Experimental statistics Labor market indicator: LinkedIn Hiring Rate


Current data of the indicator is available from the Economic Pulse Monitor (German only).

LinkedIn Hiring Rate: Background

Since the outbreak of the Corona pandemic in the spring of 2020, and continuing with the Russian war of aggression in Ukraine, social and political interest in readily available economic indicators has increased noticeably. In times of crisis, the focus is often on the impact on employment. In this context, the available, traditional labor market leading indicators often only depict the willingness of companies to hire new employees (e.g. IFO employment barometer, indicators from job advertisements, etc.). In times of a shortage of skilled workers and excess demand on the labor market, however, an indicator is needed that also takes into account the realized labor supply of people in Germany.

In cooperation with the world’s largest digital network for professionals LinkedIn, the Federal Statistical Office has developed labor market quick indicators based on statistically recorded activities on the LinkedIn platform. The project also aims to quantify the extent to which activities on the platform -which has more than 19 million active members in Germany, Austria and Switzerland- can reflect current trends on the labor market at an early stage. The underlying concept for the indicator is the so-called LinkedIn Hiring Rate (LHR). It represents what proportion of the network's active users add a new, current employer on their profile within a given month. In addition, the indicator is normalized to a base year (2016):

Such an indicator can potentially provide very timely information about new employment relationships. In addition, it can theoretically be calculated according to a wealth of characteristics, each of which is stored in LinkedIn profiles. These include, for example, the economic sector of the activity or the federal state. However, an initial analysis also shows that the calculation for the German labor market only makes sense on a monthly basis, as the vast majority of all employment relationships in Germany are started on the first of each month. A weekly or even daily representation is therefore of very little value.

With regard to interpretation, it is important to note that the hiring rate is a measure of entry into new employment, i.e. a flow variable. For example, it only indirectly tells us something about the absolute number of employed persons. A rising trend indicates an increase in the number of newly employed persons, but does not tell us anything about people leaving the labor force. A decreasing trend, on the other hand, does not necessarily indicate a decrease in employment, but initially only indicates that there are fewer new employed persons. Since no information on the number of people leaving the labor force can be obtained, the indicator cannot be used to derive any statements on the absolute change in the number of employed persons.

Figure 1:


A graphical representation of the indicator for all of Germany for the years 2016 to January 2023 (Figure 1) shows a clear seasonal figure, a dip in the first corona wave in 2020, and otherwise a positive trend for the years 2016 to 2021. Since the spring of 2022, there are signs of a trend reversal (which would imply a slowdown in the growth of entered employment).

Comparison with the actual transition to employment subject to social insurance contributions

In order to better assess the quality of the indicator, comparable data that are representative for Germany are needed. A special dataset available from the German Federal Employment Agency includes all transitions into employment subject to social security contributions, both from unemployment, another gainful activity - subject to social security contributions or not - and from inactivity. Therefore, these figures are very close to the true value that the LHR would ideally portray. Since they are based on social insurance data and thus represent a complete survey, they can also be regarded as a reference in qualitative terms. In terms of timeliness, however, this evaluation is only available with a lag of around 6 months and is therefore not suitable as an early indicator.

Figure 2:


Figure 2 shows the data of the Federal Employment Agency since 2010, also standardized to 2016, in comparison with the LinkedIn Hiring Rate, which is available for the years starting in 2016. The graph suggests that the two statistics are correlated and show a similar season. However, while the comparative data for 2010 to 2021 also shows a positive trend, it is much weaker than in the LinkedIn data. One possible reason for the discrepancy could be that LinkedIn users in Germany do not represent a random sample from the labor force, but are particularly active in the labor market and change jobs more frequently than the average working person.

When comparing in terms of month-on-month changes, the trend only comes into play as a minimal level difference. This makes it possible to show more precisely the extent to which the underlying dynamics of LinkedIn profile changes match the reference data. Figure 3 confirms a promising correlation and therefore the potential to derive short-term statements about the labor market from the LinkedIn indicator (correlation coefficient=0.84, p<0.00).

Figure 3:


To better understand what information is included in the indicator and how the sample of users is biased, a comparison by economic sector is a promising approach1. Since the difference between the Hiring Rate and the reference data is primarily due to the divergent trend, this is a logical first starting point for an analysis by industry (Figure 4). Indeed, the trend is more pronounced in most sectors of the economy in the LHR. However, the degree to which the trend is due to biases in the selection of LinkedIn users and the degree to which it is due to other effects associated with the use of the platform cannot be conclusively clarified.

Figure 4:


The comparison of the correlation of the previous month's changes by economic sector (figure 5) shows that the LHR for information-intensive service sectors reflects the underlying movements on the labor market particularly well. The indicator therefore is particularly useful in these sectors. By contrast, it performs comparatively poorly for sectors such as construction, mining and education. However, the direct graphical comparison also shows that the LinkedIn data in the least correlated area of education and teaching (WZ section P) seem to say something about the actual development of new hires (figure 6). The correlation in the most highly correlated economic sector J (information and communication) is, of course, even more convincing in comparison.

Figure 5:


Figure 6:


The LinkedIn data represent some economic sectors better than others. Inaccuracies in the overall economic view presumably stem to a large extent from those economic sectors that are less representatively represented on LinkedIn. Nonetheless, it seems clear that the data contain measurable information about actual hiring. The question is how best to use this information.

Depiction in terms of the labor-market cycle (Hodrick Prescott-Filter)

One possibility is to disregard the long-term uptrend altogether in view of the only very flat long-term trend in the reference data. Medium-term deviations from the long-term trend can be interpreted as cyclical movements and used as a quick indicator. To extract this cyclical information in the best possible way, a transformation using the so-called Hodrick-Prescott filter is useful2. This involves splitting the seasonally adjusted time series into a trend and a cycle component, with the latter representing only the business cycle and other short-term effects (see Fig. 7). The business cycle is usually defined as the multi-year cyclical fluctuation of series values around a long-term development path (trend) and presented as a percentage deviation from the latter: A value of 7, for example, indicates a deviation from the long-term trend of +7% and thus unusually many new hires. This representation corresponds to that in e.g. the Business Cycle Monitor of the Federal Statistical Office.

Figure 7:


Figure 8 shows that the cyclical representation constructed in this way approximates well the actual cyclicality of new hires resulting from the reference data. However, the average deviation from the actual value is still 3.4 percentage points. Above all, however, cyclical turning points seem to be well recognized - as far as this is possible to assess in the period under consideration so far. This also applies to most industries.

Figure 8:


In summary, the LinkedIn Hiring Rate can provide usable information about new hires and thus the labor market in Germany at an early stage. Since this is "naturally occurring", non-representative data from a social network whose use is entirely voluntary, the data exhibits biases and depicts certain areas of the labor market better than others. Nevertheless, the LinkedIn indicator is a useful addition to the indicators available so far, also because, unlike most others, it depicts the confluence of labor supply and demand.

Since the number of users of the LinkedIn platform as well as its usage is constantly increasing, it can be assumed that this first assessment of the data cannot be considered final and that the quality of the LHR will tend to increase. Also, there is certainly more data from the social network, for example on job search, which could potentially contribute useful information on the labor market.

Overall, the indicator has the potential to be offered both in the experimental data offering and in the “pulse monitor for the economy” (part of the Germany dashboard), as well as to support the acceleration of the official provision of results for employment figures.


1: For the LinkedIn data, jobs are assigned to economic sectors with the help of the respective employers specified and their profiles also stored on the platform.

2: See Robert J. Hodrick und Edward C. Prescott (1997): Postwar business cycles: an empirical investigation in Journal of Money, Credit and Banking, Vol. 29 Nr. 1.


More information on the topic from official statistics