysis and provides these data.

FROM INPUT–OUTPUT TABLES TO SUPPLY-AND-USE TABLES

In 1991, Statistics Netherlands introduced the supply-and-use tables as part of the national accounts. Since then, the supply-and-use tables have been the main statistics on the production structure of the Dutch economy. They form the basis from which input–output tables are derived. The time series of supply-and-use tables starts in 1987. However, there is a need for a time series since 1970 because benchmark revisions of the Dutch national accounts would become far easier if such time series were available. Therefore, a method has been developed to derive supply-and-use tables from existing input–output tables. This article presents the algorithm.

Forecasting U.S. Output Growth with Non-Linear Models in the Presence of Data Uncertainty

We consider the impact of data revisions on the forecast performance of a SETAR regime-switching model of U.S. output growth. The impact of data uncertainty in real-time forecasting will affect a model's forecast performance via the effect on the model parameter estimates as well as via the forecast being conditioned on data measured with error. We find that benchmark revisions do affect the performance of the non-linear model of the growth rate, and that the performance relative to a linear comparator deteriorates in real-time compared to a pseudo out-of-sample forecasting exercise.

Comprehensive Benchmark Revisions for The Conference Board Leading Economic Index® for the United States

Following an extensive reevaluation of existing indicators included in The Conference Board Leading Economic Index® for The United States, we propose a comprehensive revision of the composite index. In this paper we present the case for replacing three of the components and making a minor adjustment to one other component. The resulting index addresses structural changes that have occurred in the U.S. economy in the last several decades. The changes in the LEI composition include: 1) incorporating in the LEI a new Leading Credit Index (LCI) rather than real money supply (M2) starting in 1990 (real M2 remains in the index before 1990); 2) replacing the ISM Supplier Delivery Index with the ISM New Orders Index; 3) replacing the Reuters/University of Michigan Consumer Expectations Index with an equally weighted average of consumer expectations of business and economic conditions using questions from surveys conducted by Reuters/University of Michigan and The Conference Board; and 4) replacing “New Orders for (nondefense) Capital Goods” with “New Orders for (nondefense) Capital Goods excluding Aircraft.” These changes are assessed using turning point analysis, probit models and an indicator scoring system based on Markov Switching models. Real time out-of-sample forecasting exercises are used to confirm that the changes to the composition help the LEI forecast more accurately future economic conditions.

The efficiency of the benchmark revisions to the current employment statistics (CES) data

We find evidence of a cyclical and seasonal bias in the annual benchmark revisions to the monthly level of non-farm payroll employment. We propose a change to the methodology that would remove the seasonal bias and ensure that the benchmark process does not artificially influence the estimated seasonal pattern in the raw data.

Do benchmark revisions affect the consumption-to-output and investment-to-output ratios in Germany?

The balanced growth and stochastic growth theory implies stable consumption-to-output and investment-to-output ratios. It is tested by cointegration techniques for three different German data vintages. Systems cointegration tests are helpful in revealing inconsistencies across vintages. Differencing and rebasing, often used to adjust for benchmark revisions, are generally not sufficient to ensure consistent real-time macroeconomic data. Vintage transformation functions estimated by cointegrating regressions are more flexible. Empirically, the cointegrating property between consumption and output, as well as between investment and output, is often found, whereas the one-to-one relationship is mostly rejected. Moreover, the linear transformation function is helpful in describing the relation between two older final vintages. This function seems to be insufficient if the most recent data collection framework is involved.

Dealing with Benchmark Revisions in Real‐Time Data: The Case of German Production and Orders Statistics*

AbstractBenchmark revisions in non‐stationary real‐time data may adversely affect the results of regular revision analysis and the estimates of long‐run economic relationships. Cointegration analysis can reveal the nature of vintage heterogeneity and guide the adjustment of real‐time data for benchmark revisions. Affine vintage transformation functions estimated by cointegration regressions are a flexible tool, whereas differencing and rebasing work well only under certain circumstances. Inappropriate vintage transformation may cause observed revision statistics to be affected by nuisance parameters. Using real‐time data of German industrial production and orders, the econometric techniques are exemplified and the theoretical claims are examined empirically.

Evaluating Real-Time Forecasts in Real-Time

The accuracy of real-time forecasts of macroeconomic variables that are subject to revisions may crucially depend on the choice of data used to compare the forecasts against. We put forward a flexible time - varying parameter regression framework to obtain early estimates of the final value of macroeconomic variables based upon the initial data release that may be used as actuals in current forecast evaluation. We allow for structural changes in the regression parameters to accommodate benchmark revisions and definitional changes, which fundamentally change the statistical properties of the variable of interest, including the relationship between the final value and the initial release. The usefulness of our approach is demonstrated through an empirical application comparing the accuracy of forecasts of US GDP growth rates from the Survey of Professional Forecasters and the Greenbook.

Mismeasured Personal Saving and the Permanent Income Hypothesis

Is it possible to forecast using poorly measured data? According to the permanent income hypothesis, a low personal saving rate should predict rising future income (Campbell, 1987). However, the U.S. personal saving rate is initially poorly measured and has been repeatedly revised upward in benchmark revisions. The authors use both conventional and real-time estimates of the personal saving rate in vector autoregressions to forecast real disposable income; using the level of the personal saving rate in real time would have almost invariably made forecasts worse, but first differences of the personal saving rate are predictive. They also test the lay hypothesis that a low personal saving rate has implications for consumption growth and find no evidence of forecasting ability.

How to Treat Benchmark Revisions? The Case of German Production and Orders Statistics

Elements of an econometric examination of benchmark revisions in real-time data are suggested.

Benchmark Revisions and the U.S. Personal Saving Rate

Initially published estimates of the personal saving rate from 1965 Q3 to 1999 Q2, which averaged 5.3 percent, have been revised up 2.8 percentage points to 8.1 percent, as we document. We show that much of the initial variation in the personal saving rate across time was meaningless noise. Nominal disposable personal income has been revised upward an average of 8.4 percent: one dollar in 12 was originally missing! We use both conventional and real-time estimates of the personal saving rate to forecast real disposable income, gross domestic product, and personal consumption and show that the personal saving rate in real-time almost invariably makes forecasts worse. Thus, while the personal saving rate may have some forecasting power once we know the true saving rate, as Campbell (1987) and Ireland (1995) have argued, as a practical matter it is useless to forecasters.

Macroeconometric models and changes in measurement concepts: an overview

This paper introduces the latest report of the Seminar on Model Comparisons, being a set of articles that describe subsequent effects on representative US macroeconometric models of the most recent benchmark revisions in the US National Income and Product Accounts. These accounts state the definitional relationships between measured macroeconomic variables. They also provide an evolving view of the characteristics of the US economy, this evolution occuring as the result of occasional changes in the accounting constructs. In some cases, the changes simply reflect the effect of differences in the availability of the underlying primary measurements. However, in other cases, they are made in order to bring the accounting concepts into better conformity with economic concepts, or for the sake of harmonizing the US accounts with those of other countries. In all cases, such changes may nevertheless affect the particular use of these accounts, implying the need for users to make a careful evaluation of the implications.

Comparing Time‐Series and Survey Forecasts of Weekly Changes in Money: A Methodological Note

measured as the difference between the announced change and the expected change derived from the Money Market Services, Inc. weekly survey of government securities dealers. Unlike other survey measures of expectations (the most notable example of which is the Livingston survey data), these expectations are formed by active market participants.' Consequently, the survey's forecast, made public before the Fed's announcement of the actual change in the money stock, is commonly thought to represent the market's expectation. The concern in this paper is not with the impact of unanticipated money changes on nominal interest rates. The purpose is to compare the relative forecasting abilities of the survey predictions and those obtained from a timeseries model. This question is of interest, because the weekly money series used in previous research includes certain weeks in which exogenous factors may significantly alter the series' stochastic properties (see, for example, Cornell [1], Urich and Wachtel [9]). Thus, the influence of these abnormal weeks-weeks that include benchmark revisions, changes in seasonal factors, and weeks in which there is a social security effect-is examined by comparing forecasting results for two samples; one includes these weeks and the other deletes them. Following Pearce [6], this approach allows us to test the hypothesis that the survey responses are at least weak-form rational and to see how sensitive this test is to the sample of weeks used. Unlike previous studies, we use two survey forecast series: the median forecast, made public prior to the announced change in Ml, and the mean series, constructed from individual respondents' forecasts. Zarnowitz's [10] examination of the ASA-NBER survey forecasts indicates that the mean forecast is, on average, statistically superior to the median. This finding is based on the idea that the mean forecast uses a more informationally diverse data set. Consequently, it is